Merge pull request #791 from JaneLiuL/master

Bump to k8s 1.24.0

.gitignore

@@ -4,3 +4,4 @@ vendordiff.patch
 .idea/
 *.code-workspace
 .vscode/
+kind

Dockerfile

@@ -11,7 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-FROM golang:1.16.1
+FROM golang:1.17.7
 
 WORKDIR /go/src/sigs.k8s.io/descheduler
 COPY . .

Makefile

@@ -24,8 +24,8 @@ ARCHS = amd64 arm arm64
 
 LDFLAGS=-ldflags "-X ${LDFLAG_LOCATION}.version=${VERSION} -X ${LDFLAG_LOCATION}.buildDate=${BUILD} -X ${LDFLAG_LOCATION}.gitbranch=${BRANCH} -X ${LDFLAG_LOCATION}.gitsha1=${SHA1}"
 
-GOLANGCI_VERSION := v1.30.0
-HAS_GOLANGCI := $(shell ls _output/bin/golangci-lint)
+GOLANGCI_VERSION := v1.46.1
+HAS_GOLANGCI := $(shell ls _output/bin/golangci-lint 2> /dev/null)
 
 # REGISTRY is the container registry to push
 # into. The default is to push to the staging
@@ -43,7 +43,7 @@ IMAGE_GCLOUD:=$(REGISTRY)/descheduler:$(VERSION)
 # In the future binaries can be uploaded to
 # GCS bucket gs://k8s-staging-descheduler.
 
-HAS_HELM := $(shell which helm)
+HAS_HELM := $(shell which helm 2> /dev/null)
 
 all: build
 
@@ -127,18 +127,21 @@ gen:
 verify-gen:
 	./hack/verify-conversions.sh
 	./hack/verify-deep-copies.sh
+	./hack/verify-defaulters.sh
 
 lint:
 ifndef HAS_GOLANGCI
-	curl -sfL https://install.goreleaser.com/github.com/golangci/golangci-lint.sh | sh -s -- -b ./_output/bin ${GOLANGCI_VERSION}
+	curl -sSfL https://raw.githubusercontent.com/golangci/golangci-lint/master/install.sh | sh -s -- -b ./_output/bin ${GOLANGCI_VERSION}
 endif
 	./_output/bin/golangci-lint run
 
-lint-chart:
+lint-chart: ensure-helm-install
+	helm lint ./charts/descheduler
+
+test-helm: ensure-helm-install
+	./test/run-helm-tests.sh
+
+ensure-helm-install:
 ifndef HAS_HELM
 	curl -fsSL -o get_helm.sh https://raw.githubusercontent.com/helm/helm/master/scripts/get-helm-3 && chmod 700 ./get_helm.sh && ./get_helm.sh
 endif
-	helm lint ./charts/descheduler
-
-test-helm:
-	./test/run-helm-tests.sh

OWNERS

@@ -10,6 +10,8 @@ reviewers:
 - seanmalloy
 - ingvagabund
 - lixiang233
+- a7i
+- janeliul
 emeritus_approvers:
 - aveshagarwal
 - k82cn

README.md

@@ -1,4 +1,4 @@
-[![Go Report Card](https://goreportcard.com/badge/kubernetes-sigs/descheduler)](https://goreportcard.com/report/sigs.k8s.io/descheduler)
+[![Go Report Card](https://goreportcard.com/badge/sigs.k8s.io/descheduler)](https://goreportcard.com/report/sigs.k8s.io/descheduler)
 ![Release Charts](https://github.com/kubernetes-sigs/descheduler/workflows/Release%20Charts/badge.svg)
 
 # Descheduler for Kubernetes
@@ -42,6 +42,7 @@ Table of Contents
   - [RemovePodsViolatingTopologySpreadConstraint](#removepodsviolatingtopologyspreadconstraint)
   - [RemovePodsHavingTooManyRestarts](#removepodshavingtoomanyrestarts)
   - [PodLifeTime](#podlifetime)
+  - [RemoveFailedPods](#removefailedpods)
 - [Filter Pods](#filter-pods)
   - [Namespace filtering](#namespace-filtering)
   - [Priority filtering](#priority-filtering)
@@ -49,6 +50,8 @@ Table of Contents
   - [Node Fit filtering](#node-fit-filtering)
 - [Pod Evictions](#pod-evictions)
   - [Pod Disruption Budget (PDB)](#pod-disruption-budget-pdb)
+- [High Availability](#high-availability)
+  - [Configure HA Mode](#configure-ha-mode)
 - [Metrics](#metrics)
 - [Compatibility Matrix](#compatibility-matrix)
 - [Getting Involved and Contributing](#getting-involved-and-contributing)
@@ -102,17 +105,17 @@ See the [resources | Kustomize](https://kubectl.docs.kubernetes.io/references/ku
 
 Run As A Job
 ```
-kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/job?ref=v0.21.0' | kubectl apply -f -
+kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/job?ref=v0.24.0' | kubectl apply -f -
 ```
 
 Run As A CronJob
 ```
-kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/cronjob?ref=v0.21.0' | kubectl apply -f -
+kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/cronjob?ref=v0.24.0' | kubectl apply -f -
 ```
 
 Run As A Deployment
 ```
-kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/deployment?ref=v0.21.0' | kubectl apply -f -
+kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/deployment?ref=v0.24.0' | kubectl apply -f -
 ```
 
 ## User Guide
@@ -121,36 +124,29 @@ See the [user guide](docs/user-guide.md) in the `/docs` directory.
 
 ## Policy and Strategies
 
-Descheduler's policy is configurable and includes strategies that can be enabled or disabled.
-Nine strategies
-1. `RemoveDuplicates`
-2. `LowNodeUtilization`
-3. `HighNodeUtilization`
-4. `RemovePodsViolatingInterPodAntiAffinity`
-5. `RemovePodsViolatingNodeAffinity`
-6. `RemovePodsViolatingNodeTaints`
-7. `RemovePodsViolatingTopologySpreadConstraint`
-8. `RemovePodsHavingTooManyRestarts`
-9. `PodLifeTime`  
-are currently implemented. As part of the policy, the
-parameters associated with the strategies can be configured too. By default, all strategies are enabled.
+Descheduler's policy is configurable and includes strategies that can be enabled or disabled. By default, all strategies are enabled.
 
-The following diagram provides a visualization of most of the strategies to help
-categorize how strategies fit together.
+The policy includes a common configuration that applies to all the strategies:
+| Name | Default Value | Description |
+|------|---------------|-------------|
+| `nodeSelector` | `nil` | limiting the nodes which are processed |
+| `evictLocalStoragePods` | `false` | allows eviction of pods with local storage |
+| `evictSystemCriticalPods` | `false` | [Warning: Will evict Kubernetes system pods] allows eviction of pods with any priority, including system pods like kube-dns |
+| `ignorePvcPods` | `false` | set whether PVC pods should be evicted or ignored |
+| `maxNoOfPodsToEvictPerNode` | `nil` | maximum number of pods evicted from each node (summed through all strategies) |
+| `maxNoOfPodsToEvictPerNamespace` | `nil` | maximum number of pods evicted from each namespace (summed through all strategies) |
+| `evictFailedBarePods` | `false` | allow eviction of pods without owner references and in failed phase |
 
-![Strategies diagram](strategies_diagram.png)
+As part of the policy, the parameters associated with each strategy can be configured.
+See each strategy for details on available parameters.
 
-The policy also includes common configuration for all the strategies:
-- `nodeSelector` - limiting the nodes which are processed
-- `evictLocalStoragePods` - allows eviction of pods with local storage
-- `evictSystemCriticalPods` - [Warning: Will evict Kubernetes system pods] allows eviction of pods with any priority, including system pods like kube-dns
-- `ignorePvcPods` - set whether PVC pods should be evicted or ignored (defaults to `false`)
-- `maxNoOfPodsToEvictPerNode` - maximum number of pods evicted from each node (summed through all strategies)
+**Policy:**
 
 ```yaml
 apiVersion: "descheduler/v1alpha1"
 kind: "DeschedulerPolicy"
 nodeSelector: prod=dev
+evictFailedBarePods: false
 evictLocalStoragePods: true
 evictSystemCriticalPods: true
 maxNoOfPodsToEvictPerNode: 40
@@ -159,6 +155,11 @@ strategies:
   ...
 ```
 
+The following diagram provides a visualization of most of the strategies to help
+categorize how strategies fit together.
+
+![Strategies diagram](strategies_diagram.png)
+
 ### RemoveDuplicates
 
 This strategy makes sure that there is only one pod associated with a ReplicaSet (RS),
@@ -220,6 +221,17 @@ These thresholds, `thresholds` and `targetThresholds`, could be tuned as per you
 strategy evicts pods from `overutilized nodes` (those with usage above `targetThresholds`) to `underutilized nodes`
 (those with usage below `thresholds`), it will abort if any number of `underutilized nodes` or `overutilized nodes` is zero.
 
+Additionally, the strategy accepts a `useDeviationThresholds` parameter.
+If that parameter is set to `true`, the thresholds are considered as percentage deviations from mean resource usage.
+`thresholds` will be deducted from the mean among all nodes and `targetThresholds` will be added to the mean.
+A resource consumption above (resp. below) this window is considered as overutilization (resp. underutilization).
+
+**NOTE:** Node resource consumption is determined by the requests and limits of pods, not actual usage.
+This approach is chosen in order to maintain consistency with the kube-scheduler, which follows the same
+design for scheduling pods onto nodes. This means that resource usage as reported by Kubelet (or commands
+like `kubectl top`) may differ from the calculated consumption, due to these components reporting
+actual usage metrics. Implementing metrics-based descheduling is currently TODO for the project.
+
 **Parameters:**
 
 |Name|Type|
@@ -227,6 +239,7 @@ strategy evicts pods from `overutilized nodes` (those with usage above `targetTh
 |`thresholds`|map(string:int)|
 |`targetThresholds`|map(string:int)|
 |`numberOfNodes`|int|
+|`useDeviationThresholds`|bool|
 |`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
 |`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
 |`nodeFit`|bool (see [node fit filtering](#node-fit-filtering))|
@@ -267,9 +280,11 @@ under utilized frequently or for a short period of time. By default, `numberOfNo
 
 ### HighNodeUtilization
 
-This strategy finds nodes that are under utilized and evicts pods in the hope that these pods will be scheduled compactly into fewer nodes.
-This strategy **must** be used with the
-scheduler strategy `MostRequestedPriority`. The parameters of this strategy are configured under `nodeResourceUtilizationThresholds`.
+This strategy finds nodes that are under utilized and evicts pods from the nodes in the hope that these pods will be
+scheduled compactly into fewer nodes.  Used in conjunction with node auto-scaling, this strategy is intended to help
+trigger down scaling of under utilized nodes.
+This strategy **must** be used with the scheduler scoring strategy `MostAllocated`. The parameters of this strategy are
+configured under `nodeResourceUtilizationThresholds`.
 
 The under utilization of nodes is determined by a configurable threshold `thresholds`. The threshold
 `thresholds` can be configured for cpu, memory, number of pods, and extended resources in terms of percentage. The percentage is
@@ -285,6 +300,12 @@ strategy evicts pods from `underutilized nodes` (those with usage below `thresho
 so that they can be recreated in appropriately utilized nodes.
 The strategy will abort if any number of `underutilized nodes` or `appropriately utilized nodes` is zero.
 
+**NOTE:** Node resource consumption is determined by the requests and limits of pods, not actual usage.
+This approach is chosen in order to maintain consistency with the kube-scheduler, which follows the same
+design for scheduling pods onto nodes. This means that resource usage as reported by Kubelet (or commands
+like `kubectl top`) may differ from the calculated consumption, due to these components reporting
+actual usage metrics. Implementing metrics-based descheduling is currently TODO for the project.
+
 **Parameters:**
 
 |Name|Type|
@@ -399,10 +420,17 @@ pod "podA" with a toleration to tolerate a taint ``key=value:NoSchedule`` schedu
 node. If the node's taint is subsequently updated/removed, taint is no longer satisfied by its pods' tolerations
 and will be evicted.
 
+Node taints can be excluded from consideration by specifying a list of excludedTaints. If a node taint key **or**
+key=value matches an excludedTaints entry, the taint will be ignored.
+
+For example, excludedTaints entry "dedicated" would match all taints with key "dedicated", regardless of value.
+excludedTaints entry "dedicated=special-user" would match taints with key "dedicated" and value "special-user".
+
 **Parameters:**
 
 |Name|Type|
 |---|---|
+|`excludedTaints`|list(string)|
 |`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
 |`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
 |`namespaces`|(see [namespace filtering](#namespace-filtering))|
@@ -417,6 +445,10 @@ kind: "DeschedulerPolicy"
 strategies:
   "RemovePodsViolatingNodeTaints":
     enabled: true
+    params:
+      excludedTaints:
+      - dedicated=special-user # exclude taints with key "dedicated" and value "special-user"
+      - reserved # exclude all taints with key "reserved"
 ````
 
 ### RemovePodsViolatingTopologySpreadConstraint
@@ -458,9 +490,9 @@ strategies:
 
 This strategy makes sure that pods having too many restarts are removed from nodes. For example a pod with EBS/PD that
 can't get the volume/disk attached to the instance, then the pod should be re-scheduled to other nodes. Its parameters
-include `podRestartThreshold`, which is the number of restarts at which a pod should be evicted, and `includingInitContainers`,
-which determines whether init container restarts should be factored into that calculation.
-|`labelSelector`|(see [label filtering](#label-filtering))|
+include `podRestartThreshold`, which is the number of restarts (summed over all eligible containers) at which a pod
+should be evicted, and `includingInitContainers`, which determines whether init container restarts should be factored
+into that calculation.
 
 **Parameters:**
 
@@ -471,6 +503,7 @@ which determines whether init container restarts should be factored into that ca
 |`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
 |`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
 |`namespaces`|(see [namespace filtering](#namespace-filtering))|
+|`labelSelector`|(see [label filtering](#label-filtering))|
 |`nodeFit`|bool (see [node fit filtering](#node-fit-filtering))|
 
 **Example:**
@@ -520,6 +553,47 @@ strategies:
          - "Pending"
 ```
 
+### RemoveFailedPods
+
+This strategy evicts pods that are in failed status phase.
+You can provide an optional parameter to filter by failed `reasons`.
+`reasons` can be expanded to include reasons of InitContainers as well by setting the optional parameter `includingInitContainers` to `true`.
+You can specify an optional parameter `minPodLifetimeSeconds` to evict pods that are older than specified seconds.
+Lastly, you can specify the optional parameter `excludeOwnerKinds` and if a pod
+has any of these `Kind`s listed as an `OwnerRef`, that pod will not be considered for eviction.
+
+**Parameters:**
+
+|Name|Type|
+|---|---|
+|`minPodLifetimeSeconds`|uint|
+|`excludeOwnerKinds`|list(string)|
+|`reasons`|list(string)|
+|`includingInitContainers`|bool|
+|`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
+|`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
+|`namespaces`|(see [namespace filtering](#namespace-filtering))|
+|`labelSelector`|(see [label filtering](#label-filtering))|
+|`nodeFit`|bool (see [node fit filtering](#node-fit-filtering))|
+
+**Example:**
+
+```yaml
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "RemoveFailedPods":
+     enabled: true
+     params:
+       failedPods:
+         reasons:
+         - "NodeAffinity"
+         includingInitContainers: true
+         excludeOwnerKinds:
+         - "Job"
+         minPodLifetimeSeconds: 3600
+```
+
 ## Filter Pods
 
 ### Namespace filtering
@@ -532,6 +606,7 @@ The following strategies accept a `namespaces` parameter which allows to specify
 * `RemovePodsViolatingInterPodAntiAffinity`
 * `RemoveDuplicates`
 * `RemovePodsViolatingTopologySpreadConstraint`
+* `RemoveFailedPods`
 
 For example:
 
@@ -614,7 +689,7 @@ does not exist, descheduler won't create it and will throw an error.
 
 ### Label filtering
 
-The following strategies can configure a [standard kubernetes labelSelector](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.21/#labelselector-v1-meta)
+The following strategies can configure a [standard kubernetes labelSelector](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.24/#labelselector-v1-meta)
 to filter pods by their labels:
 
 * `PodLifeTime`
@@ -623,6 +698,7 @@ to filter pods by their labels:
 * `RemovePodsViolatingNodeAffinity`
 * `RemovePodsViolatingInterPodAntiAffinity`
 * `RemovePodsViolatingTopologySpreadConstraint`
+* `RemoveFailedPods`
 
 This allows running strategies among pods the descheduler is interested in.
 
@@ -657,11 +733,13 @@ The following strategies accept a `nodeFit` boolean parameter which can optimize
 * `RemovePodsViolatingNodeTaints`
 * `RemovePodsViolatingTopologySpreadConstraint`
 * `RemovePodsHavingTooManyRestarts`
+* `RemoveFailedPods`
 
  If set to `true` the descheduler will consider whether or not the pods that meet eviction criteria will fit on other nodes before evicting them. If a pod cannot be rescheduled to another node, it will not be evicted. Currently the following criteria are considered when setting `nodeFit` to `true`:
 - A `nodeSelector` on the pod
-- Any `Tolerations` on the pod and any `Taints` on the other nodes
+- Any `tolerations` on the pod and any `taints` on the other nodes
 - `nodeAffinity` on the pod
+- Resource `requests` made by the pod and the resources available on other nodes
 - Whether any of the other nodes are marked as `unschedulable`
 
 E.g.
@@ -671,17 +749,17 @@ apiVersion: "descheduler/v1alpha1"
 kind: "DeschedulerPolicy"
 strategies:
   "LowNodeUtilization":
-     enabled: true
-     params:
-       nodeResourceUtilizationThresholds:
-         thresholds:
-           "cpu" : 20
-           "memory": 20
-           "pods": 20
-         targetThresholds:
-           "cpu" : 50
-           "memory": 50
-           "pods": 50
+    enabled: true
+    params:
+      nodeResourceUtilizationThresholds:
+        thresholds:
+          "cpu": 20
+          "memory": 20
+          "pods": 20
+        targetThresholds:
+          "cpu": 50
+          "memory": 50
+          "pods": 50
         nodeFit: true
 ```
 
@@ -697,8 +775,8 @@ Using Deployments instead of ReplicationControllers provides an automated rollou
 When the descheduler decides to evict pods from a node, it employs the following general mechanism:
 
 * [Critical pods](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/) (with priorityClassName set to system-cluster-critical or system-node-critical) are never evicted (unless `evictSystemCriticalPods: true` is set).
-* Pods (static or mirrored pods or stand alone pods) not part of an ReplicationController, ReplicaSet(Deployment), StatefulSet, or Job are
-never evicted because these pods won't be recreated.
+* Pods (static or mirrored pods or standalone pods) not part of an ReplicationController, ReplicaSet(Deployment), StatefulSet, or Job are
+never evicted because these pods won't be recreated. (Standalone pods in failed status phase can be evicted by setting `evictFailedBarePods: true`)
 * Pods associated with DaemonSets are never evicted.
 * Pods with local storage are never evicted (unless `evictLocalStoragePods: true` is set).
 * Pods with PVCs are evicted (unless `ignorePvcPods: true` is set).
@@ -707,6 +785,7 @@ best effort pods are evicted before burstable and guaranteed pods.
 * All types of pods with the annotation `descheduler.alpha.kubernetes.io/evict` are eligible for eviction. This
   annotation is used to override checks which prevent eviction and users can select which pod is evicted.
   Users should know how and if the pod will be recreated.
+* Pods with a non-nil DeletionTimestamp are not evicted by default.
 
 Setting `--v=4` or greater on the Descheduler will log all reasons why any pod is not evictable.
 
@@ -715,6 +794,23 @@ Setting `--v=4` or greater on the Descheduler will log all reasons why any pod i
 Pods subject to a Pod Disruption Budget(PDB) are not evicted if descheduling violates its PDB. The pods
 are evicted by using the eviction subresource to handle PDB.
 
+## High Availability
+
+In High Availability mode, Descheduler starts [leader election](https://github.com/kubernetes/client-go/tree/master/tools/leaderelection) process in Kubernetes. You can activate HA mode
+if you choose to deploy your application as Deployment.
+
+Deployment starts with 1 replica by default. If you want to use more than 1 replica, you must consider
+enable High Availability mode since we don't want to run descheduler pods simultaneously.
+
+### Configure HA Mode
+
+The leader election process can be enabled by setting `--leader-elect` in the CLI. You can also set
+`--set=leaderElection.enabled=true` flag if you are using Helm.
+
+To get best results from HA mode some additional configurations might require:
+* Configure a [podAntiAffinity](https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node) rule if you want to schedule onto a node only if that node is in the same zone as at least one already-running descheduler
+* Set the replica count greater than 1
+
 ## Metrics
 
 | name	| type	| description |
@@ -734,16 +830,18 @@ v0.18 should work with k8s v1.18, v1.17, and v1.16.
 Starting with descheduler release v0.18 the minor version of descheduler matches the minor version of the k8s client
 packages that it is compiled with.
 
-Descheduler  | Supported Kubernetes Version
--------------|-----------------------------
-v0.21        | v1.21
-v0.20        | v1.20
-v0.19        | v1.19
-v0.18        | v1.18
-v0.10        | v1.17
-v0.4-v0.9    | v1.9+
-v0.1-v0.3    | v1.7-v1.8
-
+| Descheduler | Supported Kubernetes Version |
+|-------------|------------------------------|
+| v0.24       | v1.24                        |
+| v0.23       | v1.23                        |
+| v0.22       | v1.22                        |
+| v0.21       | v1.21                        |
+| v0.20       | v1.20                        |
+| v0.19       | v1.19                        |
+| v0.18       | v1.18                        |
+| v0.10       | v1.17                        |
+| v0.4-v0.9   | v1.9+                        |
+| v0.1-v0.3   | v1.7-v1.8                    |
 
 ## Getting Involved and Contributing
 

charts/descheduler/Chart.yaml

@@ -1,7 +1,7 @@
 apiVersion: v1
 name: descheduler
-version: 0.21.0
-appVersion: 0.21.0
+version: 0.23.1
+appVersion: 0.23.0
 description: Descheduler for Kubernetes is used to rebalance clusters by evicting pods that can potentially be scheduled on better nodes. In the current implementation, descheduler does not schedule replacement of evicted pods but relies on the default scheduler for that.
 keywords:
 - kubernetes

charts/descheduler/README.md

@@ -43,24 +43,44 @@ The command removes all the Kubernetes components associated with the chart and
 
 The following table lists the configurable parameters of the _descheduler_ chart and their default values.
 
-| Parameter                      | Description                                                                                                           | Default                              |
-| ------------------------------ | --------------------------------------------------------------------------------------------------------------------- | ------------------------------------ |
-| `image.repository`             | Docker repository to use                                                                                              | `k8s.gcr.io/descheduler/descheduler` |
-| `image.tag`                    | Docker tag to use                                                                                                     | `v[chart appVersion]`                |
-| `image.pullPolicy`             | Docker image pull policy                                                                                              | `IfNotPresent`                       |
-| `nameOverride`                 | String to partially override `descheduler.fullname` template (will prepend the release name)                          | `""`                                 |
-| `fullnameOverride`             | String to fully override `descheduler.fullname` template                                                              | `""`                                 |
-| `cronJobApiVersion`            | CronJob API Group Version                                                                                             | `"batch/v1"`                         |
-| `schedule`                     | The cron schedule to run the _descheduler_ job on                                                                     | `"*/2 * * * *"`                      |
-| `startingDeadlineSeconds`      | If set, configure `startingDeadlineSeconds` for the _descheduler_ job                                                 | `nil`                                |
-| `successfulJobsHistoryLimit`   | If set, configure `successfulJobsHistoryLimit` for the _descheduler_ job                                              | `nil`                                |
-| `failedJobsHistoryLimit`       | If set, configure `failedJobsHistoryLimit` for the _descheduler_ job                                                  | `nil`                                |
-| `cmdOptions`                   | The options to pass to the _descheduler_ command                                                                      | _see values.yaml_                    |
-| `deschedulerPolicy.strategies` | The _descheduler_ strategies to apply                                                                                 | _see values.yaml_                    |
-| `priorityClassName`            | The name of the priority class to add to pods                                                                         | `system-cluster-critical`            |
-| `rbac.create`                  | If `true`, create & use RBAC resources                                                                                | `true`                               |
-| `podSecurityPolicy.create`     | If `true`, create PodSecurityPolicy                                                                                   | `true`                               |
-| `resources`                    | Descheduler container CPU and memory requests/limits                                                                  | _see values.yaml_                    |
-| `serviceAccount.create`        | If `true`, create a service account for the cron job                                                                  | `true`                               |
-| `serviceAccount.name`          | The name of the service account to use, if not set and create is true a name is generated using the fullname template | `nil`                                |
-| `nodeSelector`                 | Node selectors to run the descheduler cronjob on specific nodes                                                       | `nil`                                |
+| Parameter                           | Description                                                                                                           | Default                              |
+|-------------------------------------|-----------------------------------------------------------------------------------------------------------------------|--------------------------------------|
+| `kind`                              | Use as CronJob or Deployment                                                                                          | `CronJob`                            |
+| `image.repository`                  | Docker repository to use                                                                                              | `k8s.gcr.io/descheduler/descheduler` |
+| `image.tag`                         | Docker tag to use                                                                                                     | `v[chart appVersion]`                |
+| `image.pullPolicy`                  | Docker image pull policy                                                                                              | `IfNotPresent`                       |
+| `imagePullSecrets`                  | Docker repository secrets                                                                                             | `[]`                                 |
+| `nameOverride`                      | String to partially override `descheduler.fullname` template (will prepend the release name)                          | `""`                                 |
+| `fullnameOverride`                  | String to fully override `descheduler.fullname` template                                                              | `""`                                 |
+| `cronJobApiVersion`                 | CronJob API Group Version                                                                                             | `"batch/v1"`                         |
+| `schedule`                          | The cron schedule to run the _descheduler_ job on                                                                     | `"*/2 * * * *"`                      |
+| `startingDeadlineSeconds`           | If set, configure `startingDeadlineSeconds` for the _descheduler_ job                                                 | `nil`                                |
+| `successfulJobsHistoryLimit`        | If set, configure `successfulJobsHistoryLimit` for the _descheduler_ job                                              | `nil`                                |
+| `failedJobsHistoryLimit`            | If set, configure `failedJobsHistoryLimit` for the _descheduler_ job                                                  | `nil`                                |
+| `deschedulingInterval`              | If using kind:Deployment, sets time between consecutive descheduler executions.                                       | `5m`                                 |
+| `replicas`                          | The replica count for Deployment                                                                                      | `1`                                  |
+| `leaderElection`                    | The options for high availability when running replicated components                                                  | _see values.yaml_                    |
+| `cmdOptions`                        | The options to pass to the _descheduler_ command                                                                      | _see values.yaml_                    |
+| `deschedulerPolicy.strategies`      | The _descheduler_ strategies to apply                                                                                 | _see values.yaml_                    |
+| `priorityClassName`                 | The name of the priority class to add to pods                                                                         | `system-cluster-critical`            |
+| `rbac.create`                       | If `true`, create & use RBAC resources                                                                                | `true`                               |
+| `podSecurityPolicy.create`          | If `true`, create PodSecurityPolicy                                                                                   | `true`                               |
+| `resources`                         | Descheduler container CPU and memory requests/limits                                                                  | _see values.yaml_                    |
+| `serviceAccount.create`             | If `true`, create a service account for the cron job                                                                  | `true`                               |
+| `serviceAccount.name`               | The name of the service account to use, if not set and create is true a name is generated using the fullname template | `nil`                                |
+| `serviceAccount.annotations`        | Specifies custom annotations for the serviceAccount                                                                   | `{}`                                 |
+| `nodeSelector`                      | Node selectors to run the descheduler cronjob/deployment on specific nodes                                            | `nil`                                |
+| `service.enabled`                   | If `true`, create a service for deployment                                                                            | `false`                              |
+| `serviceMonitor.enabled`            | If `true`, create a ServiceMonitor for deployment                                                                     | `false`                              |
+| `serviceMonitor.namespace`          | The namespace where Prometheus expects to find service monitors                                                       | `nil`                                |
+| `serviceMonitor.interval`           | The scrape interval. If not set, the Prometheus default scrape interval is used                                       | `nil`                                |
+| `serviceMonitor.honorLabels`        | Keeps the scraped data's labels when labels are on collisions with target labels.                                     | `true`                               |
+| `serviceMonitor.insecureSkipVerify` | Skip TLS certificate validation when scraping                                                                         | `true`                               |
+| `serviceMonitor.serverName`         | Name of the server to use when validating TLS certificate                                                             | `nil`                                |
+| `serviceMonitor.metricRelabelings`  | MetricRelabelConfigs to apply to samples after scraping, but before ingestion                                         | `[]`                                 |
+| `serviceMonitor.relabelings`        | RelabelConfigs to apply to samples before scraping                                                                    | `[]`                                 |
+| `affinity`                          | Node affinity to run the descheduler cronjob/deployment on specific nodes                                             | `nil`                                |
+| `tolerations`                       | tolerations to run the descheduler cronjob/deployment on specific nodes                                               | `nil`                                |
+| `suspend`                           | Set spec.suspend in descheduler cronjob                                                                               | `false`                              |
+| `commonLabels`                      | Labels to apply to all resources                                                                                      | `{}`                                 |
+| `livenessProbe`                     | Liveness probe configuration for the descheduler container                                                            | _see values.yaml_                    |

charts/descheduler/templates/NOTES.txt

@@ -1 +1,7 @@
-Descheduler installed as a cron job.
+Descheduler installed as a {{ .Values.kind }}.
+
+{{- if eq .Values.kind "Deployment" }}
+{{- if eq .Values.replicas 1.0}}
+WARNING: You set replica count as 1 and workload kind as Deployment however leaderElection is not enabled. Consider enabling Leader Election for HA mode.
+{{- end}}
+{{- end}}

charts/descheduler/templates/_helpers.tpl

@@ -42,6 +42,17 @@ app.kubernetes.io/instance: {{ .Release.Name }}
 app.kubernetes.io/version: {{ .Chart.AppVersion | quote }}
 {{- end }}
 app.kubernetes.io/managed-by: {{ .Release.Service }}
+{{- if .Values.commonLabels}}
+{{ toYaml .Values.commonLabels }}
+{{- end }}
+{{- end -}}
+
+{{/*
+Selector labels
+*/}}
+{{- define "descheduler.selectorLabels" -}}
+app.kubernetes.io/name: {{ include "descheduler.name" . }}
+app.kubernetes.io/instance: {{ .Release.Name }}
 {{- end -}}
 
 {{/*
@@ -54,3 +65,37 @@ Create the name of the service account to use
     {{ default "default" .Values.serviceAccount.name }}
 {{- end -}}
 {{- end -}}
+
+{{/*
+Leader Election
+*/}}
+{{- define "descheduler.leaderElection"}}
+{{- if .Values.leaderElection -}}
+- --leader-elect
+- {{ default false .Values.leaderElection.enabled }}
+{{- if .Values.leaderElection.leaseDuration }}
+- --leader-elect-lease-duration
+- {{ .Values.leaderElection.leaseDuration }}
+{{- end }}
+{{- if .Values.leaderElection.renewDeadline }}
+- --leader-elect-renew-deadline
+- {{ .Values.leaderElection.renewDeadline }}
+{{- end }}
+{{- if .Values.leaderElection.retryPeriod }}
+- --leader-elect-retry-period
+- {{ .Values.leaderElection.retryPeriod }}
+{{- end }}
+{{- if .Values.leaderElection.resourceLock }}
+- --leader-elect-resource-lock
+- {{ .Values.leaderElection.resourceLock }}
+{{- end }}
+{{- if .Values.leaderElection.resourceName }}
+- --leader-elect-resource-name
+- {{ .Values.leaderElection.resourceName }}
+{{- end }}
+{{- if .Values.leaderElection.resourceNamescape }}
+- --leader-elect-resource-namespace
+- {{ .Values.leaderElection.resourceNamescape }}
+{{- end -}}
+{{- end }}
+{{- end }}

charts/descheduler/templates/clusterrole.yaml

@@ -14,7 +14,7 @@ rules:
   verbs: ["get", "watch", "list"]
 - apiGroups: [""]
   resources: ["namespaces"]
-  verbs: ["get", "list"]
+  verbs: ["get", "watch", "list"]
 - apiGroups: [""]
   resources: ["pods"]
   verbs: ["get", "watch", "list", "delete"]
@@ -24,6 +24,15 @@ rules:
 - apiGroups: ["scheduling.k8s.io"]
   resources: ["priorityclasses"]
   verbs: ["get", "watch", "list"]
+{{- if .Values.leaderElection.enabled }}
+- apiGroups: ["coordination.k8s.io"]
+  resources: ["leases"]
+  verbs: ["create"]
+- apiGroups: ["coordination.k8s.io"]
+  resources: ["leases"]
+  resourceNames: ["descheduler"]
+  verbs: ["get", "patch", "delete"]
+{{- end }}
 {{- if .Values.podSecurityPolicy.create }}
 - apiGroups: ['policy']
   resources: ['podsecuritypolicies']

charts/descheduler/templates/configmap.yaml

@@ -2,6 +2,7 @@ apiVersion: v1
 kind: ConfigMap
 metadata:
   name: {{ template "descheduler.fullname" . }}
+  namespace: {{ .Release.Namespace }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
 data:

charts/descheduler/templates/cronjob.yaml

@@ -1,11 +1,16 @@
+{{- if eq .Values.kind "CronJob" }}
 apiVersion: {{ .Values.cronJobApiVersion | default "batch/v1" }}
 kind: CronJob
 metadata:
   name: {{ template "descheduler.fullname" . }}
+  namespace: {{ .Release.Namespace }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
 spec:
   schedule: {{ .Values.schedule | quote }}
+  {{- if .Values.suspend }}
+  suspend: {{ .Values.suspend }}
+  {{- end }}
   concurrencyPolicy: "Forbid"
   {{- if .Values.startingDeadlineSeconds }}
   startingDeadlineSeconds: {{ .Values.startingDeadlineSeconds }}
@@ -27,8 +32,7 @@ spec:
             {{- .Values.podAnnotations | toYaml | nindent 12 }}
             {{- end }}
           labels:
-            app.kubernetes.io/name: {{ include "descheduler.name" . }}
-            app.kubernetes.io/instance: {{ .Release.Name }}
+            {{- include "descheduler.selectorLabels" . | nindent 12 }}
             {{- if .Values.podLabels }}
             {{- .Values.podLabels | toYaml | nindent 12 }}
             {{- end }}
@@ -37,11 +41,23 @@ spec:
           nodeSelector:
             {{- toYaml . | nindent 12 }}
           {{- end }}
+          {{- with .Values.affinity }}
+          affinity:
+            {{- toYaml . | nindent 12 }}
+          {{- end }}
+          {{- with .Values.tolerations }}
+          tolerations:
+            {{- toYaml . | nindent 12 }}
+          {{- end }}
           {{- if .Values.priorityClassName }}
           priorityClassName: {{ .Values.priorityClassName }}
           {{- end }}
           serviceAccountName: {{ template "descheduler.serviceAccountName" . }}
           restartPolicy: "Never"
+          {{- with .Values.imagePullSecrets }}
+          imagePullSecrets:
+          {{- toYaml . | nindent 10 }}
+          {{- end }}
           containers:
             - name: {{ .Chart.Name }}
               image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default (printf "v%s" .Chart.AppVersion) }}"
@@ -57,6 +73,8 @@ spec:
                 - {{ $value | quote }}
                 {{- end }}
                 {{- end }}
+              livenessProbe:
+                {{- toYaml .Values.livenessProbe | nindent 16 }}
               resources:
                 {{- toYaml .Values.resources | nindent 16 }}
               securityContext:
@@ -74,3 +92,4 @@ spec:
           - name: policy-volume
             configMap:
               name: {{ template "descheduler.fullname" . }}
+{{- end }}

charts/descheduler/templates/deployment.yaml

@@ -0,0 +1,94 @@
+{{- if eq .Values.kind "Deployment" }}
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ template "descheduler.fullname" . }}
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "descheduler.labels" . | nindent 4 }}
+spec:
+  {{- if gt .Values.replicas 1.0}}
+  {{- if not .Values.leaderElection.enabled }}
+  {{- fail "You must set leaderElection to use more than 1 replica"}}
+  {{- end}}
+  replicas: {{ required "leaderElection required for running more than one replica" .Values.replicas }}
+  {{- else }}
+  replicas: 1
+  {{- end }}
+  selector:
+    matchLabels:
+      {{- include "descheduler.selectorLabels" . | nindent 6 }}
+  template:
+    metadata:
+      labels:
+        {{- include "descheduler.selectorLabels" . | nindent 8 }}
+        {{- if .Values.podLabels }}
+        {{- .Values.podLabels | toYaml | nindent 8 }}
+        {{- end }}
+      annotations:
+        checksum/config: {{ include (print $.Template.BasePath "/configmap.yaml") . | sha256sum }}
+        {{- if .Values.podAnnotations }}
+        {{- .Values.podAnnotations | toYaml | nindent 8 }}
+        {{- end }}
+    spec:
+      {{- if .Values.priorityClassName }}
+      priorityClassName: {{ .Values.priorityClassName }}
+      {{- end }}
+      serviceAccountName: {{ template "descheduler.serviceAccountName" . }}
+      {{- with .Values.imagePullSecrets }}
+      imagePullSecrets:
+      {{- toYaml . | nindent 10 }}
+      {{- end }}
+      containers:
+        - name: {{ .Chart.Name }}
+          image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default (printf "v%s" .Chart.AppVersion) }}"
+          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          command:
+            - "/bin/descheduler"
+          args:
+            - "--policy-config-file"
+            - "/policy-dir/policy.yaml"
+            - "--descheduling-interval"
+            - {{ required "deschedulingInterval required for running as Deployment" .Values.deschedulingInterval }}
+            {{- range $key, $value := .Values.cmdOptions }}
+            - {{ printf "--%s" $key | quote }}
+            {{- if $value }}
+            - {{ $value | quote }}
+            {{- end }}
+            {{- end }}
+            {{- include "descheduler.leaderElection" . | nindent 12 }}
+          ports:
+            - containerPort: 10258
+              protocol: TCP
+          livenessProbe:
+            {{- toYaml .Values.livenessProbe | nindent 12 }}
+          resources:
+            {{- toYaml .Values.resources | nindent 12 }}
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop:
+                - ALL
+            privileged: false
+            readOnlyRootFilesystem: true
+            runAsNonRoot: true
+          volumeMounts:
+            - mountPath: /policy-dir
+              name: policy-volume
+      volumes:
+        - name: policy-volume
+          configMap:
+            name: {{ template "descheduler.fullname" . }}
+      {{- with .Values.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+{{- end }}

charts/descheduler/templates/podsecuritypolicy.yaml

@@ -3,6 +3,7 @@ apiVersion: policy/v1beta1
 kind: PodSecurityPolicy
 metadata:
   name: {{ template "descheduler.fullname" . }}
+  namespace: {{ .Release.Namespace }}
   annotations:
     seccomp.security.alpha.kubernetes.io/allowedProfileNames: 'docker/default,runtime/default'
     seccomp.security.alpha.kubernetes.io/defaultProfileName:  'runtime/default'

charts/descheduler/templates/service.yaml

@@ -0,0 +1,21 @@
+{{- if eq .Values.kind "Deployment" }}
+{{- if eq .Values.service.enabled true }}
+apiVersion: v1
+kind: Service
+metadata:
+  labels:
+    {{- include "descheduler.labels" . | nindent 4 }}
+  name: {{ template "descheduler.fullname" . }}
+  namespace: {{ .Release.Namespace }}
+spec:
+  clusterIP: None
+  ports:
+  - name: http-metrics
+    port: 10258
+    protocol: TCP
+    targetPort: 10258
+  selector:
+    {{- include "descheduler.selectorLabels" . | nindent 4 }}
+  type: ClusterIP
+{{- end }}
+{{- end }}

charts/descheduler/templates/serviceaccount.yaml

@@ -3,6 +3,10 @@ apiVersion: v1
 kind: ServiceAccount
 metadata:
   name: {{ template "descheduler.serviceAccountName" . }}
+  namespace: {{ .Release.Namespace }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
+{{- if .Values.serviceAccount.annotations }}
+  annotations: {{ toYaml .Values.serviceAccount.annotations | nindent 4 }}
+{{- end }}
 {{- end -}}

charts/descheduler/templates/servicemonitor.yaml

@@ -0,0 +1,41 @@
+{{- if eq .Values.kind "Deployment" }}
+{{- if eq .Values.serviceMonitor.enabled true }}
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: {{ template "descheduler.fullname" . }}-servicemonitor
+  namespace: {{ .Values.serviceMonitor.namespace | default .Release.Namespace }}
+  labels:
+    {{- include "descheduler.labels" . | nindent 4 }}
+spec:
+  jobLabel: jobLabel
+  namespaceSelector:
+    matchNames:
+    - {{ .Release.Namespace }}
+  selector:
+    matchLabels:
+      {{- include "descheduler.selectorLabels" . | nindent 6 }}
+  endpoints:
+  - honorLabels: {{ .Values.serviceMonitor.honorLabels | default true }}
+    port: http-metrics
+    {{- if .Values.serviceMonitor.interval }}
+    interval: {{ .Values.serviceMonitor.interval }}
+    {{- end }}
+    scheme: https
+    tlsConfig:
+      {{- if eq .Values.serviceMonitor.insecureSkipVerify true }}
+      insecureSkipVerify: true
+      {{- end }}
+      {{- if .Values.serviceMonitor.serverName }}
+      serverName: {{ .Values.serviceMonitor.serverName }}
+      {{- end}}
+{{- if .Values.serviceMonitor.metricRelabelings }}
+    metricRelabelings:
+{{ tpl (toYaml .Values.serviceMonitor.metricRelabelings | indent 4) . }}
+{{- end }}
+{{- if .Values.serviceMonitor.relabelings }}
+    relabelings:
+{{ tpl (toYaml .Values.serviceMonitor.relabelings | indent 4) . }}
+{{- end }}
+{{- end }}
+{{- end }}

charts/descheduler/templates/tests/test-descheduler-pod.yaml

@@ -2,6 +2,7 @@ apiVersion: v1
 kind: Pod
 metadata:
     name: descheduler-test-pod
+    namespace: {{ .Release.Namespace }}
     annotations:
       "helm.sh/hook": test
 spec:
@@ -26,4 +27,4 @@ spec:
             curl -LO https://storage.googleapis.com/kubernetes-release/release/$(curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt)/bin/linux/amd64/kubectl &&
             chmod +x ./kubectl &&
             mv ./kubectl /usr/local/bin/kubectl &&
-            /usr/local/bin/kubectl get pods --namespace kube-system --token "$(cat /var/run/secrets/kubernetes.io/serviceaccount/token)" | grep "descheduler" | grep "Completed"
+            /usr/local/bin/kubectl get pods --namespace {{ .Release.Namespace }} --token "$(cat /var/run/secrets/kubernetes.io/serviceaccount/token)" | grep "descheduler" | grep "Completed"

charts/descheduler/values.yaml

@@ -2,12 +2,18 @@
 # This is a YAML-formatted file.
 # Declare variables to be passed into your templates.
 
+# CronJob or Deployment
+kind: CronJob
+
 image:
   repository: k8s.gcr.io/descheduler/descheduler
   # Overrides the image tag whose default is the chart version
   tag: ""
   pullPolicy: IfNotPresent
 
+imagePullSecrets:
+#   - name: container-registry-secret
+
 resources:
   requests:
     cpu: 500m
@@ -19,19 +25,45 @@ resources:
 nameOverride: ""
 fullnameOverride: ""
 
+# labels that'll be applied to all resources
+commonLabels: {}
+
 cronJobApiVersion: "batch/v1"  # Use "batch/v1beta1" for k8s version < 1.21.0. TODO(@7i) remove with 1.23 release
 schedule: "*/2 * * * *"
+suspend: false
 #startingDeadlineSeconds: 200
 #successfulJobsHistoryLimit: 1
 #failedJobsHistoryLimit: 1
 
+# Required when running as a Deployment
+deschedulingInterval: 5m
+
+# Specifies the replica count for Deployment
+# Set leaderElection if you want to use more than 1 replica
+# Set affinity.podAntiAffinity rule if you want to schedule onto a node
+# only if that node is in the same zone as at least one already-running descheduler
+replicas: 1
+
+# Specifies whether Leader Election resources should be created
+# Required when running as a Deployment
+leaderElection: {}
+#  enabled: true
+#  leaseDuration: 15
+#  renewDeadline: 10
+#  retryPeriod: 2
+#  resourceLock: "leases"
+#  resourceName: "descheduler"
+#  resourceNamescape: "kube-system"
+
 cmdOptions:
   v: 3
-  # evict-local-storage-pods:
-  # max-pods-to-evict-per-node: 10
-  # node-selector: "key1=value1,key2=value2"
 
 deschedulerPolicy:
+  # nodeSelector: "key1=value1,key2=value2"
+  # maxNoOfPodsToEvictPerNode: 10
+  # maxNoOfPodsToEvictPerNamespace: 10
+  # ignorePvcPods: true
+  # evictLocalStoragePods: true
   strategies:
     RemoveDuplicates:
       enabled: true
@@ -62,6 +94,31 @@ priorityClassName: system-cluster-critical
 nodeSelector: {}
 #  foo: bar
 
+affinity: {}
+# nodeAffinity:
+#   requiredDuringSchedulingIgnoredDuringExecution:
+#     nodeSelectorTerms:
+#     - matchExpressions:
+#       - key: kubernetes.io/e2e-az-name
+#         operator: In
+#         values:
+#         - e2e-az1
+#         - e2e-az2
+#  podAntiAffinity:
+#    requiredDuringSchedulingIgnoredDuringExecution:
+#      - labelSelector:
+#          matchExpressions:
+#            - key: app.kubernetes.io/name
+#              operator: In
+#              values:
+#                - descheduler
+#        topologyKey: "kubernetes.io/hostname"
+tolerations: []
+# - key: 'management'
+#   operator: 'Equal'
+#   value: 'tool'
+#   effect: 'NoSchedule'
+
 rbac:
   # Specifies whether RBAC resources should be created
   create: true
@@ -76,3 +133,37 @@ serviceAccount:
   # The name of the ServiceAccount to use.
   # If not set and create is true, a name is generated using the fullname template
   name:
+  # Specifies custom annotations for the serviceAccount
+  annotations: {}
+
+livenessProbe:
+  failureThreshold: 3
+  httpGet:
+    path: /healthz
+    port: 10258
+    scheme: HTTPS
+  initialDelaySeconds: 3
+  periodSeconds: 10
+
+service:
+  enabled: false
+
+serviceMonitor:
+  enabled: false
+  # The namespace where Prometheus expects to find service monitors.
+  # namespace: ""
+  interval: ""
+  # honorLabels: true
+  insecureSkipVerify: true
+  serverName: null
+  metricRelabelings: []
+    # - action: keep
+    #   regex: 'descheduler_(build_info|pods_evicted)'
+    #   sourceLabels: [__name__]
+  relabelings: []
+    # - sourceLabels: [__meta_kubernetes_pod_node_name]
+    #   separator: ;
+    #   regex: ^(.*)$
+    #   targetLabel: nodename
+    #   replacement: $1
+    #   action: replace

cloudbuild.yaml

@@ -7,7 +7,7 @@ timeout: 1200s
 options:
   substitution_option: ALLOW_LOOSE
 steps:
-  - name: 'gcr.io/k8s-testimages/gcb-docker-gcloud:v20190906-745fed4'
+  - name: 'gcr.io/k8s-staging-test-infra/gcb-docker-gcloud:v20211118-2f2d816b90'
     entrypoint: make
     env:
     - DOCKER_CLI_EXPERIMENTAL=enabled

cmd/descheduler/app/options/options.go

@@ -19,15 +19,15 @@ package options
 
 import (
 	"github.com/spf13/pflag"
-
-	utilerrors "k8s.io/apimachinery/pkg/util/errors"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	apiserveroptions "k8s.io/apiserver/pkg/server/options"
 	clientset "k8s.io/client-go/kubernetes"
-	"k8s.io/component-base/logs"
-
+	componentbaseconfig "k8s.io/component-base/config"
+	componentbaseoptions "k8s.io/component-base/config/options"
 	"sigs.k8s.io/descheduler/pkg/apis/componentconfig"
 	"sigs.k8s.io/descheduler/pkg/apis/componentconfig/v1alpha1"
 	deschedulerscheme "sigs.k8s.io/descheduler/pkg/descheduler/scheme"
+	"time"
 )
 
 const (
@@ -39,7 +39,6 @@ type DeschedulerServer struct {
 	componentconfig.DeschedulerConfiguration
 
 	Client         clientset.Interface
-	Logs           *logs.Options
 	SecureServing  *apiserveroptions.SecureServingOptionsWithLoopback
 	DisableMetrics bool
 }
@@ -56,20 +55,22 @@ func NewDeschedulerServer() (*DeschedulerServer, error) {
 
 	return &DeschedulerServer{
 		DeschedulerConfiguration: *cfg,
-		Logs:                     logs.NewOptions(),
 		SecureServing:            secureServing,
 	}, nil
 }
 
-// Validation checks for DeschedulerServer.
-func (s *DeschedulerServer) Validate() error {
-	var errs []error
-	errs = append(errs, s.Logs.Validate()...)
-	return utilerrors.NewAggregate(errs)
-}
-
 func newDefaultComponentConfig() (*componentconfig.DeschedulerConfiguration, error) {
-	versionedCfg := v1alpha1.DeschedulerConfiguration{}
+	versionedCfg := v1alpha1.DeschedulerConfiguration{
+		LeaderElection: componentbaseconfig.LeaderElectionConfiguration{
+			LeaderElect:       false,
+			LeaseDuration:     metav1.Duration{Duration: 137 * time.Second},
+			RenewDeadline:     metav1.Duration{Duration: 107 * time.Second},
+			RetryPeriod:       metav1.Duration{Duration: 26 * time.Second},
+			ResourceLock:      "leases",
+			ResourceName:      "descheduler",
+			ResourceNamespace: "kube-system",
+		},
+	}
 	deschedulerscheme.Scheme.Default(&versionedCfg)
 	cfg := componentconfig.DeschedulerConfiguration{}
 	if err := deschedulerscheme.Scheme.Convert(&versionedCfg, &cfg, nil); err != nil {
@@ -85,13 +86,9 @@ func (rs *DeschedulerServer) AddFlags(fs *pflag.FlagSet) {
 	fs.StringVar(&rs.KubeconfigFile, "kubeconfig", rs.KubeconfigFile, "File with  kube configuration.")
 	fs.StringVar(&rs.PolicyConfigFile, "policy-config-file", rs.PolicyConfigFile, "File with descheduler policy configuration.")
 	fs.BoolVar(&rs.DryRun, "dry-run", rs.DryRun, "execute descheduler in dry run mode.")
-	// node-selector query causes descheduler to run only on nodes that matches the node labels in the query
-	fs.StringVar(&rs.NodeSelector, "node-selector", rs.NodeSelector, "DEPRECATED: selector (label query) to filter on, supports '=', '==', and '!='.(e.g. -l key1=value1,key2=value2)")
-	// max-no-pods-to-evict limits the maximum number of pods to be evicted per node by descheduler.
-	fs.IntVar(&rs.MaxNoOfPodsToEvictPerNode, "max-pods-to-evict-per-node", rs.MaxNoOfPodsToEvictPerNode, "DEPRECATED: limits the maximum number of pods to be evicted per node by descheduler")
-	// evict-local-storage-pods allows eviction of pods that are using local storage. This is false by default.
-	fs.BoolVar(&rs.EvictLocalStoragePods, "evict-local-storage-pods", rs.EvictLocalStoragePods, "DEPRECATED: enables evicting pods using local storage by descheduler")
 	fs.BoolVar(&rs.DisableMetrics, "disable-metrics", rs.DisableMetrics, "Disables metrics. The metrics are by default served through https://localhost:10258/metrics. Secure address, resp. port can be changed through --bind-address, resp. --secure-port flags.")
 
+	componentbaseoptions.BindLeaderElectionFlags(&rs.LeaderElection, fs)
+
 	rs.SecureServing.AddFlags(fs)
 }

cmd/descheduler/app/server.go

@@ -19,8 +19,11 @@ package app
 
 import (
 	"context"
-	"flag"
 	"io"
+	"os/signal"
+	"syscall"
+
+	"k8s.io/apiserver/pkg/server/healthz"
 
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
 	"sigs.k8s.io/descheduler/pkg/descheduler"
@@ -30,7 +33,9 @@ import (
 	apiserver "k8s.io/apiserver/pkg/server"
 	"k8s.io/apiserver/pkg/server/mux"
 	restclient "k8s.io/client-go/rest"
-	aflag "k8s.io/component-base/cli/flag"
+	"k8s.io/component-base/config"
+	_ "k8s.io/component-base/logs/json/register"
+	"k8s.io/component-base/logs/registry"
 	"k8s.io/component-base/metrics/legacyregistry"
 	"k8s.io/klog/v2"
 )
@@ -48,8 +53,7 @@ func NewDeschedulerCommand(out io.Writer) *cobra.Command {
 		Short: "descheduler",
 		Long:  `The descheduler evicts pods which may be bound to less desired nodes`,
 		Run: func(cmd *cobra.Command, args []string) {
-			s.Logs.LogFormat = s.Logging.Format
-			s.Logs.Apply()
+			// s.Logs.Config.Format = s.Logging.Format
 
 			// LoopbackClientConfig is a config for a privileged loopback connection
 			var LoopbackClientConfig *restclient.Config
@@ -59,36 +63,46 @@ func NewDeschedulerCommand(out io.Writer) *cobra.Command {
 				return
 			}
 
-			if err := s.Validate(); err != nil {
-				klog.ErrorS(err, "failed to validate server configuration")
+			factory, _ := registry.LogRegistry.Get(s.Logging.Format)
+			if factory == nil {
+				klog.ClearLogger()
+			} else {
+				log, logrFlush := factory.Create(config.LoggingConfiguration{})
+				defer logrFlush()
+				klog.SetLogger(log)
+			}
+
+			ctx, done := signal.NotifyContext(context.Background(), syscall.SIGINT, syscall.SIGTERM)
+
+			pathRecorderMux := mux.NewPathRecorderMux("descheduler")
+			if !s.DisableMetrics {
+				pathRecorderMux.Handle("/metrics", legacyregistry.HandlerWithReset())
+			}
+
+			healthz.InstallHandler(pathRecorderMux, healthz.NamedCheck("Descheduler", healthz.PingHealthz.Check))
+
+			stoppedCh, _, err := SecureServing.Serve(pathRecorderMux, 0, ctx.Done())
+			if err != nil {
+				klog.Fatalf("failed to start secure server: %v", err)
 				return
 			}
 
-			if !s.DisableMetrics {
-				ctx := context.TODO()
-				pathRecorderMux := mux.NewPathRecorderMux("descheduler")
-				pathRecorderMux.Handle("/metrics", legacyregistry.HandlerWithReset())
-
-				if _, err := SecureServing.Serve(pathRecorderMux, 0, ctx.Done()); err != nil {
-					klog.Fatalf("failed to start secure server: %v", err)
-					return
-				}
-			}
-
-			err := Run(s)
+			err = Run(ctx, s)
 			if err != nil {
 				klog.ErrorS(err, "descheduler server")
 			}
+
+			done()
+			// wait for metrics server to close
+			<-stoppedCh
 		},
 	}
 	cmd.SetOut(out)
 	flags := cmd.Flags()
-	flags.SetNormalizeFunc(aflag.WordSepNormalizeFunc)
-	flags.AddGoFlagSet(flag.CommandLine)
 	s.AddFlags(flags)
 	return cmd
 }
 
-func Run(rs *options.DeschedulerServer) error {
-	return descheduler.Run(rs)
+func Run(ctx context.Context, rs *options.DeschedulerServer) error {
+	return descheduler.Run(ctx, rs)
 }

cmd/descheduler/descheduler.go

@@ -17,22 +17,23 @@ limitations under the License.
 package main
 
 import (
-	"fmt"
-	"k8s.io/component-base/logs"
 	"os"
+
+	"k8s.io/component-base/cli"
+	"k8s.io/klog/v2"
 	"sigs.k8s.io/descheduler/cmd/descheduler/app"
 )
 
+func init() {
+	klog.SetOutput(os.Stdout)
+	klog.InitFlags(nil)
+}
+
 func main() {
 	out := os.Stdout
 	cmd := app.NewDeschedulerCommand(out)
 	cmd.AddCommand(app.NewVersionCommand())
 
-	logs.InitLogs()
-	defer logs.FlushLogs()
-
-	if err := cmd.Execute(); err != nil {
-		fmt.Println(err)
-		os.Exit(1)
-	}
+	code := cli.Run(cmd)
+	os.Exit(code)
 }

docs/proposals.md

@@ -0,0 +1,16 @@
+# Proposals
+This document walk you through about all the enhancements proposals for descheduler.
+
+## Descheduler v1alpha2 Design Proposal
+```yaml
+title: Descheduler v1alpha2 Design Proposal
+authors:
+- "@damemi"
+link: 
+- https://docs.google.com/document/d/1S1JCh-0F-QCJvBBG-kbmXiHAJFF8doArhDIAKbOj93I/edit#heading=h.imbp1ctnc8lx
+- https://github.com/kubernetes-sigs/descheduler/issues/679
+owning-sig: sig-scheduling
+creation-date: 2021-05-01
+status: implementable
+```
+

docs/release-guide.md

@@ -9,7 +9,7 @@
 3. Push the release branch to the descheuler repo and ensure branch protection is enabled (not required for patch releases)
 4. Tag the repository from the `master` branch (from the `release-1.18` branch for a patch release) and push the tag `VERSION=v0.18.0 git tag -m $VERSION $VERSION; git push origin $VERSION`
 5. Publish a draft release using the tag you just created
-6. Perform the [image promotion process](https://github.com/kubernetes/k8s.io/tree/master/k8s.gcr.io#image-promoter)
+6. Perform the [image promotion process](https://github.com/kubernetes/k8s.io/tree/main/k8s.gcr.io#image-promoter)
 7. Publish release
 8. Email `kubernetes-sig-scheduling@googlegroups.com` to announce the release
 
@@ -22,7 +22,7 @@
 5. Checkout the tag you just created and make sure your repo is clean by git's standards `git checkout $VERSION`
 6. Build and push the container image to the staging registry `VERSION=$VERSION make push-all`
 7. Publish a draft release using the tag you just created
-8. Perform the [image promotion process](https://github.com/kubernetes/k8s.io/tree/master/k8s.gcr.io#image-promoter)
+8. Perform the [image promotion process](https://github.com/kubernetes/k8s.io/tree/main/k8s.gcr.io#image-promoter)
 9. Publish release
 10. Email `kubernetes-sig-scheduling@googlegroups.com` to announce the release
 

docs/user-guide.md

@@ -2,13 +2,16 @@
 
 Starting with descheduler release v0.10.0 container images are available in the official k8s container registry.
 
-Descheduler Version | Container Image                                     | Architectures           |
-------------------- |-----------------------------------------------------|-------------------------|
-v0.21.0             | k8s.gcr.io/descheduler/descheduler:v0.21.0          | AMD64<br>ARM64<br>ARMv7 |
-v0.20.0             | k8s.gcr.io/descheduler/descheduler:v0.20.0          | AMD64<br>ARM64          |
-v0.19.0             | k8s.gcr.io/descheduler/descheduler:v0.19.0          | AMD64                   |
-v0.18.0             | k8s.gcr.io/descheduler/descheduler:v0.18.0          | AMD64                   |
-v0.10.0             | k8s.gcr.io/descheduler/descheduler:v0.10.0          | AMD64                   |
+Descheduler Version | Container Image                            | Architectures           |
+------------------- |--------------------------------------------|-------------------------|
+v0.24.0             | k8s.gcr.io/descheduler/descheduler:v0.24.0 | AMD64<br>ARM64<br>ARMv7 |
+v0.23.1             | k8s.gcr.io/descheduler/descheduler:v0.23.1 | AMD64<br>ARM64<br>ARMv7 |
+v0.22.0             | k8s.gcr.io/descheduler/descheduler:v0.22.0 | AMD64<br>ARM64<br>ARMv7 |
+v0.21.0             | k8s.gcr.io/descheduler/descheduler:v0.21.0 | AMD64<br>ARM64<br>ARMv7 |
+v0.20.0             | k8s.gcr.io/descheduler/descheduler:v0.20.0 | AMD64<br>ARM64          |
+v0.19.0             | k8s.gcr.io/descheduler/descheduler:v0.19.0 | AMD64                   |
+v0.18.0             | k8s.gcr.io/descheduler/descheduler:v0.18.0 | AMD64                   |
+v0.10.0             | k8s.gcr.io/descheduler/descheduler:v0.10.0 | AMD64                   |
 
 Note that multi-arch container images cannot be pulled by [kind](https://kind.sigs.k8s.io) from a registry. Therefore
 starting with descheduler release v0.20.0 use the below process to download the official descheduler
@@ -33,31 +36,52 @@ Usage:
   descheduler [command]
 
 Available Commands:
+  completion  generate the autocompletion script for the specified shell
   help        Help about any command
   version     Version of descheduler
 
 Flags:
-      --add-dir-header                   If true, adds the file directory to the header of the log messages
-      --alsologtostderr                  log to standard error as well as files
-      --descheduling-interval duration   Time interval between two consecutive descheduler executions. Setting this value instructs the descheduler to run in a continuous loop at the interval specified.
-      --dry-run                          execute descheduler in dry run mode.
-      --evict-local-storage-pods         DEPRECATED: enables evicting pods using local storage by descheduler
-  -h, --help                             help for descheduler
-      --kubeconfig string                File with  kube configuration.
-      --log-backtrace-at traceLocation   when logging hits line file:N, emit a stack trace (default :0)
-      --log-dir string                   If non-empty, write log files in this directory
-      --log-file string                  If non-empty, use this log file
-      --log-file-max-size uint           Defines the maximum size a log file can grow to. Unit is megabytes. If the value is 0, the maximum file size is unlimited. (default 1800)
-      --log-flush-frequency duration     Maximum number of seconds between log flushes (default 5s)
-      --logtostderr                      log to standard error instead of files (default true)
-      --max-pods-to-evict-per-node int   DEPRECATED: limits the maximum number of pods to be evicted per node by descheduler
-      --node-selector string             DEPRECATED: selector (label query) to filter on, supports '=', '==', and '!='.(e.g. -l key1=value1,key2=value2)
-      --policy-config-file string        File with descheduler policy configuration.
-      --skip-headers                     If true, avoid header prefixes in the log messages
-      --skip-log-headers                 If true, avoid headers when opening log files
-      --stderrthreshold severity         logs at or above this threshold go to stderr (default 2)
-  -v, --v Level                          number for the log level verbosity
-      --vmodule moduleSpec               comma-separated list of pattern=N settings for file-filtered logging
+      --add-dir-header                           If true, adds the file directory to the header of the log messages (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --alsologtostderr                          log to standard error as well as files (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --bind-address ip                          The IP address on which to listen for the --secure-port port. The associated interface(s) must be reachable by the rest of the cluster, and by CLI/web clients. If blank or an unspecified address (0.0.0.0 or ::), all interfaces will be used. (default 0.0.0.0)
+      --cert-dir string                          The directory where the TLS certs are located. If --tls-cert-file and --tls-private-key-file are provided, this flag will be ignored. (default "apiserver.local.config/certificates")
+      --descheduling-interval duration           Time interval between two consecutive descheduler executions. Setting this value instructs the descheduler to run in a continuous loop at the interval specified.
+      --disable-metrics                          Disables metrics. The metrics are by default served through https://localhost:10258/metrics. Secure address, resp. port can be changed through --bind-address, resp. --secure-port flags.
+      --dry-run                                  execute descheduler in dry run mode.
+  -h, --help                                     help for descheduler
+      --http2-max-streams-per-connection int     The limit that the server gives to clients for the maximum number of streams in an HTTP/2 connection. Zero means to use golang's default.
+      --kubeconfig string                        File with  kube configuration.
+      --leader-elect                             Start a leader election client and gain leadership before executing the main loop. Enable this when running replicated components for high availability.
+      --leader-elect-lease-duration duration     The duration that non-leader candidates will wait after observing a leadership renewal until attempting to acquire leadership of a led but unrenewed leader slot. This is effectively the maximum duration that a leader can be stopped before it is replaced by another candidate. This is only applicable if leader election is enabled. (default 15s)
+      --leader-elect-renew-deadline duration     The interval between attempts by the acting master to renew a leadership slot before it stops leading. This must be less than or equal to the lease duration. This is only applicable if leader election is enabled. (default 10s)
+      --leader-elect-resource-lock string        The type of resource object that is used for locking during leader election. Supported options are 'endpoints', 'configmaps', 'leases', 'endpointsleases' and 'configmapsleases'. (default "leases")
+      --leader-elect-resource-name string        The name of resource object that is used for locking during leader election. (default "descheduler")
+      --leader-elect-resource-namespace string   The namespace of resource object that is used for locking during leader election. (default "kube-system")
+      --leader-elect-retry-period duration       The duration the clients should wait between attempting acquisition and renewal of a leadership. This is only applicable if leader election is enabled. (default 2s)
+      --log-backtrace-at traceLocation           when logging hits line file:N, emit a stack trace (default :0) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --log-dir string                           If non-empty, write log files in this directory (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --log-file string                          If non-empty, use this log file (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --log-file-max-size uint                   Defines the maximum size a log file can grow to. Unit is megabytes. If the value is 0, the maximum file size is unlimited. (default 1800) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --log-flush-frequency duration             Maximum number of seconds between log flushes (default 5s)
+      --logging-format string                    Sets the log format. Permitted formats: "text", "json". Non-default formats don't honor these flags: --add-dir-header, --alsologtostderr, --log-backtrace-at, --log-dir, --log-file, --log-file-max-size, --logtostderr, --skip-headers, --skip-log-headers, --stderrthreshold, --log-flush-frequency.\nNon-default choices are currently alpha and subject to change without warning. (default "text")
+      --logtostderr                              log to standard error instead of files (default true) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --one-output                               If true, only write logs to their native severity level (vs also writing to each lower severity level) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --permit-address-sharing                   If true, SO_REUSEADDR will be used when binding the port. This allows binding to wildcard IPs like 0.0.0.0 and specific IPs in parallel, and it avoids waiting for the kernel to release sockets in TIME_WAIT state. [default=false]
+      --permit-port-sharing                      If true, SO_REUSEPORT will be used when binding the port, which allows more than one instance to bind on the same address and port. [default=false]
+      --policy-config-file string                File with descheduler policy configuration.
+      --secure-port int                          The port on which to serve HTTPS with authentication and authorization. If 0, don't serve HTTPS at all. (default 10258)
+      --skip-headers                             If true, avoid header prefixes in the log messages (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --skip-log-headers                         If true, avoid headers when opening log files (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --stderrthreshold severity                 logs at or above this threshold go to stderr (default 2) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
+      --tls-cert-file string                     File containing the default x509 Certificate for HTTPS. (CA cert, if any, concatenated after server cert). If HTTPS serving is enabled, and --tls-cert-file and --tls-private-key-file are not provided, a self-signed certificate and key are generated for the public address and saved to the directory specified by --cert-dir.
+      --tls-cipher-suites strings                Comma-separated list of cipher suites for the server. If omitted, the default Go cipher suites will be used. 
+                                                 Preferred values: TLS_AES_128_GCM_SHA256, TLS_AES_256_GCM_SHA384, TLS_CHACHA20_POLY1305_SHA256, TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA, TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA, TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384, TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305, TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256, TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256, TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384, TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305, TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256, TLS_RSA_WITH_AES_128_CBC_SHA, TLS_RSA_WITH_AES_128_GCM_SHA256, TLS_RSA_WITH_AES_256_CBC_SHA, TLS_RSA_WITH_AES_256_GCM_SHA384. 
+                                                 Insecure values: TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, TLS_ECDHE_ECDSA_WITH_RC4_128_SHA, TLS_ECDHE_RSA_WITH_3DES_EDE_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, TLS_ECDHE_RSA_WITH_RC4_128_SHA, TLS_RSA_WITH_3DES_EDE_CBC_SHA, TLS_RSA_WITH_AES_128_CBC_SHA256, TLS_RSA_WITH_RC4_128_SHA.
+      --tls-min-version string                   Minimum TLS version supported. Possible values: VersionTLS10, VersionTLS11, VersionTLS12, VersionTLS13
+      --tls-private-key-file string              File containing the default x509 private key matching --tls-cert-file.
+      --tls-sni-cert-key namedCertKey            A pair of x509 certificate and private key file paths, optionally suffixed with a list of domain patterns which are fully qualified domain names, possibly with prefixed wildcard segments. The domain patterns also allow IP addresses, but IPs should only be used if the apiserver has visibility to the IP address requested by a client. If no domain patterns are provided, the names of the certificate are extracted. Non-wildcard matches trump over wildcard matches, explicit domain patterns trump over extracted names. For multiple key/certificate pairs, use the --tls-sni-cert-key multiple times. Examples: "example.crt,example.key" or "foo.crt,foo.key:*.foo.com,foo.com". (default [])
+  -v, --v Level                                  number for the log level verbosity
+      --vmodule moduleSpec                       comma-separated list of pattern=N settings for file-filtered logging
 
 Use "descheduler [command] --help" for more information about a command.
 ```
@@ -88,7 +112,8 @@ strategies:
   "PodLifeTime":
     enabled: true
     params:
-      maxPodLifeTimeSeconds: 604800 # pods run for a maximum of 7 days
+      podLifeTime:
+        maxPodLifeTimeSeconds: 604800 # pods run for a maximum of 7 days
 ```
 
 ### Balance Cluster By Node Memory Utilization
@@ -116,7 +141,7 @@ strategies:
 
 #### Balance low utilization nodes
 Using `HighNodeUtilization`, descheduler will rebalance the cluster based on memory by evicting pods
-from nodes with memory utilization lower than 20%. This should be used along with scheduler strategy `MostRequestedPriority`.
+from nodes with memory utilization lower than 20%. This should be use `NodeResourcesFit` with the `MostAllocated` scoring strategy based on these [doc](https://kubernetes.io/docs/reference/scheduling/config/#scheduling-plugins).
 The evicted pods will be compacted into minimal set of nodes.
 
 ```
@@ -135,7 +160,14 @@ strategies:
 Descheduler's `RemovePodsViolatingNodeTaints` strategy can be combined with
 [Node Problem Detector](https://github.com/kubernetes/node-problem-detector/) and
 [Cluster Autoscaler](https://github.com/kubernetes/autoscaler/tree/master/cluster-autoscaler) to automatically remove
-Nodes which have problems. Node Problem Detector can detect specific Node problems and taint any Nodes which have those
-problems. The Descheduler will then deschedule workloads from those Nodes. Finally, if the descheduled Node's resource
+Nodes which have problems. Node Problem Detector can detect specific Node problems and report them to the API server.
+There is a feature called TaintNodeByCondition of the node controller that takes some conditions and turns them into taints. Currently, this only works for the default node conditions: PIDPressure, MemoryPressure, DiskPressure, Ready, and some cloud provider specific conditions.
+The Descheduler will then deschedule workloads from those Nodes. Finally, if the descheduled Node's resource
 allocation falls below the Cluster Autoscaler's scale down threshold, the Node will become a scale down candidate
 and can be removed by Cluster Autoscaler. These three components form an autohealing cycle for Node problems.
+---
+**NOTE**
+
+Once [kubernetes/node-problem-detector#565](https://github.com/kubernetes/node-problem-detector/pull/565) is available in NPD, we need to update this section.
+
+---

examples/failed-pods.yaml

@@ -0,0 +1,14 @@
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "RemoveFailedPods":
+    enabled: true
+    params:
+      failedPods:
+        reasons:
+        - "OutOfcpu"
+        - "CreateContainerConfigError"
+        includingInitContainers: true
+        excludeOwnerKinds:
+        - "Job"
+        minPodLifetimeSeconds: 3600 # 1 hour

examples/high-node-utilization.yml

@@ -1,4 +1,3 @@
----
 apiVersion: "descheduler/v1alpha1"
 kind: "DeschedulerPolicy"
 strategies:

examples/low-node-utilization.yml

@@ -1,4 +1,3 @@
----
 apiVersion: "descheduler/v1alpha1"
 kind: "DeschedulerPolicy"
 strategies:

examples/pod-life-time.yml

@@ -1,4 +1,3 @@
----
 apiVersion: "descheduler/v1alpha1"
 kind: "DeschedulerPolicy"
 strategies:

examples/policy.yaml

@@ -23,3 +23,7 @@ strategies:
        podsHavingTooManyRestarts:
          podRestartThreshold: 100
          includingInitContainers: true
+  "RemovePodsViolatingTopologySpreadConstraint":
+     enabled: true
+     params:
+       includeSoftConstraints: true

examples/topology-spread-constraint.yaml

@@ -0,0 +1,7 @@
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "RemovePodsViolatingTopologySpreadConstraint":
+     enabled: true
+     params:
+       includeSoftConstraints: true # Include 'ScheduleAnyways' constraints

go.mod

@@ -1,19 +1,117 @@
 module sigs.k8s.io/descheduler
 
-go 1.16
+go 1.17
 
 require (
 	github.com/client9/misspell v0.3.4
-	github.com/spf13/cobra v1.1.1
+	github.com/spf13/cobra v1.4.0
 	github.com/spf13/pflag v1.0.5
-	k8s.io/api v0.21.0
-	k8s.io/apimachinery v0.21.0
-	k8s.io/apiserver v0.21.0
-	k8s.io/client-go v0.21.0
-	k8s.io/code-generator v0.21.0
-	k8s.io/component-base v0.21.0
-	k8s.io/component-helpers v0.21.0
-	k8s.io/klog/v2 v2.8.0
+	k8s.io/api v0.24.0
+	k8s.io/apimachinery v0.24.0
+	k8s.io/apiserver v0.24.0
+	k8s.io/client-go v0.24.0
+	k8s.io/code-generator v0.24.0
+	k8s.io/component-base v0.24.0
+	k8s.io/component-helpers v0.24.0
+	k8s.io/klog/v2 v2.60.1
 	k8s.io/kubectl v0.20.5
 	sigs.k8s.io/mdtoc v1.0.1
 )
+
+require (
+	cloud.google.com/go v0.81.0 // indirect
+	github.com/Azure/go-autorest v14.2.0+incompatible // indirect
+	github.com/Azure/go-autorest/autorest v0.11.18 // indirect
+	github.com/Azure/go-autorest/autorest/adal v0.9.13 // indirect
+	github.com/Azure/go-autorest/autorest/date v0.3.0 // indirect
+	github.com/Azure/go-autorest/logger v0.2.1 // indirect
+	github.com/Azure/go-autorest/tracing v0.6.0 // indirect
+	github.com/BurntSushi/toml v0.3.1 // indirect
+	github.com/NYTimes/gziphandler v1.1.1 // indirect
+	github.com/PuerkitoBio/purell v1.1.1 // indirect
+	github.com/PuerkitoBio/urlesc v0.0.0-20170810143723-de5bf2ad4578 // indirect
+	github.com/beorn7/perks v1.0.1 // indirect
+	github.com/blang/semver/v4 v4.0.0 // indirect
+	github.com/cespare/xxhash/v2 v2.1.2 // indirect
+	github.com/coreos/go-semver v0.3.0 // indirect
+	github.com/coreos/go-systemd/v22 v22.3.2 // indirect
+	github.com/davecgh/go-spew v1.1.1 // indirect
+	github.com/emicklei/go-restful v2.9.5+incompatible // indirect
+	github.com/evanphx/json-patch v4.12.0+incompatible // indirect
+	github.com/felixge/httpsnoop v1.0.1 // indirect
+	github.com/form3tech-oss/jwt-go v3.2.3+incompatible // indirect
+	github.com/fsnotify/fsnotify v1.4.9 // indirect
+	github.com/go-logr/logr v1.2.0 // indirect
+	github.com/go-logr/zapr v1.2.0 // indirect
+	github.com/go-openapi/jsonpointer v0.19.5 // indirect
+	github.com/go-openapi/jsonreference v0.19.5 // indirect
+	github.com/go-openapi/swag v0.19.14 // indirect
+	github.com/gogo/protobuf v1.3.2 // indirect
+	github.com/golang/groupcache v0.0.0-20210331224755-41bb18bfe9da // indirect
+	github.com/golang/protobuf v1.5.2 // indirect
+	github.com/gomarkdown/markdown v0.0.0-20200824053859-8c8b3816f167 // indirect
+	github.com/google/gnostic v0.5.7-v3refs // indirect
+	github.com/google/go-cmp v0.5.5 // indirect
+	github.com/google/gofuzz v1.1.0 // indirect
+	github.com/google/uuid v1.1.2 // indirect
+	github.com/grpc-ecosystem/go-grpc-prometheus v1.2.0 // indirect
+	github.com/grpc-ecosystem/grpc-gateway v1.16.0 // indirect
+	github.com/imdario/mergo v0.3.5 // indirect
+	github.com/inconshreveable/mousetrap v1.0.0 // indirect
+	github.com/josharian/intern v1.0.0 // indirect
+	github.com/json-iterator/go v1.1.12 // indirect
+	github.com/mailru/easyjson v0.7.6 // indirect
+	github.com/matttproud/golang_protobuf_extensions v1.0.2-0.20181231171920-c182affec369 // indirect
+	github.com/mmarkdown/mmark v2.0.40+incompatible // indirect
+	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
+	github.com/modern-go/reflect2 v1.0.2 // indirect
+	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
+	github.com/pkg/errors v0.9.1 // indirect
+	github.com/prometheus/client_golang v1.12.1 // indirect
+	github.com/prometheus/client_model v0.2.0 // indirect
+	github.com/prometheus/common v0.32.1 // indirect
+	github.com/prometheus/procfs v0.7.3 // indirect
+	go.etcd.io/etcd/api/v3 v3.5.1 // indirect
+	go.etcd.io/etcd/client/pkg/v3 v3.5.1 // indirect
+	go.etcd.io/etcd/client/v3 v3.5.1 // indirect
+	go.opentelemetry.io/contrib v0.20.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.20.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.20.0 // indirect
+	go.opentelemetry.io/otel v0.20.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp v0.20.0 // indirect
+	go.opentelemetry.io/otel/metric v0.20.0 // indirect
+	go.opentelemetry.io/otel/sdk v0.20.0 // indirect
+	go.opentelemetry.io/otel/sdk/export/metric v0.20.0 // indirect
+	go.opentelemetry.io/otel/sdk/metric v0.20.0 // indirect
+	go.opentelemetry.io/otel/trace v0.20.0 // indirect
+	go.opentelemetry.io/proto/otlp v0.7.0 // indirect
+	go.uber.org/atomic v1.7.0 // indirect
+	go.uber.org/multierr v1.6.0 // indirect
+	go.uber.org/zap v1.19.0 // indirect
+	golang.org/x/crypto v0.0.0-20220214200702-86341886e292 // indirect
+	golang.org/x/mod v0.6.0-dev.0.20220106191415-9b9b3d81d5e3 // indirect
+	golang.org/x/net v0.0.0-20220127200216-cd36cc0744dd // indirect
+	golang.org/x/oauth2 v0.0.0-20211104180415-d3ed0bb246c8 // indirect
+	golang.org/x/sync v0.0.0-20210220032951-036812b2e83c // indirect
+	golang.org/x/sys v0.0.0-20220209214540-3681064d5158 // indirect
+	golang.org/x/term v0.0.0-20210927222741-03fcf44c2211 // indirect
+	golang.org/x/text v0.3.7 // indirect
+	golang.org/x/time v0.0.0-20220210224613-90d013bbcef8 // indirect
+	golang.org/x/tools v0.1.10-0.20220218145154-897bd77cd717 // indirect
+	golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1 // indirect
+	google.golang.org/appengine v1.6.7 // indirect
+	google.golang.org/genproto v0.0.0-20220107163113-42d7afdf6368 // indirect
+	google.golang.org/grpc v1.40.0 // indirect
+	google.golang.org/protobuf v1.27.1 // indirect
+	gopkg.in/inf.v0 v0.9.1 // indirect
+	gopkg.in/natefinch/lumberjack.v2 v2.0.0 // indirect
+	gopkg.in/yaml.v2 v2.4.0 // indirect
+	gopkg.in/yaml.v3 v3.0.0-20210107192922-496545a6307b // indirect
+	k8s.io/gengo v0.0.0-20211129171323-c02415ce4185 // indirect
+	k8s.io/kube-openapi v0.0.0-20220328201542-3ee0da9b0b42 // indirect
+	k8s.io/utils v0.0.0-20220210201930-3a6ce19ff2f9 // indirect
+	sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.0.30 // indirect
+	sigs.k8s.io/json v0.0.0-20211208200746-9f7c6b3444d2 // indirect
+	sigs.k8s.io/structured-merge-diff/v4 v4.2.1 // indirect
+	sigs.k8s.io/yaml v1.2.0 // indirect
+)

hack/e2e-gce/gcloud_create_cluster.sh

@@ -18,15 +18,15 @@ E2E_GCE_HOME=$DESCHEDULER_ROOT/hack/e2e-gce
 
 create_cluster() {
 	echo "#################### Creating instances ##########################"
-        gcloud compute instances create descheduler-$master_uuid --image="ubuntu-1604-xenial-v20180306" --image-project="ubuntu-os-cloud" --zone=us-east1-b
+	gcloud compute instances create descheduler-$master_uuid --image-family="ubuntu-1804-lts" --image-project="ubuntu-os-cloud" --zone=us-east1-b
 	# Keeping the --zone here so as to make sure that e2e's can run locally.
 	echo "gcloud compute instances delete descheduler-$master_uuid --zone=us-east1-b --quiet" > $E2E_GCE_HOME/delete_cluster.sh
-       
-	gcloud compute instances create descheduler-$node1_uuid --image="ubuntu-1604-xenial-v20180306" --image-project="ubuntu-os-cloud" --zone=us-east1-b
+
+	gcloud compute instances create descheduler-$node1_uuid --image-family="ubuntu-1804-lts" --image-project="ubuntu-os-cloud" --zone=us-east1-b
 	echo "gcloud compute instances delete descheduler-$node1_uuid --zone=us-east1-b --quiet" >> $E2E_GCE_HOME/delete_cluster.sh
 
-	gcloud compute instances create descheduler-$node2_uuid --image="ubuntu-1604-xenial-v20180306" --image-project="ubuntu-os-cloud" --zone=us-east1-b
-	echo "gcloud compute instances delete descheduler-$node2_uuid --zone=us-east1-b --quiet" >> $E2E_GCE_HOME/delete_cluster.sh
+	gcloud compute instances create descheduler-$node2_uuid --image-family="ubuntu-1804-lts" --image-project="ubuntu-os-cloud" --zone=us-east1-b
+	echo "gcloud compute instances delete descheduler-$node2_uuid --zone=us-east1-c --quiet" >> $E2E_GCE_HOME/delete_cluster.sh
 
 	# Delete the firewall port created for master.
 	echo "gcloud compute firewall-rules delete kubeapiserver-$master_uuid --quiet" >> $E2E_GCE_HOME/delete_cluster.sh
@@ -44,10 +44,10 @@ generate_kubeadm_instance_files() {
 
 
 transfer_install_files() {
-	gcloud compute scp  $E2E_GCE_HOME/kubeadm_preinstall.sh descheduler-$master_uuid:/tmp --zone=us-east1-b
+	gcloud compute scp $E2E_GCE_HOME/kubeadm_preinstall.sh descheduler-$master_uuid:/tmp --zone=us-east1-b
 	gcloud compute scp $E2E_GCE_HOME/kubeadm_install.sh descheduler-$master_uuid:/tmp --zone=us-east1-b
-	gcloud compute scp  $E2E_GCE_HOME/kubeadm_preinstall.sh descheduler-$node1_uuid:/tmp --zone=us-east1-b
-	gcloud compute scp  $E2E_GCE_HOME/kubeadm_preinstall.sh descheduler-$node2_uuid:/tmp --zone=us-east1-b
+	gcloud compute scp $E2E_GCE_HOME/kubeadm_preinstall.sh descheduler-$node1_uuid:/tmp --zone=us-east1-b
+	gcloud compute scp $E2E_GCE_HOME/kubeadm_preinstall.sh descheduler-$node2_uuid:/tmp --zone=us-east1-c
 }
 
 
@@ -55,19 +55,19 @@ install_kube() {
 	# Docker installation.
 	gcloud compute ssh descheduler-$master_uuid --command "sudo apt-get update; sudo apt-get install -y docker.io" --zone=us-east1-b
 	gcloud compute ssh descheduler-$node1_uuid --command "sudo apt-get update; sudo apt-get install -y docker.io" --zone=us-east1-b
-	gcloud compute ssh descheduler-$node2_uuid --command "sudo apt-get update; sudo apt-get install -y docker.io" --zone=us-east1-b
+	gcloud compute ssh descheduler-$node2_uuid --command "sudo apt-get update; sudo apt-get install -y docker.io" --zone=us-east1-c
 	# kubeadm installation.
 	# 1. Transfer files to master, nodes.
 	transfer_install_files
 	# 2. Install kubeadm.
 	#TODO: Add rm /tmp/kubeadm_install.sh
 	# Open port for kube API server	
-	gcloud compute firewall-rules create kubeapiserver-$master_uuid --allow tcp:6443 --source-tags=descheduler-$master_uuid  --source-ranges=0.0.0.0/0 --description="Opening api server port" 
+	gcloud compute firewall-rules create kubeapiserver-$master_uuid --allow tcp:6443 --source-tags=descheduler-$master_uuid  --source-ranges=0.0.0.0/0 --description="Opening api server port"
 
 	gcloud compute ssh descheduler-$master_uuid --command "sudo chmod 755 /tmp/kubeadm_preinstall.sh; sudo /tmp/kubeadm_preinstall.sh" --zone=us-east1-b
 	kubeadm_join_command=$(gcloud compute ssh descheduler-$master_uuid --command "sudo chmod 755 /tmp/kubeadm_install.sh; sudo /tmp/kubeadm_install.sh" --zone=us-east1-b|grep 'kubeadm join')
 	
-	# Copy the kubeconfig file onto /tmp for e2e tests. 
+	# Copy the kubeconfig file onto /tmp for e2e tests.
 	gcloud compute ssh descheduler-$master_uuid --command "sudo cp /etc/kubernetes/admin.conf /tmp; sudo chmod 777 /tmp/admin.conf" --zone=us-east1-b
 	gcloud compute scp descheduler-$master_uuid:/tmp/admin.conf /tmp/admin.conf --zone=us-east1-b
 
@@ -75,16 +75,15 @@ install_kube() {
 	gcloud compute ssh descheduler-$master_uuid --command "sudo kubectl apply -f https://raw.githubusercontent.com/cloudnativelabs/kube-router/master/daemonset/kubeadm-kuberouter.yaml --kubeconfig /etc/kubernetes/admin.conf" --zone=us-east1-b
 	echo $kubeadm_join_command > $E2E_GCE_HOME/kubeadm_join.sh
 
-	# Copy kubeadm_join to every node. 
+	# Copy kubeadm_join to every node.
 	#TODO: Put these in a loop, so that extension becomes possible.	
 	gcloud compute ssh descheduler-$node1_uuid --command "sudo chmod 755 /tmp/kubeadm_preinstall.sh; sudo /tmp/kubeadm_preinstall.sh" --zone=us-east1-b
 	gcloud compute scp $E2E_GCE_HOME/kubeadm_join.sh descheduler-$node1_uuid:/tmp --zone=us-east1-b
 	gcloud compute ssh descheduler-$node1_uuid --command "sudo chmod 755 /tmp/kubeadm_join.sh; sudo /tmp/kubeadm_join.sh" --zone=us-east1-b
 	
-	gcloud compute ssh descheduler-$node2_uuid --command "sudo chmod 755 /tmp/kubeadm_preinstall.sh; sudo /tmp/kubeadm_preinstall.sh" --zone=us-east1-b
-	gcloud compute scp $E2E_GCE_HOME/kubeadm_join.sh descheduler-$node2_uuid:/tmp --zone=us-east1-b
-	gcloud compute ssh descheduler-$node2_uuid --command "sudo chmod 755 /tmp/kubeadm_join.sh; sudo /tmp/kubeadm_join.sh" --zone=us-east1-b
-
+	gcloud compute ssh descheduler-$node2_uuid --command "sudo chmod 755 /tmp/kubeadm_preinstall.sh; sudo /tmp/kubeadm_preinstall.sh" --zone=us-east1-c
+	gcloud compute scp $E2E_GCE_HOME/kubeadm_join.sh descheduler-$node2_uuid:/tmp --zone=us-east1-c
+	gcloud compute ssh descheduler-$node2_uuid --command "sudo chmod 755 /tmp/kubeadm_join.sh; sudo /tmp/kubeadm_join.sh" --zone=us-east1-c
 }
 
 

hack/kind_config.yaml

@@ -3,4 +3,16 @@ apiVersion: kind.x-k8s.io/v1alpha4
 nodes:
 - role: control-plane
 - role: worker
+  kubeadmConfigPatches:
+    - |
+      kind: JoinConfiguration
+      nodeRegistration:
+        kubeletExtraArgs:
+          node-labels: "topology.kubernetes.io/zone=local-a"
 - role: worker
+  kubeadmConfigPatches:
+    - |
+      kind: JoinConfiguration
+      nodeRegistration:
+        kubeletExtraArgs:
+          node-labels: "topology.kubernetes.io/zone=local-b"

hack/tools.go

@@ -1,3 +1,4 @@
+//go:build tools
 // +build tools
 
 /*

hack/update-gofmt.sh

@@ -23,7 +23,7 @@ DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
 
 GO_VERSION=($(go version))
 
-if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.14|go1.15|go1.16') ]]; then
+if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.14|go1.15|go1.16|go1.17') ]]; then
   echo "Unknown go version '${GO_VERSION[2]}', skipping gofmt."
   exit 1
 fi

hack/verify-conversions.sh

@@ -28,7 +28,7 @@ pushd "${DESCHEDULER_ROOT}" > /dev/null 2>&1
 if ! _out="$(diff -Naupr pkg/ "${_deschedulertmp}/pkg/")"; then
     echo "Generated output differs:" >&2
     echo "${_out}" >&2
-    echo "Generated conversions verify failed. Please run ./hack/update-generated-conversions.sh (and commit the result)"
+    echo "Generated conversions verify failed. Please run ./hack/update-generated-conversions.sh"
     exit 1
 fi
 popd > /dev/null 2>&1

hack/verify-deep-copies.sh

@@ -28,7 +28,7 @@ pushd "${DESCHEDULER_ROOT}" > /dev/null 2>&1
 if ! _out="$(diff -Naupr pkg/ "${_deschedulertmp}/pkg/")"; then
     echo "Generated deep-copies output differs:" >&2
     echo "${_out}" >&2
-    echo "Generated deep-copies verify failed. Please run ./hack/update-generated-deep-copies.sh (and commit the result)"
+    echo "Generated deep-copies verify failed. Please run ./hack/update-generated-deep-copies.sh"
     exit 1
 fi
 popd > /dev/null 2>&1

hack/verify-defaulters.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
+_tmpdir="$(mktemp -d "${DESCHEDULER_ROOT}/_tmp/kube-verify.XXXXXX")"
+
+_deschedulertmp="${_tmpdir}"
+mkdir -p "${_deschedulertmp}"
+
+git archive --format=tar --prefix=descheduler/ "$(git write-tree)" | (cd "${_deschedulertmp}" && tar xf -)
+_deschedulertmp="${_deschedulertmp}/descheduler"
+
+pushd "${_deschedulertmp}" > /dev/null 2>&1
+go build -o "${OS_OUTPUT_BINPATH}/defaulter-gen" "k8s.io/code-generator/cmd/defaulter-gen"
+
+${OS_OUTPUT_BINPATH}/defaulter-gen \
+            --go-header-file "hack/boilerplate/boilerplate.go.txt" \
+            --input-dirs "${PRJ_PREFIX}/pkg/apis/componentconfig/v1alpha1,${PRJ_PREFIX}/pkg/api/v1alpha1" \
+            --extra-peer-dirs "${PRJ_PREFIX}/pkg/apis/componentconfig/v1alpha1,${PRJ_PREFIX}/pkg/api/v1alpha1" \
+            --output-file-base zz_generated.defaults
+popd > /dev/null 2>&1
+
+pushd "${DESCHEDULER_ROOT}" > /dev/null 2>&1
+if ! _out="$(diff -Naupr pkg/ "${_deschedulertmp}/pkg/")"; then
+    echo "Generated defaulters output differs:" >&2
+    echo "${_out}" >&2
+    echo "Generated defaulters verify failed. Please run ./hack/update-generated-defaulters.sh"
+fi
+popd > /dev/null 2>&1
+
+echo "Generated Defaulters verified."

hack/verify-gofmt.sh

@@ -23,7 +23,7 @@ DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
 
 GO_VERSION=($(go version))
 
-if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.14|go1.15|go1.16') ]]; then
+if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.14|go1.15|go1.16|go1.17|go1.18') ]]; then
   echo "Unknown go version '${GO_VERSION[2]}', skipping gofmt."
   exit 1
 fi

kubernetes/base/rbac.yaml

@@ -12,7 +12,7 @@ rules:
   verbs: ["get", "watch", "list"]
 - apiGroups: [""]
   resources: ["namespaces"]
-  verbs: ["get", "list"]
+  verbs: ["get", "watch", "list"]
 - apiGroups: [""]
   resources: ["pods"]
   verbs: ["get", "watch", "list", "delete"]
@@ -22,6 +22,13 @@ rules:
 - apiGroups: ["scheduling.k8s.io"]
   resources: ["priorityclasses"]
   verbs: ["get", "watch", "list"]
+- apiGroups: ["coordination.k8s.io"]
+  resources: ["leases"]
+  verbs: ["create"]
+- apiGroups: ["coordination.k8s.io"]
+  resources: ["leases"]
+  resourceNames: ["descheduler"]
+  verbs: ["get", "patch", "delete"]
 ---
 apiVersion: v1
 kind: ServiceAccount
@@ -41,4 +48,3 @@ subjects:
   - name: descheduler-sa
     kind: ServiceAccount
     namespace: kube-system
-

kubernetes/cronjob/cronjob.yaml

@@ -16,7 +16,7 @@ spec:
           priorityClassName: system-cluster-critical
           containers:
           - name: descheduler
-            image: k8s.gcr.io/descheduler/descheduler:v0.21.0
+            image: k8s.gcr.io/descheduler/descheduler:v0.24.0
             volumeMounts:
             - mountPath: /policy-dir
               name: policy-volume
@@ -31,6 +31,14 @@ spec:
               requests:
                 cpu: "500m"
                 memory: "256Mi"
+            livenessProbe:
+              failureThreshold: 3
+              httpGet:
+                path: /healthz
+                port: 10258
+                scheme: HTTPS
+              initialDelaySeconds: 3
+              periodSeconds: 10
             securityContext:
               allowPrivilegeEscalation: false
               capabilities:

kubernetes/deployment/deployment.yaml

@@ -19,7 +19,7 @@ spec:
       serviceAccountName: descheduler-sa
       containers:
         - name: descheduler
-          image: k8s.gcr.io/descheduler/descheduler:v0.21.0
+          image: k8s.gcr.io/descheduler/descheduler:v0.24.0
           imagePullPolicy: IfNotPresent
           command:
             - "/bin/descheduler"
@@ -33,6 +33,14 @@ spec:
           ports:
           - containerPort: 10258
             protocol: TCP
+          livenessProbe:
+            failureThreshold: 3
+            httpGet:
+              path: /healthz
+              port: 10258
+              scheme: HTTPS
+            initialDelaySeconds: 3
+            periodSeconds: 10
           resources:
             requests:
               cpu: 500m

kubernetes/job/job.yaml

@@ -14,7 +14,7 @@ spec:
       priorityClassName: system-cluster-critical
       containers:
         - name: descheduler
-          image: k8s.gcr.io/descheduler/descheduler:v0.21.0
+          image: k8s.gcr.io/descheduler/descheduler:v0.24.0
           volumeMounts:
           - mountPath: /policy-dir
             name: policy-volume
@@ -29,6 +29,14 @@ spec:
             requests:
               cpu: "500m"
               memory: "256Mi"
+          livenessProbe:
+            failureThreshold: 3
+            httpGet:
+              path: /healthz
+              port: 10258
+              scheme: HTTPS
+            initialDelaySeconds: 3
+            periodSeconds: 10
           securityContext:
             allowPrivilegeEscalation: false
             capabilities:

metrics/metrics.go

@@ -34,9 +34,9 @@ var (
 		&metrics.CounterOpts{
 			Subsystem:      DeschedulerSubsystem,
 			Name:           "pods_evicted",
-			Help:           "Number of evicted pods, by the result, by the strategy, by the namespace. 'failed' result means a pod could not be evicted",
+			Help:           "Number of evicted pods, by the result, by the strategy, by the namespace, by the node name. 'error' result means a pod could not be evicted",
 			StabilityLevel: metrics.ALPHA,
-		}, []string{"result", "strategy", "namespace"})
+		}, []string{"result", "strategy", "namespace", "node"})
 
 	buildInfo = metrics.NewGauge(
 		&metrics.GaugeOpts{

pkg/api/types.go

@@ -32,6 +32,9 @@ type DeschedulerPolicy struct {
 	// NodeSelector for a set of nodes to operate over
 	NodeSelector *string
 
+	// EvictFailedBarePods allows pods without ownerReferences and in failed phase to be evicted.
+	EvictFailedBarePods *bool
+
 	// EvictLocalStoragePods allows pods using local storage to be evicted.
 	EvictLocalStoragePods *bool
 
@@ -42,7 +45,10 @@ type DeschedulerPolicy struct {
 	IgnorePVCPods *bool
 
 	// MaxNoOfPodsToEvictPerNode restricts maximum of pods to be evicted per node.
-	MaxNoOfPodsToEvictPerNode *int
+	MaxNoOfPodsToEvictPerNode *uint
+
+	// MaxNoOfPodsToEvictPerNamespace restricts maximum of pods to be evicted per namespace.
+	MaxNoOfPodsToEvictPerNamespace *uint
 }
 
 type StrategyName string
@@ -75,21 +81,25 @@ type StrategyParameters struct {
 	PodsHavingTooManyRestarts         *PodsHavingTooManyRestarts
 	PodLifeTime                       *PodLifeTime
 	RemoveDuplicates                  *RemoveDuplicates
+	FailedPods                        *FailedPods
 	IncludeSoftConstraints            bool
 	Namespaces                        *Namespaces
 	ThresholdPriority                 *int32
 	ThresholdPriorityClassName        string
 	LabelSelector                     *metav1.LabelSelector
 	NodeFit                           bool
+	IncludePreferNoSchedule           bool
+	ExcludedTaints                    []string
 }
 
 type Percentage float64
 type ResourceThresholds map[v1.ResourceName]Percentage
 
 type NodeResourceUtilizationThresholds struct {
-	Thresholds       ResourceThresholds
-	TargetThresholds ResourceThresholds
-	NumberOfNodes    int
+	UseDeviationThresholds bool
+	Thresholds             ResourceThresholds
+	TargetThresholds       ResourceThresholds
+	NumberOfNodes          int
 }
 
 type PodsHavingTooManyRestarts struct {
@@ -105,3 +115,10 @@ type PodLifeTime struct {
 	MaxPodLifeTimeSeconds *uint
 	PodStatusPhases       []string
 }
+
+type FailedPods struct {
+	ExcludeOwnerKinds       []string
+	MinPodLifetimeSeconds   *uint
+	Reasons                 []string
+	IncludingInitContainers bool
+}

pkg/api/v1alpha1/types.go

@@ -32,6 +32,9 @@ type DeschedulerPolicy struct {
 	// NodeSelector for a set of nodes to operate over
 	NodeSelector *string `json:"nodeSelector,omitempty"`
 
+	// EvictFailedBarePods allows pods without ownerReferences and in failed phase to be evicted.
+	EvictFailedBarePods *bool `json:"evictFailedBarePods,omitempty"`
+
 	// EvictLocalStoragePods allows pods using local storage to be evicted.
 	EvictLocalStoragePods *bool `json:"evictLocalStoragePods,omitempty"`
 
@@ -43,6 +46,9 @@ type DeschedulerPolicy struct {
 
 	// MaxNoOfPodsToEvictPerNode restricts maximum of pods to be evicted per node.
 	MaxNoOfPodsToEvictPerNode *int `json:"maxNoOfPodsToEvictPerNode,omitempty"`
+
+	// MaxNoOfPodsToEvictPerNamespace restricts maximum of pods to be evicted per namespace.
+	MaxNoOfPodsToEvictPerNamespace *int `json:"maxNoOfPodsToEvictPerNamespace,omitempty"`
 }
 
 type StrategyName string
@@ -73,21 +79,25 @@ type StrategyParameters struct {
 	PodsHavingTooManyRestarts         *PodsHavingTooManyRestarts         `json:"podsHavingTooManyRestarts,omitempty"`
 	PodLifeTime                       *PodLifeTime                       `json:"podLifeTime,omitempty"`
 	RemoveDuplicates                  *RemoveDuplicates                  `json:"removeDuplicates,omitempty"`
+	FailedPods                        *FailedPods                        `json:"failedPods,omitempty"`
 	IncludeSoftConstraints            bool                               `json:"includeSoftConstraints"`
 	Namespaces                        *Namespaces                        `json:"namespaces"`
 	ThresholdPriority                 *int32                             `json:"thresholdPriority"`
 	ThresholdPriorityClassName        string                             `json:"thresholdPriorityClassName"`
 	LabelSelector                     *metav1.LabelSelector              `json:"labelSelector"`
 	NodeFit                           bool                               `json:"nodeFit"`
+	IncludePreferNoSchedule           bool                               `json:"includePreferNoSchedule"`
+	ExcludedTaints                    []string                           `json:"excludedTaints,omitempty"`
 }
 
 type Percentage float64
 type ResourceThresholds map[v1.ResourceName]Percentage
 
 type NodeResourceUtilizationThresholds struct {
-	Thresholds       ResourceThresholds `json:"thresholds,omitempty"`
-	TargetThresholds ResourceThresholds `json:"targetThresholds,omitempty"`
-	NumberOfNodes    int                `json:"numberOfNodes,omitempty"`
+	UseDeviationThresholds bool               `json:"useDeviationThresholds,omitempty"`
+	Thresholds             ResourceThresholds `json:"thresholds,omitempty"`
+	TargetThresholds       ResourceThresholds `json:"targetThresholds,omitempty"`
+	NumberOfNodes          int                `json:"numberOfNodes,omitempty"`
 }
 
 type PodsHavingTooManyRestarts struct {
@@ -103,3 +113,10 @@ type PodLifeTime struct {
 	MaxPodLifeTimeSeconds *uint    `json:"maxPodLifeTimeSeconds,omitempty"`
 	PodStatusPhases       []string `json:"podStatusPhases,omitempty"`
 }
+
+type FailedPods struct {
+	ExcludeOwnerKinds       []string `json:"excludeOwnerKinds,omitempty"`
+	MinPodLifetimeSeconds   *uint    `json:"minPodLifetimeSeconds,omitempty"`
+	Reasons                 []string `json:"reasons,omitempty"`
+	IncludingInitContainers bool     `json:"includingInitContainers,omitempty"`
+}

pkg/api/v1alpha1/zz_generated.conversion.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -56,6 +57,16 @@ func RegisterConversions(s *runtime.Scheme) error {
 	}); err != nil {
 		return err
 	}
+	if err := s.AddGeneratedConversionFunc((*FailedPods)(nil), (*api.FailedPods)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha1_FailedPods_To_api_FailedPods(a.(*FailedPods), b.(*api.FailedPods), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*api.FailedPods)(nil), (*FailedPods)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_FailedPods_To_v1alpha1_FailedPods(a.(*api.FailedPods), b.(*FailedPods), scope)
+	}); err != nil {
+		return err
+	}
 	if err := s.AddGeneratedConversionFunc((*Namespaces)(nil), (*api.Namespaces)(nil), func(a, b interface{}, scope conversion.Scope) error {
 		return Convert_v1alpha1_Namespaces_To_api_Namespaces(a.(*Namespaces), b.(*api.Namespaces), scope)
 	}); err != nil {
@@ -122,10 +133,24 @@ func RegisterConversions(s *runtime.Scheme) error {
 func autoConvert_v1alpha1_DeschedulerPolicy_To_api_DeschedulerPolicy(in *DeschedulerPolicy, out *api.DeschedulerPolicy, s conversion.Scope) error {
 	out.Strategies = *(*api.StrategyList)(unsafe.Pointer(&in.Strategies))
 	out.NodeSelector = (*string)(unsafe.Pointer(in.NodeSelector))
+	out.EvictFailedBarePods = (*bool)(unsafe.Pointer(in.EvictFailedBarePods))
 	out.EvictLocalStoragePods = (*bool)(unsafe.Pointer(in.EvictLocalStoragePods))
 	out.EvictSystemCriticalPods = (*bool)(unsafe.Pointer(in.EvictSystemCriticalPods))
 	out.IgnorePVCPods = (*bool)(unsafe.Pointer(in.IgnorePVCPods))
-	out.MaxNoOfPodsToEvictPerNode = (*int)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNode))
+	if in.MaxNoOfPodsToEvictPerNode != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNode, &out.MaxNoOfPodsToEvictPerNode
+		*out = new(uint)
+		**out = uint(**in)
+	} else {
+		out.MaxNoOfPodsToEvictPerNode = nil
+	}
+	if in.MaxNoOfPodsToEvictPerNamespace != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
+		*out = new(uint)
+		**out = uint(**in)
+	} else {
+		out.MaxNoOfPodsToEvictPerNamespace = nil
+	}
 	return nil
 }
 
@@ -137,10 +162,24 @@ func Convert_v1alpha1_DeschedulerPolicy_To_api_DeschedulerPolicy(in *Descheduler
 func autoConvert_api_DeschedulerPolicy_To_v1alpha1_DeschedulerPolicy(in *api.DeschedulerPolicy, out *DeschedulerPolicy, s conversion.Scope) error {
 	out.Strategies = *(*StrategyList)(unsafe.Pointer(&in.Strategies))
 	out.NodeSelector = (*string)(unsafe.Pointer(in.NodeSelector))
+	out.EvictFailedBarePods = (*bool)(unsafe.Pointer(in.EvictFailedBarePods))
 	out.EvictLocalStoragePods = (*bool)(unsafe.Pointer(in.EvictLocalStoragePods))
 	out.EvictSystemCriticalPods = (*bool)(unsafe.Pointer(in.EvictSystemCriticalPods))
 	out.IgnorePVCPods = (*bool)(unsafe.Pointer(in.IgnorePVCPods))
-	out.MaxNoOfPodsToEvictPerNode = (*int)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNode))
+	if in.MaxNoOfPodsToEvictPerNode != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNode, &out.MaxNoOfPodsToEvictPerNode
+		*out = new(int)
+		**out = int(**in)
+	} else {
+		out.MaxNoOfPodsToEvictPerNode = nil
+	}
+	if in.MaxNoOfPodsToEvictPerNamespace != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
+		*out = new(int)
+		**out = int(**in)
+	} else {
+		out.MaxNoOfPodsToEvictPerNamespace = nil
+	}
 	return nil
 }
 
@@ -173,6 +212,32 @@ func Convert_api_DeschedulerStrategy_To_v1alpha1_DeschedulerStrategy(in *api.Des
 	return autoConvert_api_DeschedulerStrategy_To_v1alpha1_DeschedulerStrategy(in, out, s)
 }
 
+func autoConvert_v1alpha1_FailedPods_To_api_FailedPods(in *FailedPods, out *api.FailedPods, s conversion.Scope) error {
+	out.ExcludeOwnerKinds = *(*[]string)(unsafe.Pointer(&in.ExcludeOwnerKinds))
+	out.MinPodLifetimeSeconds = (*uint)(unsafe.Pointer(in.MinPodLifetimeSeconds))
+	out.Reasons = *(*[]string)(unsafe.Pointer(&in.Reasons))
+	out.IncludingInitContainers = in.IncludingInitContainers
+	return nil
+}
+
+// Convert_v1alpha1_FailedPods_To_api_FailedPods is an autogenerated conversion function.
+func Convert_v1alpha1_FailedPods_To_api_FailedPods(in *FailedPods, out *api.FailedPods, s conversion.Scope) error {
+	return autoConvert_v1alpha1_FailedPods_To_api_FailedPods(in, out, s)
+}
+
+func autoConvert_api_FailedPods_To_v1alpha1_FailedPods(in *api.FailedPods, out *FailedPods, s conversion.Scope) error {
+	out.ExcludeOwnerKinds = *(*[]string)(unsafe.Pointer(&in.ExcludeOwnerKinds))
+	out.MinPodLifetimeSeconds = (*uint)(unsafe.Pointer(in.MinPodLifetimeSeconds))
+	out.Reasons = *(*[]string)(unsafe.Pointer(&in.Reasons))
+	out.IncludingInitContainers = in.IncludingInitContainers
+	return nil
+}
+
+// Convert_api_FailedPods_To_v1alpha1_FailedPods is an autogenerated conversion function.
+func Convert_api_FailedPods_To_v1alpha1_FailedPods(in *api.FailedPods, out *FailedPods, s conversion.Scope) error {
+	return autoConvert_api_FailedPods_To_v1alpha1_FailedPods(in, out, s)
+}
+
 func autoConvert_v1alpha1_Namespaces_To_api_Namespaces(in *Namespaces, out *api.Namespaces, s conversion.Scope) error {
 	out.Include = *(*[]string)(unsafe.Pointer(&in.Include))
 	out.Exclude = *(*[]string)(unsafe.Pointer(&in.Exclude))
@@ -196,6 +261,7 @@ func Convert_api_Namespaces_To_v1alpha1_Namespaces(in *api.Namespaces, out *Name
 }
 
 func autoConvert_v1alpha1_NodeResourceUtilizationThresholds_To_api_NodeResourceUtilizationThresholds(in *NodeResourceUtilizationThresholds, out *api.NodeResourceUtilizationThresholds, s conversion.Scope) error {
+	out.UseDeviationThresholds = in.UseDeviationThresholds
 	out.Thresholds = *(*api.ResourceThresholds)(unsafe.Pointer(&in.Thresholds))
 	out.TargetThresholds = *(*api.ResourceThresholds)(unsafe.Pointer(&in.TargetThresholds))
 	out.NumberOfNodes = in.NumberOfNodes
@@ -208,6 +274,7 @@ func Convert_v1alpha1_NodeResourceUtilizationThresholds_To_api_NodeResourceUtili
 }
 
 func autoConvert_api_NodeResourceUtilizationThresholds_To_v1alpha1_NodeResourceUtilizationThresholds(in *api.NodeResourceUtilizationThresholds, out *NodeResourceUtilizationThresholds, s conversion.Scope) error {
+	out.UseDeviationThresholds = in.UseDeviationThresholds
 	out.Thresholds = *(*ResourceThresholds)(unsafe.Pointer(&in.Thresholds))
 	out.TargetThresholds = *(*ResourceThresholds)(unsafe.Pointer(&in.TargetThresholds))
 	out.NumberOfNodes = in.NumberOfNodes
@@ -289,12 +356,15 @@ func autoConvert_v1alpha1_StrategyParameters_To_api_StrategyParameters(in *Strat
 	out.PodsHavingTooManyRestarts = (*api.PodsHavingTooManyRestarts)(unsafe.Pointer(in.PodsHavingTooManyRestarts))
 	out.PodLifeTime = (*api.PodLifeTime)(unsafe.Pointer(in.PodLifeTime))
 	out.RemoveDuplicates = (*api.RemoveDuplicates)(unsafe.Pointer(in.RemoveDuplicates))
+	out.FailedPods = (*api.FailedPods)(unsafe.Pointer(in.FailedPods))
 	out.IncludeSoftConstraints = in.IncludeSoftConstraints
 	out.Namespaces = (*api.Namespaces)(unsafe.Pointer(in.Namespaces))
 	out.ThresholdPriority = (*int32)(unsafe.Pointer(in.ThresholdPriority))
 	out.ThresholdPriorityClassName = in.ThresholdPriorityClassName
 	out.LabelSelector = (*v1.LabelSelector)(unsafe.Pointer(in.LabelSelector))
 	out.NodeFit = in.NodeFit
+	out.IncludePreferNoSchedule = in.IncludePreferNoSchedule
+	out.ExcludedTaints = *(*[]string)(unsafe.Pointer(&in.ExcludedTaints))
 	return nil
 }
 
@@ -309,12 +379,15 @@ func autoConvert_api_StrategyParameters_To_v1alpha1_StrategyParameters(in *api.S
 	out.PodsHavingTooManyRestarts = (*PodsHavingTooManyRestarts)(unsafe.Pointer(in.PodsHavingTooManyRestarts))
 	out.PodLifeTime = (*PodLifeTime)(unsafe.Pointer(in.PodLifeTime))
 	out.RemoveDuplicates = (*RemoveDuplicates)(unsafe.Pointer(in.RemoveDuplicates))
+	out.FailedPods = (*FailedPods)(unsafe.Pointer(in.FailedPods))
 	out.IncludeSoftConstraints = in.IncludeSoftConstraints
 	out.Namespaces = (*Namespaces)(unsafe.Pointer(in.Namespaces))
 	out.ThresholdPriority = (*int32)(unsafe.Pointer(in.ThresholdPriority))
 	out.ThresholdPriorityClassName = in.ThresholdPriorityClassName
 	out.LabelSelector = (*v1.LabelSelector)(unsafe.Pointer(in.LabelSelector))
 	out.NodeFit = in.NodeFit
+	out.IncludePreferNoSchedule = in.IncludePreferNoSchedule
+	out.ExcludedTaints = *(*[]string)(unsafe.Pointer(&in.ExcludedTaints))
 	return nil
 }
 

pkg/api/v1alpha1/zz_generated.deepcopy.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -41,6 +42,11 @@ func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
 		*out = new(string)
 		**out = **in
 	}
+	if in.EvictFailedBarePods != nil {
+		in, out := &in.EvictFailedBarePods, &out.EvictFailedBarePods
+		*out = new(bool)
+		**out = **in
+	}
 	if in.EvictLocalStoragePods != nil {
 		in, out := &in.EvictLocalStoragePods, &out.EvictLocalStoragePods
 		*out = new(bool)
@@ -61,6 +67,11 @@ func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
 		*out = new(int)
 		**out = **in
 	}
+	if in.MaxNoOfPodsToEvictPerNamespace != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
+		*out = new(int)
+		**out = **in
+	}
 	return
 }
 
@@ -103,6 +114,37 @@ func (in *DeschedulerStrategy) DeepCopy() *DeschedulerStrategy {
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *FailedPods) DeepCopyInto(out *FailedPods) {
+	*out = *in
+	if in.ExcludeOwnerKinds != nil {
+		in, out := &in.ExcludeOwnerKinds, &out.ExcludeOwnerKinds
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
+	if in.MinPodLifetimeSeconds != nil {
+		in, out := &in.MinPodLifetimeSeconds, &out.MinPodLifetimeSeconds
+		*out = new(uint)
+		**out = **in
+	}
+	if in.Reasons != nil {
+		in, out := &in.Reasons, &out.Reasons
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new FailedPods.
+func (in *FailedPods) DeepCopy() *FailedPods {
+	if in == nil {
+		return nil
+	}
+	out := new(FailedPods)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *Namespaces) DeepCopyInto(out *Namespaces) {
 	*out = *in
@@ -294,6 +336,11 @@ func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
 		*out = new(RemoveDuplicates)
 		(*in).DeepCopyInto(*out)
 	}
+	if in.FailedPods != nil {
+		in, out := &in.FailedPods, &out.FailedPods
+		*out = new(FailedPods)
+		(*in).DeepCopyInto(*out)
+	}
 	if in.Namespaces != nil {
 		in, out := &in.Namespaces, &out.Namespaces
 		*out = new(Namespaces)
@@ -309,6 +356,11 @@ func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
 		*out = new(v1.LabelSelector)
 		(*in).DeepCopyInto(*out)
 	}
+	if in.ExcludedTaints != nil {
+		in, out := &in.ExcludedTaints, &out.ExcludedTaints
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
 	return
 }
 

pkg/api/v1alpha1/zz_generated.defaults.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.

pkg/api/zz_generated.deepcopy.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -41,6 +42,11 @@ func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
 		*out = new(string)
 		**out = **in
 	}
+	if in.EvictFailedBarePods != nil {
+		in, out := &in.EvictFailedBarePods, &out.EvictFailedBarePods
+		*out = new(bool)
+		**out = **in
+	}
 	if in.EvictLocalStoragePods != nil {
 		in, out := &in.EvictLocalStoragePods, &out.EvictLocalStoragePods
 		*out = new(bool)
@@ -58,7 +64,12 @@ func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
 	}
 	if in.MaxNoOfPodsToEvictPerNode != nil {
 		in, out := &in.MaxNoOfPodsToEvictPerNode, &out.MaxNoOfPodsToEvictPerNode
-		*out = new(int)
+		*out = new(uint)
+		**out = **in
+	}
+	if in.MaxNoOfPodsToEvictPerNamespace != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
+		*out = new(uint)
 		**out = **in
 	}
 	return
@@ -103,6 +114,37 @@ func (in *DeschedulerStrategy) DeepCopy() *DeschedulerStrategy {
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *FailedPods) DeepCopyInto(out *FailedPods) {
+	*out = *in
+	if in.ExcludeOwnerKinds != nil {
+		in, out := &in.ExcludeOwnerKinds, &out.ExcludeOwnerKinds
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
+	if in.MinPodLifetimeSeconds != nil {
+		in, out := &in.MinPodLifetimeSeconds, &out.MinPodLifetimeSeconds
+		*out = new(uint)
+		**out = **in
+	}
+	if in.Reasons != nil {
+		in, out := &in.Reasons, &out.Reasons
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new FailedPods.
+func (in *FailedPods) DeepCopy() *FailedPods {
+	if in == nil {
+		return nil
+	}
+	out := new(FailedPods)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *Namespaces) DeepCopyInto(out *Namespaces) {
 	*out = *in
@@ -294,6 +336,11 @@ func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
 		*out = new(RemoveDuplicates)
 		(*in).DeepCopyInto(*out)
 	}
+	if in.FailedPods != nil {
+		in, out := &in.FailedPods, &out.FailedPods
+		*out = new(FailedPods)
+		(*in).DeepCopyInto(*out)
+	}
 	if in.Namespaces != nil {
 		in, out := &in.Namespaces, &out.Namespaces
 		*out = new(Namespaces)
@@ -309,6 +356,11 @@ func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
 		*out = new(v1.LabelSelector)
 		(*in).DeepCopyInto(*out)
 	}
+	if in.ExcludedTaints != nil {
+		in, out := &in.ExcludedTaints, &out.ExcludedTaints
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
 	return
 }
 

pkg/apis/componentconfig/types.go

@@ -53,6 +53,9 @@ type DeschedulerConfiguration struct {
 	// IgnorePVCPods sets whether PVC pods should be allowed to be evicted
 	IgnorePVCPods bool
 
+	// LeaderElection starts Deployment using leader election loop
+	LeaderElection componentbaseconfig.LeaderElectionConfiguration
+
 	// Logging specifies the options of logging.
 	// Refer [Logs Options](https://github.com/kubernetes/component-base/blob/master/logs/options.go) for more information.
 	Logging componentbaseconfig.LoggingConfiguration

pkg/apis/componentconfig/v1alpha1/types.go

@@ -17,9 +17,10 @@ limitations under the License.
 package v1alpha1
 
 import (
-	componentbaseconfig "k8s.io/component-base/config"
 	"time"
 
+	componentbaseconfig "k8s.io/component-base/config"
+
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 )
 
@@ -53,6 +54,9 @@ type DeschedulerConfiguration struct {
 	// IgnorePVCPods sets whether PVC pods should be allowed to be evicted
 	IgnorePVCPods bool `json:"ignorePvcPods,omitempty"`
 
+	// LeaderElection starts Deployment using leader election loop
+	LeaderElection componentbaseconfig.LeaderElectionConfiguration `json:"leaderElection,omitempty"`
+
 	// Logging specifies the options of logging.
 	// Refer [Logs Options](https://github.com/kubernetes/component-base/blob/master/logs/options.go) for more information.
 	Logging componentbaseconfig.LoggingConfiguration `json:"logging,omitempty"`

pkg/apis/componentconfig/v1alpha1/zz_generated.conversion.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -57,6 +58,7 @@ func autoConvert_v1alpha1_DeschedulerConfiguration_To_componentconfig_Deschedule
 	out.MaxNoOfPodsToEvictPerNode = in.MaxNoOfPodsToEvictPerNode
 	out.EvictLocalStoragePods = in.EvictLocalStoragePods
 	out.IgnorePVCPods = in.IgnorePVCPods
+	out.LeaderElection = in.LeaderElection
 	out.Logging = in.Logging
 	return nil
 }
@@ -75,6 +77,7 @@ func autoConvert_componentconfig_DeschedulerConfiguration_To_v1alpha1_Deschedule
 	out.MaxNoOfPodsToEvictPerNode = in.MaxNoOfPodsToEvictPerNode
 	out.EvictLocalStoragePods = in.EvictLocalStoragePods
 	out.IgnorePVCPods = in.IgnorePVCPods
+	out.LeaderElection = in.LeaderElection
 	out.Logging = in.Logging
 	return nil
 }

pkg/apis/componentconfig/v1alpha1/zz_generated.deepcopy.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -28,7 +29,8 @@ import (
 func (in *DeschedulerConfiguration) DeepCopyInto(out *DeschedulerConfiguration) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
-	out.Logging = in.Logging
+	out.LeaderElection = in.LeaderElection
+	in.Logging.DeepCopyInto(&out.Logging)
 	return
 }
 

pkg/apis/componentconfig/v1alpha1/zz_generated.defaults.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.

pkg/apis/componentconfig/zz_generated.deepcopy.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2021 The Kubernetes Authors.
+Copyright 2022 The Kubernetes Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -28,7 +29,8 @@ import (
 func (in *DeschedulerConfiguration) DeepCopyInto(out *DeschedulerConfiguration) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
-	out.Logging = in.Logging
+	out.LeaderElection = in.LeaderElection
+	in.Logging.DeepCopyInto(&out.Logging)
 	return
 }
 

pkg/descheduler/descheduler.go

@@ -19,14 +19,22 @@ package descheduler
 import (
 	"context"
 	"fmt"
-	"sigs.k8s.io/descheduler/pkg/descheduler/strategies/nodeutilization"
 
 	v1 "k8s.io/api/core/v1"
-	clientset "k8s.io/client-go/kubernetes"
-	"k8s.io/klog/v2"
-
+	policy "k8s.io/api/policy/v1beta1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/client-go/informers"
+	clientset "k8s.io/client-go/kubernetes"
+	fakeclientset "k8s.io/client-go/kubernetes/fake"
+	core "k8s.io/client-go/testing"
+	"k8s.io/klog/v2"
+
+	corev1informers "k8s.io/client-go/informers/core/v1"
+	schedulingv1informers "k8s.io/client-go/informers/scheduling/v1"
+
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
 	"sigs.k8s.io/descheduler/metrics"
 	"sigs.k8s.io/descheduler/pkg/api"
@@ -34,13 +42,14 @@ import (
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
 	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
 	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/descheduler/strategies"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies/nodeutilization"
 )
 
-func Run(rs *options.DeschedulerServer) error {
+func Run(ctx context.Context, rs *options.DeschedulerServer) error {
 	metrics.Register()
 
-	ctx := context.Background()
 	rsclient, err := client.CreateClient(rs.KubeconfigFile)
 	if err != nil {
 		return err
@@ -60,18 +69,123 @@ func Run(rs *options.DeschedulerServer) error {
 		return err
 	}
 
-	stopChannel := make(chan struct{})
-	return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion, stopChannel)
+	runFn := func() error {
+		return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion)
+	}
+
+	if rs.LeaderElection.LeaderElect && rs.DeschedulingInterval.Seconds() == 0 {
+		return fmt.Errorf("leaderElection must be used with deschedulingInterval")
+	}
+
+	if rs.LeaderElection.LeaderElect && !rs.DryRun {
+		if err := NewLeaderElection(runFn, rsclient, &rs.LeaderElection, ctx); err != nil {
+			return fmt.Errorf("leaderElection: %w", err)
+		}
+		return nil
+	}
+
+	return runFn()
 }
 
-type strategyFunction func(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor)
+type strategyFunction func(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc)
 
-func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string, stopChannel chan struct{}) error {
+func cachedClient(
+	realClient clientset.Interface,
+	podInformer corev1informers.PodInformer,
+	nodeInformer corev1informers.NodeInformer,
+	namespaceInformer corev1informers.NamespaceInformer,
+	priorityClassInformer schedulingv1informers.PriorityClassInformer,
+) (clientset.Interface, error) {
+	fakeClient := fakeclientset.NewSimpleClientset()
+	// simulate a pod eviction by deleting a pod
+	fakeClient.PrependReactor("create", "pods", func(action core.Action) (bool, runtime.Object, error) {
+		if action.GetSubresource() == "eviction" {
+			createAct, matched := action.(core.CreateActionImpl)
+			if !matched {
+				return false, nil, fmt.Errorf("unable to convert action to core.CreateActionImpl")
+			}
+			eviction, matched := createAct.Object.(*policy.Eviction)
+			if !matched {
+				return false, nil, fmt.Errorf("unable to convert action object into *policy.Eviction")
+			}
+			if err := fakeClient.Tracker().Delete(action.GetResource(), eviction.GetNamespace(), eviction.GetName()); err != nil {
+				return false, nil, fmt.Errorf("unable to delete pod %v/%v: %v", eviction.GetNamespace(), eviction.GetName(), err)
+			}
+			return true, nil, nil
+		}
+		// fallback to the default reactor
+		return false, nil, nil
+	})
+
+	klog.V(3).Infof("Pulling resources for the cached client from the cluster")
+	pods, err := podInformer.Lister().List(labels.Everything())
+	if err != nil {
+		return nil, fmt.Errorf("unable to list pods: %v", err)
+	}
+
+	for _, item := range pods {
+		if _, err := fakeClient.CoreV1().Pods(item.Namespace).Create(context.TODO(), item, metav1.CreateOptions{}); err != nil {
+			return nil, fmt.Errorf("unable to copy pod: %v", err)
+		}
+	}
+
+	nodes, err := nodeInformer.Lister().List(labels.Everything())
+	if err != nil {
+		return nil, fmt.Errorf("unable to list nodes: %v", err)
+	}
+
+	for _, item := range nodes {
+		if _, err := fakeClient.CoreV1().Nodes().Create(context.TODO(), item, metav1.CreateOptions{}); err != nil {
+			return nil, fmt.Errorf("unable to copy node: %v", err)
+		}
+	}
+
+	namespaces, err := namespaceInformer.Lister().List(labels.Everything())
+	if err != nil {
+		return nil, fmt.Errorf("unable to list namespaces: %v", err)
+	}
+
+	for _, item := range namespaces {
+		if _, err := fakeClient.CoreV1().Namespaces().Create(context.TODO(), item, metav1.CreateOptions{}); err != nil {
+			return nil, fmt.Errorf("unable to copy node: %v", err)
+		}
+	}
+
+	priorityClasses, err := priorityClassInformer.Lister().List(labels.Everything())
+	if err != nil {
+		return nil, fmt.Errorf("unable to list priorityclasses: %v", err)
+	}
+
+	for _, item := range priorityClasses {
+		if _, err := fakeClient.SchedulingV1().PriorityClasses().Create(context.TODO(), item, metav1.CreateOptions{}); err != nil {
+			return nil, fmt.Errorf("unable to copy priorityclass: %v", err)
+		}
+	}
+
+	return fakeClient, nil
+}
+
+func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string) error {
 	sharedInformerFactory := informers.NewSharedInformerFactory(rs.Client, 0)
 	nodeInformer := sharedInformerFactory.Core().V1().Nodes()
+	podInformer := sharedInformerFactory.Core().V1().Pods()
+	namespaceInformer := sharedInformerFactory.Core().V1().Namespaces()
+	priorityClassInformer := sharedInformerFactory.Scheduling().V1().PriorityClasses()
 
-	sharedInformerFactory.Start(stopChannel)
-	sharedInformerFactory.WaitForCacheSync(stopChannel)
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	// create the informers
+	namespaceInformer.Informer()
+	priorityClassInformer.Informer()
+
+	getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+	if err != nil {
+		return fmt.Errorf("build get pods assigned to node function error: %v", err)
+	}
+
+	sharedInformerFactory.Start(ctx.Done())
+	sharedInformerFactory.WaitForCacheSync(ctx.Done())
 
 	strategyFuncs := map[api.StrategyName]strategyFunction{
 		"RemoveDuplicates":                            strategies.RemoveDuplicatePods,
@@ -83,18 +197,27 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 		"RemovePodsHavingTooManyRestarts":             strategies.RemovePodsHavingTooManyRestarts,
 		"PodLifeTime":                                 strategies.PodLifeTime,
 		"RemovePodsViolatingTopologySpreadConstraint": strategies.RemovePodsViolatingTopologySpreadConstraint,
+		"RemoveFailedPods":                            strategies.RemoveFailedPods,
 	}
 
-	nodeSelector := rs.NodeSelector
+	var nodeSelector string
 	if deschedulerPolicy.NodeSelector != nil {
 		nodeSelector = *deschedulerPolicy.NodeSelector
 	}
 
-	evictLocalStoragePods := rs.EvictLocalStoragePods
+	var evictLocalStoragePods bool
 	if deschedulerPolicy.EvictLocalStoragePods != nil {
 		evictLocalStoragePods = *deschedulerPolicy.EvictLocalStoragePods
 	}
 
+	evictBarePods := false
+	if deschedulerPolicy.EvictFailedBarePods != nil {
+		evictBarePods = *deschedulerPolicy.EvictFailedBarePods
+		if evictBarePods {
+			klog.V(1).InfoS("Warning: EvictFailedBarePods is set to True. This could cause eviction of pods without ownerReferences.")
+		}
+	}
+
 	evictSystemCriticalPods := false
 	if deschedulerPolicy.EvictSystemCriticalPods != nil {
 		evictSystemCriticalPods = *deschedulerPolicy.EvictSystemCriticalPods
@@ -108,40 +231,70 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 		ignorePvcPods = *deschedulerPolicy.IgnorePVCPods
 	}
 
-	maxNoOfPodsToEvictPerNode := rs.MaxNoOfPodsToEvictPerNode
-	if deschedulerPolicy.MaxNoOfPodsToEvictPerNode != nil {
-		maxNoOfPodsToEvictPerNode = *deschedulerPolicy.MaxNoOfPodsToEvictPerNode
-	}
-
-	wait.Until(func() {
+	wait.NonSlidingUntil(func() {
 		nodes, err := nodeutil.ReadyNodes(ctx, rs.Client, nodeInformer, nodeSelector)
 		if err != nil {
 			klog.V(1).InfoS("Unable to get ready nodes", "err", err)
-			close(stopChannel)
+			cancel()
 			return
 		}
 
 		if len(nodes) <= 1 {
 			klog.V(1).InfoS("The cluster size is 0 or 1 meaning eviction causes service disruption or degradation. So aborting..")
-			close(stopChannel)
+			cancel()
 			return
 		}
 
+		var podEvictorClient clientset.Interface
+		// When the dry mode is enable, collect all the relevant objects (mostly pods) under a fake client.
+		// So when evicting pods while running multiple strategies in a row have the cummulative effect
+		// as is when evicting pods for real.
+		if rs.DryRun {
+			klog.V(3).Infof("Building a cached client from the cluster for the dry run")
+			// Create a new cache so we start from scratch without any leftovers
+			fakeClient, err := cachedClient(rs.Client, podInformer, nodeInformer, namespaceInformer, priorityClassInformer)
+			if err != nil {
+				klog.Error(err)
+				return
+			}
+
+			fakeSharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			getPodsAssignedToNode, err = podutil.BuildGetPodsAssignedToNodeFunc(fakeSharedInformerFactory.Core().V1().Pods())
+			if err != nil {
+				klog.Errorf("build get pods assigned to node function error: %v", err)
+				return
+			}
+
+			fakeCtx, cncl := context.WithCancel(context.TODO())
+			defer cncl()
+			fakeSharedInformerFactory.Start(fakeCtx.Done())
+			fakeSharedInformerFactory.WaitForCacheSync(fakeCtx.Done())
+
+			podEvictorClient = fakeClient
+		} else {
+			podEvictorClient = rs.Client
+		}
+
+		klog.V(3).Infof("Building a pod evictor")
 		podEvictor := evictions.NewPodEvictor(
-			rs.Client,
+			podEvictorClient,
 			evictionPolicyGroupVersion,
 			rs.DryRun,
-			maxNoOfPodsToEvictPerNode,
+			deschedulerPolicy.MaxNoOfPodsToEvictPerNode,
+			deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace,
 			nodes,
+			getPodsAssignedToNode,
 			evictLocalStoragePods,
 			evictSystemCriticalPods,
 			ignorePvcPods,
+			evictBarePods,
+			!rs.DisableMetrics,
 		)
 
 		for name, strategy := range deschedulerPolicy.Strategies {
 			if f, ok := strategyFuncs[name]; ok {
 				if strategy.Enabled {
-					f(ctx, rs.Client, strategy, nodes, podEvictor)
+					f(ctx, rs.Client, strategy, nodes, podEvictor, getPodsAssignedToNode)
 				}
 			} else {
 				klog.ErrorS(fmt.Errorf("unknown strategy name"), "skipping strategy", "strategy", name)
@@ -152,9 +305,9 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 
 		// If there was no interval specified, send a signal to the stopChannel to end the wait.Until loop after 1 iteration
 		if rs.DeschedulingInterval.Seconds() == 0 {
-			close(stopChannel)
+			cancel()
 		}
-	}, rs.DeschedulingInterval, stopChannel)
+	}, rs.DeschedulingInterval, ctx.Done())
 
 	return nil
 }

pkg/descheduler/descheduler_test.go

@@ -35,9 +35,6 @@ func TestTaintsUpdated(t *testing.T) {
 		},
 	}
 
-	stopChannel := make(chan struct{})
-	defer close(stopChannel)
-
 	rs, err := options.NewDeschedulerServer()
 	if err != nil {
 		t.Fatalf("Unable to initialize server: %v", err)
@@ -47,7 +44,7 @@ func TestTaintsUpdated(t *testing.T) {
 	errChan := make(chan error, 1)
 	defer close(errChan)
 	go func() {
-		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1", stopChannel)
+		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1")
 		errChan <- err
 	}()
 	select {
@@ -55,7 +52,7 @@ func TestTaintsUpdated(t *testing.T) {
 		if err != nil {
 			t.Fatalf("Unable to run descheduler strategies: %v", err)
 		}
-	case <-time.After(300 * time.Millisecond):
+	case <-time.After(1 * time.Second):
 		// Wait for few cycles and then verify the only pod still exists
 	}
 
@@ -101,3 +98,69 @@ func TestTaintsUpdated(t *testing.T) {
 		t.Fatalf("Unable to evict pod, node taint did not get propagated to descheduler strategies")
 	}
 }
+
+func TestRootCancel(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	client := fakeclientset.NewSimpleClientset(n1, n2)
+	dp := &api.DeschedulerPolicy{
+		Strategies: api.StrategyList{}, // no strategies needed for this test
+	}
+
+	rs, err := options.NewDeschedulerServer()
+	if err != nil {
+		t.Fatalf("Unable to initialize server: %v", err)
+	}
+	rs.Client = client
+	rs.DeschedulingInterval = 100 * time.Millisecond
+	errChan := make(chan error, 1)
+	defer close(errChan)
+
+	go func() {
+		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1")
+		errChan <- err
+	}()
+	cancel()
+	select {
+	case err := <-errChan:
+		if err != nil {
+			t.Fatalf("Unable to run descheduler strategies: %v", err)
+		}
+	case <-time.After(1 * time.Second):
+		t.Fatal("Root ctx should have canceled immediately")
+	}
+}
+
+func TestRootCancelWithNoInterval(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	client := fakeclientset.NewSimpleClientset(n1, n2)
+	dp := &api.DeschedulerPolicy{
+		Strategies: api.StrategyList{}, // no strategies needed for this test
+	}
+
+	rs, err := options.NewDeschedulerServer()
+	if err != nil {
+		t.Fatalf("Unable to initialize server: %v", err)
+	}
+	rs.Client = client
+	rs.DeschedulingInterval = 0
+	errChan := make(chan error, 1)
+	defer close(errChan)
+
+	go func() {
+		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1")
+		errChan <- err
+	}()
+	cancel()
+	select {
+	case err := <-errChan:
+		if err != nil {
+			t.Fatalf("Unable to run descheduler strategies: %v", err)
+		}
+	case <-time.After(1 * time.Second):
+		t.Fatal("Root ctx should have canceled immediately")
+	}
+}

pkg/descheduler/evictions/evictions.go

@@ -45,57 +45,73 @@ const (
 )
 
 // nodePodEvictedCount keeps count of pods evicted on node
-type nodePodEvictedCount map[*v1.Node]int
+type nodePodEvictedCount map[*v1.Node]uint
+type namespacePodEvictCount map[string]uint
 
 type PodEvictor struct {
-	client                  clientset.Interface
-	nodes                   []*v1.Node
-	policyGroupVersion      string
-	dryRun                  bool
-	maxPodsToEvictPerNode   int
-	nodepodCount            nodePodEvictedCount
-	evictLocalStoragePods   bool
-	evictSystemCriticalPods bool
-	ignorePvcPods           bool
+	client                     clientset.Interface
+	nodes                      []*v1.Node
+	nodeIndexer                podutil.GetPodsAssignedToNodeFunc
+	policyGroupVersion         string
+	dryRun                     bool
+	maxPodsToEvictPerNode      *uint
+	maxPodsToEvictPerNamespace *uint
+	nodepodCount               nodePodEvictedCount
+	namespacePodCount          namespacePodEvictCount
+	evictFailedBarePods        bool
+	evictLocalStoragePods      bool
+	evictSystemCriticalPods    bool
+	ignorePvcPods              bool
+	metricsEnabled             bool
 }
 
 func NewPodEvictor(
 	client clientset.Interface,
 	policyGroupVersion string,
 	dryRun bool,
-	maxPodsToEvictPerNode int,
+	maxPodsToEvictPerNode *uint,
+	maxPodsToEvictPerNamespace *uint,
 	nodes []*v1.Node,
+	nodeIndexer podutil.GetPodsAssignedToNodeFunc,
 	evictLocalStoragePods bool,
 	evictSystemCriticalPods bool,
 	ignorePvcPods bool,
+	evictFailedBarePods bool,
+	metricsEnabled bool,
 ) *PodEvictor {
 	var nodePodCount = make(nodePodEvictedCount)
+	var namespacePodCount = make(namespacePodEvictCount)
 	for _, node := range nodes {
 		// Initialize podsEvicted till now with 0.
 		nodePodCount[node] = 0
 	}
 
 	return &PodEvictor{
-		client:                  client,
-		nodes:                   nodes,
-		policyGroupVersion:      policyGroupVersion,
-		dryRun:                  dryRun,
-		maxPodsToEvictPerNode:   maxPodsToEvictPerNode,
-		nodepodCount:            nodePodCount,
-		evictLocalStoragePods:   evictLocalStoragePods,
-		evictSystemCriticalPods: evictSystemCriticalPods,
-		ignorePvcPods:           ignorePvcPods,
+		client:                     client,
+		nodes:                      nodes,
+		nodeIndexer:                nodeIndexer,
+		policyGroupVersion:         policyGroupVersion,
+		dryRun:                     dryRun,
+		maxPodsToEvictPerNode:      maxPodsToEvictPerNode,
+		maxPodsToEvictPerNamespace: maxPodsToEvictPerNamespace,
+		nodepodCount:               nodePodCount,
+		namespacePodCount:          namespacePodCount,
+		evictLocalStoragePods:      evictLocalStoragePods,
+		evictSystemCriticalPods:    evictSystemCriticalPods,
+		evictFailedBarePods:        evictFailedBarePods,
+		ignorePvcPods:              ignorePvcPods,
+		metricsEnabled:             metricsEnabled,
 	}
 }
 
 // NodeEvicted gives a number of pods evicted for node
-func (pe *PodEvictor) NodeEvicted(node *v1.Node) int {
+func (pe *PodEvictor) NodeEvicted(node *v1.Node) uint {
 	return pe.nodepodCount[node]
 }
 
 // TotalEvicted gives a number of pods evicted through all nodes
-func (pe *PodEvictor) TotalEvicted() int {
-	var total int
+func (pe *PodEvictor) TotalEvicted() uint {
+	var total uint
 	for _, count := range pe.nodepodCount {
 		total += count
 	}
@@ -110,38 +126,51 @@ func (pe *PodEvictor) EvictPod(ctx context.Context, pod *v1.Pod, node *v1.Node,
 	if len(reasons) > 0 {
 		reason += " (" + strings.Join(reasons, ", ") + ")"
 	}
-	if pe.maxPodsToEvictPerNode > 0 && pe.nodepodCount[node]+1 > pe.maxPodsToEvictPerNode {
-		metrics.PodsEvicted.With(map[string]string{"result": "maximum number reached", "strategy": strategy, "namespace": pod.Namespace}).Inc()
-		return false, fmt.Errorf("Maximum number %v of evicted pods per %q node reached", pe.maxPodsToEvictPerNode, node.Name)
+	if pe.maxPodsToEvictPerNode != nil && pe.nodepodCount[node]+1 > *pe.maxPodsToEvictPerNode {
+		if pe.metricsEnabled {
+			metrics.PodsEvicted.With(map[string]string{"result": "maximum number of pods per node reached", "strategy": strategy, "namespace": pod.Namespace, "node": node.Name}).Inc()
+		}
+		return false, fmt.Errorf("Maximum number %v of evicted pods per %q node reached", *pe.maxPodsToEvictPerNode, node.Name)
 	}
 
-	err := evictPod(ctx, pe.client, pod, pe.policyGroupVersion, pe.dryRun)
+	if pe.maxPodsToEvictPerNamespace != nil && pe.namespacePodCount[pod.Namespace]+1 > *pe.maxPodsToEvictPerNamespace {
+		if pe.metricsEnabled {
+			metrics.PodsEvicted.With(map[string]string{"result": "maximum number of pods per namespace reached", "strategy": strategy, "namespace": pod.Namespace, "node": node.Name}).Inc()
+		}
+		return false, fmt.Errorf("Maximum number %v of evicted pods per %q namespace reached", *pe.maxPodsToEvictPerNamespace, pod.Namespace)
+	}
+
+	err := evictPod(ctx, pe.client, pod, pe.policyGroupVersion)
 	if err != nil {
 		// err is used only for logging purposes
 		klog.ErrorS(err, "Error evicting pod", "pod", klog.KObj(pod), "reason", reason)
-		metrics.PodsEvicted.With(map[string]string{"result": "error", "strategy": strategy, "namespace": pod.Namespace}).Inc()
+		if pe.metricsEnabled {
+			metrics.PodsEvicted.With(map[string]string{"result": "error", "strategy": strategy, "namespace": pod.Namespace, "node": node.Name}).Inc()
+		}
 		return false, nil
 	}
 
 	pe.nodepodCount[node]++
+	pe.namespacePodCount[pod.Namespace]++
+
+	if pe.metricsEnabled {
+		metrics.PodsEvicted.With(map[string]string{"result": "success", "strategy": strategy, "namespace": pod.Namespace, "node": node.Name}).Inc()
+	}
+
 	if pe.dryRun {
-		klog.V(1).InfoS("Evicted pod in dry run mode", "pod", klog.KObj(pod), "reason", reason)
+		klog.V(1).InfoS("Evicted pod in dry run mode", "pod", klog.KObj(pod), "reason", reason, "strategy", strategy, "node", node.Name)
 	} else {
-		klog.V(1).InfoS("Evicted pod", "pod", klog.KObj(pod), "reason", reason)
+		klog.V(1).InfoS("Evicted pod", "pod", klog.KObj(pod), "reason", reason, "strategy", strategy, "node", node.Name)
 		eventBroadcaster := record.NewBroadcaster()
 		eventBroadcaster.StartStructuredLogging(3)
 		eventBroadcaster.StartRecordingToSink(&clientcorev1.EventSinkImpl{Interface: pe.client.CoreV1().Events(pod.Namespace)})
 		r := eventBroadcaster.NewRecorder(scheme.Scheme, v1.EventSource{Component: "sigs.k8s.io.descheduler"})
 		r.Event(pod, v1.EventTypeNormal, "Descheduled", fmt.Sprintf("pod evicted by sigs.k8s.io/descheduler%s", reason))
-		metrics.PodsEvicted.With(map[string]string{"result": "success", "strategy": strategy, "namespace": pod.Namespace}).Inc()
 	}
 	return true, nil
 }
 
-func evictPod(ctx context.Context, client clientset.Interface, pod *v1.Pod, policyGroupVersion string, dryRun bool) error {
-	if dryRun {
-		return nil
-	}
+func evictPod(ctx context.Context, client clientset.Interface, pod *v1.Pod, policyGroupVersion string) error {
 	deleteOptions := &metav1.DeleteOptions{}
 	// GracePeriodSeconds ?
 	eviction := &policy.Eviction{
@@ -215,6 +244,25 @@ func (pe *PodEvictor) Evictable(opts ...func(opts *Options)) *evictable {
 	}
 
 	ev := &evictable{}
+	if pe.evictFailedBarePods {
+		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+			ownerRefList := podutil.OwnerRef(pod)
+			// Enable evictFailedBarePods to evict bare pods in failed phase
+			if len(ownerRefList) == 0 && pod.Status.Phase != v1.PodFailed {
+				return fmt.Errorf("pod does not have any ownerRefs and is not in failed phase")
+			}
+			return nil
+		})
+	} else {
+		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+			ownerRefList := podutil.OwnerRef(pod)
+			// Moved from IsEvictable function for backward compatibility
+			if len(ownerRefList) == 0 {
+				return fmt.Errorf("pod does not have any ownerRefs")
+			}
+			return nil
+		})
+	}
 	if !pe.evictSystemCriticalPods {
 		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
 			// Moved from IsEvictable function to allow for disabling
@@ -251,7 +299,7 @@ func (pe *PodEvictor) Evictable(opts ...func(opts *Options)) *evictable {
 	}
 	if options.nodeFit {
 		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
-			if !nodeutil.PodFitsAnyOtherNode(pod, pe.nodes) {
+			if !nodeutil.PodFitsAnyOtherNode(pe.nodeIndexer, pod, pe.nodes) {
 				return fmt.Errorf("pod does not fit on any other node because of nodeSelector(s), Taint(s), or nodes marked as unschedulable")
 			}
 			return nil
@@ -278,10 +326,6 @@ func (ev *evictable) IsEvictable(pod *v1.Pod) bool {
 		checkErrs = append(checkErrs, fmt.Errorf("pod is a DaemonSet pod"))
 	}
 
-	if len(ownerRefList) == 0 {
-		checkErrs = append(checkErrs, fmt.Errorf("pod does not have any ownerrefs"))
-	}
-
 	if utils.IsMirrorPod(pod) {
 		checkErrs = append(checkErrs, fmt.Errorf("pod is a mirror pod"))
 	}
@@ -290,6 +334,10 @@ func (ev *evictable) IsEvictable(pod *v1.Pod) bool {
 		checkErrs = append(checkErrs, fmt.Errorf("pod is a static pod"))
 	}
 
+	if utils.IsPodTerminating(pod) {
+		checkErrs = append(checkErrs, fmt.Errorf("pod is terminating"))
+	}
+
 	for _, c := range ev.constraints {
 		if err := c(pod); err != nil {
 			checkErrs = append(checkErrs, err)

pkg/descheduler/evictions/evictions_test.go

@@ -21,8 +21,10 @@ import (
 	"testing"
 
 	v1 "k8s.io/api/core/v1"
+	policyv1 "k8s.io/api/policy/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
 	core "k8s.io/client-go/testing"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
@@ -62,7 +64,7 @@ func TestEvictPod(t *testing.T) {
 		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
 			return true, &v1.PodList{Items: test.pods}, nil
 		})
-		got := evictPod(ctx, fakeClient, test.pod, "v1", false)
+		got := evictPod(ctx, fakeClient, test.pod, "v1")
 		if got != test.want {
 			t.Errorf("Test error for Desc: %s. Expected %v pod eviction to be %v, got %v", test.description, test.pod.Name, test.want, got)
 		}
@@ -80,9 +82,10 @@ func TestIsEvictable(t *testing.T) {
 	nodeLabelKey := "datacenter"
 	nodeLabelValue := "east"
 	type testCase struct {
-		pod                     *v1.Pod
+		description             string
+		pods                    []*v1.Pod
 		nodes                   []*v1.Node
-		runBefore               func(*v1.Pod, []*v1.Node)
+		evictFailedBarePods     bool
 		evictLocalStoragePods   bool
 		evictSystemCriticalPods bool
 		priorityThreshold       *int32
@@ -91,241 +94,309 @@ func TestIsEvictable(t *testing.T) {
 	}
 
 	testCases := []testCase{
-		{ // Normal pod eviction with normal ownerRefs
-			pod: test.BuildTestPod("p1", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+		{
+			description: "Failed pod eviction with no ownerRefs",
+			pods: []*v1.Pod{
+				test.BuildTestPod("bare_pod_failed", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Status.Phase = v1.PodFailed
+				}),
+			},
+			evictFailedBarePods: false,
+			result:              false,
+		}, {
+			description:         "Normal pod eviction with no ownerRefs and evictFailedBarePods enabled",
+			pods:                []*v1.Pod{test.BuildTestPod("bare_pod", 400, 0, n1.Name, nil)},
+			evictFailedBarePods: true,
+			result:              false,
+		}, {
+			description: "Failed pod eviction with no ownerRefs",
+			pods: []*v1.Pod{
+				test.BuildTestPod("bare_pod_failed_but_can_be_evicted", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Status.Phase = v1.PodFailed
+				}),
+			},
+			evictFailedBarePods: true,
+			result:              true,
+		}, {
+			description: "Normal pod eviction with normal ownerRefs",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Normal pod eviction with normal ownerRefs and descheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p2", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+		}, {
+			description: "Normal pod eviction with normal ownerRefs and descheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p2", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Normal pod eviction with replicaSet ownerRefs
-			pod: test.BuildTestPod("p3", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
+		}, {
+			description: "Normal pod eviction with replicaSet ownerRefs",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p3", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Normal pod eviction with replicaSet ownerRefs and descheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p4", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
+		}, {
+			description: "Normal pod eviction with replicaSet ownerRefs and descheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p4", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Normal pod eviction with statefulSet ownerRefs
-			pod: test.BuildTestPod("p18", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetStatefulSetOwnerRefList()
+		}, {
+			description: "Normal pod eviction with statefulSet ownerRefs",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p18", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Normal pod eviction with statefulSet ownerRefs and descheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p19", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.ObjectMeta.OwnerReferences = test.GetStatefulSetOwnerRefList()
+		}, {
+			description: "Normal pod eviction with statefulSet ownerRefs and descheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p19", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.ObjectMeta.OwnerReferences = test.GetStatefulSetOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Pod not evicted because it is bound to a PV and evictLocalStoragePods = false
-			pod: test.BuildTestPod("p5", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Spec.Volumes = []v1.Volume{
-					{
-						Name: "sample",
-						VolumeSource: v1.VolumeSource{
-							HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
-							EmptyDir: &v1.EmptyDirVolumeSource{
-								SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+		}, {
+			description: "Pod not evicted because it is bound to a PV and evictLocalStoragePods = false",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p5", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.Volumes = []v1.Volume{
+						{
+							Name: "sample",
+							VolumeSource: v1.VolumeSource{
+								HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
+								EmptyDir: &v1.EmptyDirVolumeSource{
+									SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+							},
 						},
-					},
-				}
+					}
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  false,
-		}, { // Pod is evicted because it is bound to a PV and evictLocalStoragePods = true
-			pod: test.BuildTestPod("p6", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Spec.Volumes = []v1.Volume{
-					{
-						Name: "sample",
-						VolumeSource: v1.VolumeSource{
-							HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
-							EmptyDir: &v1.EmptyDirVolumeSource{
-								SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+		}, {
+			description: "Pod is evicted because it is bound to a PV and evictLocalStoragePods = true",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p6", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.Volumes = []v1.Volume{
+						{
+							Name: "sample",
+							VolumeSource: v1.VolumeSource{
+								HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
+								EmptyDir: &v1.EmptyDirVolumeSource{
+									SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+							},
 						},
-					},
-				}
+					}
+				}),
 			},
 			evictLocalStoragePods:   true,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Pod is evicted because it is bound to a PV and evictLocalStoragePods = false, but it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p7", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Spec.Volumes = []v1.Volume{
-					{
-						Name: "sample",
-						VolumeSource: v1.VolumeSource{
-							HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
-							EmptyDir: &v1.EmptyDirVolumeSource{
-								SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+		}, {
+			description: "Pod is evicted because it is bound to a PV and evictLocalStoragePods = false, but it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p7", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.Volumes = []v1.Volume{
+						{
+							Name: "sample",
+							VolumeSource: v1.VolumeSource{
+								HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
+								EmptyDir: &v1.EmptyDirVolumeSource{
+									SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+							},
 						},
-					},
-				}
+					}
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Pod not evicted becasuse it is part of a daemonSet
-			pod: test.BuildTestPod("p8", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
+		}, {
+			description: "Pod not evicted becasuse it is part of a daemonSet",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p8", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  false,
-		}, { // Pod is evicted becasuse it is part of a daemonSet, but it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p9", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
+		}, {
+			description: "Pod is evicted becasuse it is part of a daemonSet, but it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p9", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Pod not evicted becasuse it is a mirror pod
-			pod: test.BuildTestPod("p10", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Annotations = test.GetMirrorPodAnnotation()
+		}, {
+			description: "Pod not evicted becasuse it is a mirror poddsa",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p10", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Annotations = test.GetMirrorPodAnnotation()
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  false,
-		}, { // Pod is evicted becasuse it is a mirror pod, but it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p11", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Annotations = test.GetMirrorPodAnnotation()
-				pod.Annotations["descheduler.alpha.kubernetes.io/evict"] = "true"
+		}, {
+			description: "Pod is evicted becasuse it is a mirror pod, but it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p11", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Annotations = test.GetMirrorPodAnnotation()
+					pod.Annotations["descheduler.alpha.kubernetes.io/evict"] = "true"
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Pod not evicted becasuse it has system critical priority
-			pod: test.BuildTestPod("p12", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				priority := utils.SystemCriticalPriority
-				pod.Spec.Priority = &priority
+		}, {
+			description: "Pod not evicted becasuse it has system critical priority",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p12", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					priority := utils.SystemCriticalPriority
+					pod.Spec.Priority = &priority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  false,
-		}, { // Pod is evicted becasuse it has system critical priority, but it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p13", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				priority := utils.SystemCriticalPriority
-				pod.Spec.Priority = &priority
-				pod.Annotations = map[string]string{
-					"descheduler.alpha.kubernetes.io/evict": "true",
-				}
+		}, {
+			description: "Pod is evicted becasuse it has system critical priority, but it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p13", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					priority := utils.SystemCriticalPriority
+					pod.Spec.Priority = &priority
+					pod.Annotations = map[string]string{
+						"descheduler.alpha.kubernetes.io/evict": "true",
+					}
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			result:                  true,
-		}, { // Pod not evicted becasuse it has a priority higher than the configured priority threshold
-			pod: test.BuildTestPod("p14", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Spec.Priority = &highPriority
+		}, {
+			description: "Pod not evicted becasuse it has a priority higher than the configured priority threshold",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p14", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.Priority = &highPriority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			priorityThreshold:       &lowPriority,
 			result:                  false,
-		}, { // Pod is evicted becasuse it has a priority higher than the configured priority threshold, but it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p15", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.Spec.Priority = &highPriority
+		}, {
+			description: "Pod is evicted becasuse it has a priority higher than the configured priority threshold, but it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p15", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.Spec.Priority = &highPriority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			priorityThreshold:       &lowPriority,
 			result:                  true,
-		}, { // Pod is evicted becasuse it has system critical priority, but evictSystemCriticalPods = true
-			pod: test.BuildTestPod("p16", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				priority := utils.SystemCriticalPriority
-				pod.Spec.Priority = &priority
+		}, {
+			description: "Pod is evicted becasuse it has system critical priority, but evictSystemCriticalPods = true",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p16", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					priority := utils.SystemCriticalPriority
+					pod.Spec.Priority = &priority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: true,
 			result:                  true,
-		}, { // Pod is evicted becasuse it has system critical priority, but evictSystemCriticalPods = true and it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p16", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				priority := utils.SystemCriticalPriority
-				pod.Spec.Priority = &priority
+		}, {
+			description: "Pod is evicted becasuse it has system critical priority, but evictSystemCriticalPods = true and it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p16", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					priority := utils.SystemCriticalPriority
+					pod.Spec.Priority = &priority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: true,
 			result:                  true,
-		}, { // Pod is evicted becasuse it has a priority higher than the configured priority threshold, but evictSystemCriticalPods = true
-			pod: test.BuildTestPod("p17", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Spec.Priority = &highPriority
+		}, {
+			description: "Pod is evicted becasuse it has a priority higher than the configured priority threshold, but evictSystemCriticalPods = true",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p17", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.Priority = &highPriority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: true,
 			priorityThreshold:       &lowPriority,
 			result:                  true,
-		}, { // Pod is evicted becasuse it has a priority higher than the configured priority threshold, but evictSystemCriticalPods = true and it has scheduler.alpha.kubernetes.io/evict annotation
-			pod: test.BuildTestPod("p17", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
-				pod.Spec.Priority = &highPriority
+		}, {
+			description: "Pod is evicted becasuse it has a priority higher than the configured priority threshold, but evictSystemCriticalPods = true and it has scheduler.alpha.kubernetes.io/evict annotation",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p17", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
+					pod.Spec.Priority = &highPriority
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: true,
 			priorityThreshold:       &lowPriority,
 			result:                  true,
-		}, { // Pod with no tolerations running on normal node, all other nodes tainted
-			pod:   test.BuildTestPod("p1", 400, 0, n1.Name, nil),
-			nodes: []*v1.Node{test.BuildTestNode("node2", 1000, 2000, 13, nil), test.BuildTestNode("node3", 1000, 2000, 13, nil)},
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-
-				for _, node := range nodes {
+		}, {
+			description: "Pod with no tolerations running on normal node, all other nodes tainted",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 1000, 2000, 13, func(node *v1.Node) {
 					node.Spec.Taints = []v1.Taint{
 						{
 							Key:    nodeTaintKey,
@@ -333,27 +404,8 @@ func TestIsEvictable(t *testing.T) {
 							Effect: v1.TaintEffectNoSchedule,
 						},
 					}
-				}
-			},
-			evictLocalStoragePods:   false,
-			evictSystemCriticalPods: false,
-			nodeFit:                 true,
-			result:                  false,
-		}, { // Pod with correct tolerations running on normal node, all other nodes tainted
-			pod: test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
-				pod.Spec.Tolerations = []v1.Toleration{
-					{
-						Key:    nodeTaintKey,
-						Value:  nodeTaintValue,
-						Effect: v1.TaintEffectNoSchedule,
-					},
-				}
-			}),
-			nodes: []*v1.Node{test.BuildTestNode("node2", 1000, 2000, 13, nil), test.BuildTestNode("node3", 1000, 2000, 13, nil)},
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-
-				for _, node := range nodes {
+				}),
+				test.BuildTestNode("node3", 1000, 2000, 13, func(node *v1.Node) {
 					node.Spec.Taints = []v1.Taint{
 						{
 							Key:    nodeTaintKey,
@@ -361,80 +413,259 @@ func TestIsEvictable(t *testing.T) {
 							Effect: v1.TaintEffectNoSchedule,
 						},
 					}
-				}
-			},
-			evictLocalStoragePods:   false,
-			evictSystemCriticalPods: false,
-			nodeFit:                 true,
-			result:                  true,
-		}, { // Pod with incorrect node selector
-			pod: test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
-				pod.Spec.NodeSelector = map[string]string{
-					nodeLabelKey: "fail",
-				}
-			}),
-			nodes: []*v1.Node{test.BuildTestNode("node2", 1000, 2000, 13, nil), test.BuildTestNode("node3", 1000, 2000, 13, nil)},
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-
-				for _, node := range nodes {
-					node.ObjectMeta.Labels = map[string]string{
-						nodeLabelKey: nodeLabelValue,
-					}
-				}
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			nodeFit:                 true,
 			result:                  false,
-		}, { // Pod with correct node selector
-			pod: test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
-				pod.Spec.NodeSelector = map[string]string{
-					nodeLabelKey: nodeLabelValue,
-				}
-			}),
-			nodes: []*v1.Node{test.BuildTestNode("node2", 1000, 2000, 13, nil), test.BuildTestNode("node3", 1000, 2000, 13, nil)},
-			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
-				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
-
-				for _, node := range nodes {
-					node.ObjectMeta.Labels = map[string]string{
-						nodeLabelKey: nodeLabelValue,
+		}, {
+			description: "Pod with correct tolerations running on normal node, all other nodes tainted",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.Tolerations = []v1.Toleration{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
 					}
-				}
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 1000, 2000, 13, func(node *v1.Node) {
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode("node3", 1000, 2000, 13, func(node *v1.Node) {
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
 			},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
 			nodeFit:                 true,
 			result:                  true,
+		}, {
+			description: "Pod with incorrect node selector",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.NodeSelector = map[string]string{
+						nodeLabelKey: "fail",
+					}
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 1000, 2000, 13, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestNode("node3", 1000, 2000, 13, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			nodeFit:                 true,
+			result:                  false,
+		}, {
+			description: "Pod with correct node selector",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.NodeSelector = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 1000, 2000, 13, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestNode("node3", 1000, 2000, 13, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			nodeFit:                 true,
+			result:                  true,
+		}, {
+			description: "Pod with correct node selector, but only available node doesn't have enough CPU",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 12, 8, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.NodeSelector = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2-TEST", 10, 16, 10, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestNode("node3-TEST", 10, 16, 10, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			nodeFit:                 true,
+			result:                  false,
+		}, {
+			description: "Pod with correct node selector, and one node has enough memory",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 12, 8, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.NodeSelector = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestPod("node2-pod-10GB-mem", 20, 10, "node2", func(pod *v1.Pod) {
+					pod.ObjectMeta.Labels = map[string]string{
+						"test": "true",
+					}
+				}),
+				test.BuildTestPod("node3-pod-10GB-mem", 20, 10, "node3", func(pod *v1.Pod) {
+					pod.ObjectMeta.Labels = map[string]string{
+						"test": "true",
+					}
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 100, 16, 10, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestNode("node3", 100, 20, 10, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			nodeFit:                 true,
+			result:                  true,
+		}, {
+			description: "Pod with correct node selector, but both nodes don't have enough memory",
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 12, 8, n1.Name, func(pod *v1.Pod) {
+					pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+					pod.Spec.NodeSelector = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestPod("node2-pod-10GB-mem", 10, 10, "node2", func(pod *v1.Pod) {
+					pod.ObjectMeta.Labels = map[string]string{
+						"test": "true",
+					}
+				}),
+				test.BuildTestPod("node3-pod-10GB-mem", 10, 10, "node3", func(pod *v1.Pod) {
+					pod.ObjectMeta.Labels = map[string]string{
+						"test": "true",
+					}
+				}),
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 100, 16, 10, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+				test.BuildTestNode("node3", 100, 16, 10, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+				}),
+			},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			nodeFit:                 true,
+			result:                  false,
 		},
 	}
 
 	for _, test := range testCases {
-		test.runBefore(test.pod, test.nodes)
-		nodes := append(test.nodes, n1)
 
-		podEvictor := &PodEvictor{
-			evictLocalStoragePods:   test.evictLocalStoragePods,
-			evictSystemCriticalPods: test.evictSystemCriticalPods,
-			nodes:                   nodes,
-		}
+		t.Run(test.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
 
-		evictable := podEvictor.Evictable()
-		var opts []func(opts *Options)
-		if test.priorityThreshold != nil {
-			opts = append(opts, WithPriorityThreshold(*test.priorityThreshold))
-		}
-		if test.nodeFit {
-			opts = append(opts, WithNodeFit(true))
-		}
-		evictable = podEvictor.Evictable(opts...)
+			nodes := append(test.nodes, n1)
 
-		result := evictable.IsEvictable(test.pod)
-		if result != test.result {
-			t.Errorf("IsEvictable should return for pod %s %t, but it returns %t", test.pod.Name, test.result, result)
-		}
+			var objs []runtime.Object
+			for _, node := range test.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range test.pods {
+				objs = append(objs, pod)
+			}
 
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := &PodEvictor{
+				client:                     fakeClient,
+				nodes:                      nodes,
+				nodeIndexer:                getPodsAssignedToNode,
+				policyGroupVersion:         policyv1.SchemeGroupVersion.String(),
+				dryRun:                     false,
+				maxPodsToEvictPerNode:      nil,
+				maxPodsToEvictPerNamespace: nil,
+				evictLocalStoragePods:      test.evictLocalStoragePods,
+				evictSystemCriticalPods:    test.evictSystemCriticalPods,
+				evictFailedBarePods:        test.evictFailedBarePods,
+			}
+
+			var opts []func(opts *Options)
+			if test.priorityThreshold != nil {
+				opts = append(opts, WithPriorityThreshold(*test.priorityThreshold))
+			}
+			if test.nodeFit {
+				opts = append(opts, WithNodeFit(true))
+			}
+			evictable := podEvictor.Evictable(opts...)
+
+			result := evictable.IsEvictable(test.pods[0])
+			if result != test.result {
+				t.Errorf("IsEvictable should return for pod %s %t, but it returns %t", test.pods[0].Name, test.result, result)
+			}
+		})
 	}
 }
 func TestPodTypes(t *testing.T) {

pkg/descheduler/leaderelection.go

@@ -0,0 +1,99 @@
+/*
+Copyright 2022 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+   http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package descheduler
+
+import (
+	"context"
+	"fmt"
+	"k8s.io/apimachinery/pkg/util/uuid"
+	clientset "k8s.io/client-go/kubernetes"
+	"k8s.io/client-go/tools/leaderelection"
+	"k8s.io/client-go/tools/leaderelection/resourcelock"
+	componentbaseconfig "k8s.io/component-base/config"
+	"k8s.io/klog/v2"
+	"os"
+)
+
+// NewLeaderElection starts the leader election code loop
+func NewLeaderElection(
+	run func() error,
+	client clientset.Interface,
+	LeaderElectionConfig *componentbaseconfig.LeaderElectionConfiguration,
+	ctx context.Context,
+) error {
+	var id string
+
+	if hostname, err := os.Hostname(); err != nil {
+		// on errors, make sure we're unique
+		id = string(uuid.NewUUID())
+	} else {
+		// add a uniquifier so that two processes on the same host don't accidentally both become active
+		id = hostname + "_" + string(uuid.NewUUID())
+	}
+
+	klog.V(3).Infof("Assigned unique lease holder id: %s", id)
+
+	if len(LeaderElectionConfig.ResourceNamespace) == 0 {
+		return fmt.Errorf("namespace may not be empty")
+	}
+
+	if len(LeaderElectionConfig.ResourceName) == 0 {
+		return fmt.Errorf("name may not be empty")
+	}
+
+	lock, err := resourcelock.New(
+		LeaderElectionConfig.ResourceLock,
+		LeaderElectionConfig.ResourceNamespace,
+		LeaderElectionConfig.ResourceName,
+		client.CoreV1(),
+		client.CoordinationV1(),
+		resourcelock.ResourceLockConfig{
+			Identity: id,
+		},
+	)
+	if err != nil {
+		return fmt.Errorf("unable to create leader election lock: %v", err)
+	}
+
+	leaderelection.RunOrDie(ctx, leaderelection.LeaderElectionConfig{
+		Lock:            lock,
+		ReleaseOnCancel: true,
+		LeaseDuration:   LeaderElectionConfig.LeaseDuration.Duration,
+		RenewDeadline:   LeaderElectionConfig.RenewDeadline.Duration,
+		RetryPeriod:     LeaderElectionConfig.RetryPeriod.Duration,
+		Callbacks: leaderelection.LeaderCallbacks{
+			OnStartedLeading: func(ctx context.Context) {
+				klog.V(1).InfoS("Started leading")
+				err := run()
+				if err != nil {
+					klog.Error(err)
+				}
+			},
+			OnStoppedLeading: func() {
+				klog.V(1).InfoS("Leader lost")
+			},
+			OnNewLeader: func(identity string) {
+				// Just got the lock
+				if identity == id {
+					return
+				}
+				klog.V(1).Infof("New leader elected: %v", identity)
+			},
+		},
+	})
+	return nil
+}

pkg/descheduler/node/node.go

@@ -18,13 +18,16 @@ package node
 
 import (
 	"context"
+	"fmt"
 
 	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/labels"
 	coreinformers "k8s.io/client-go/informers/core/v1"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
@@ -96,32 +99,90 @@ func IsReady(node *v1.Node) bool {
 	return true
 }
 
-// PodFitsAnyOtherNode checks if the given pod fits any of the given nodes, besides the node
-// the pod is already running on. The node fit is based on multiple criteria, like, pod node selector
-// matching the node label (including affinity), the taints on the node, and the node being schedulable or not.
-func PodFitsAnyOtherNode(pod *v1.Pod, nodes []*v1.Node) bool {
+// NodeFit returns true if the provided pod can be scheduled onto the provided node.
+// This function is used when the NodeFit pod filtering feature of the Descheduler is enabled.
+// This function currently considers a subset of the Kubernetes Scheduler's predicates when
+// deciding if a pod would fit on a node, but more predicates may be added in the future.
+func NodeFit(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) []error {
+	// Check node selector and required affinity
+	var errors []error
+	if ok, err := utils.PodMatchNodeSelector(pod, node); err != nil {
+		errors = append(errors, err)
+	} else if !ok {
+		errors = append(errors, fmt.Errorf("pod node selector does not match the node label"))
+	}
+	// Check taints (we only care about NoSchedule and NoExecute taints)
+	ok := utils.TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, node.Spec.Taints, func(taint *v1.Taint) bool {
+		return taint.Effect == v1.TaintEffectNoSchedule || taint.Effect == v1.TaintEffectNoExecute
+	})
+	if !ok {
+		errors = append(errors, fmt.Errorf("pod does not tolerate taints on the node"))
+	}
+	// Check if the pod can fit on a node based off it's requests
+	ok, reqErrors := fitsRequest(nodeIndexer, pod, node)
+	if !ok {
+		errors = append(errors, reqErrors...)
+	}
+	// Check if node is schedulable
+	if IsNodeUnschedulable(node) {
+		errors = append(errors, fmt.Errorf("node is not schedulable"))
+	}
 
+	return errors
+}
+
+// PodFitsAnyOtherNode checks if the given pod will fit any of the given nodes, besides the node
+// the pod is already running on. The predicates used to determine if the pod will fit can be found in the NodeFit function.
+func PodFitsAnyOtherNode(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, nodes []*v1.Node) bool {
 	for _, node := range nodes {
 		// Skip node pod is already on
 		if node.Name == pod.Spec.NodeName {
 			continue
 		}
-		// Check node selector and required affinity
-		ok, err := utils.PodMatchNodeSelector(pod, node)
-		if err != nil || !ok {
-			continue
-		}
-		// Check taints (we only care about NoSchedule and NoExecute taints)
-		ok = utils.TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, node.Spec.Taints, func(taint *v1.Taint) bool {
-			return taint.Effect == v1.TaintEffectNoSchedule || taint.Effect == v1.TaintEffectNoExecute
-		})
-		if !ok {
-			continue
-		}
-		// Check if node is schedulable
-		if !IsNodeUnschedulable(node) {
-			klog.V(2).InfoS("Pod can possibly be scheduled on a different node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+
+		errors := NodeFit(nodeIndexer, pod, node)
+		if len(errors) == 0 {
+			klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
 			return true
+		} else {
+			klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+			for _, err := range errors {
+				klog.V(4).InfoS(err.Error())
+			}
+		}
+	}
+	return false
+}
+
+// PodFitsAnyNode checks if the given pod will fit any of the given nodes. The predicates used
+// to determine if the pod will fit can be found in the NodeFit function.
+func PodFitsAnyNode(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, nodes []*v1.Node) bool {
+	for _, node := range nodes {
+		errors := NodeFit(nodeIndexer, pod, node)
+		if len(errors) == 0 {
+			klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+			return true
+		} else {
+			klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+			for _, err := range errors {
+				klog.V(4).InfoS(err.Error())
+			}
+		}
+	}
+	return false
+}
+
+// PodFitsCurrentNode checks if the given pod will fit onto the given node. The predicates used
+// to determine if the pod will fit can be found in the NodeFit function.
+func PodFitsCurrentNode(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) bool {
+	errors := NodeFit(nodeIndexer, pod, node)
+	if len(errors) == 0 {
+		klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+		return true
+	} else {
+		klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+		for _, err := range errors {
+			klog.V(4).InfoS(err.Error())
 		}
 	}
 	return false
@@ -133,39 +194,95 @@ func IsNodeUnschedulable(node *v1.Node) bool {
 	return node.Spec.Unschedulable
 }
 
-// PodFitsAnyNode checks if the given pod fits any of the given nodes, based on
-// multiple criteria, like, pod node selector matching the node label, node
-// being schedulable or not.
-func PodFitsAnyNode(pod *v1.Pod, nodes []*v1.Node) bool {
-	for _, node := range nodes {
+// fitsRequest determines if a pod can fit on a node based on its resource requests. It returns true if
+// the pod will fit.
+func fitsRequest(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) (bool, []error) {
+	var insufficientResources []error
 
-		ok, err := utils.PodMatchNodeSelector(pod, node)
-		if err != nil || !ok {
-			continue
-		}
-		if !IsNodeUnschedulable(node) {
-			klog.V(2).InfoS("Pod can possibly be scheduled on a different node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-			return true
-		}
+	// Get pod requests
+	podRequests, _ := utils.PodRequestsAndLimits(pod)
+	resourceNames := make([]v1.ResourceName, 0, len(podRequests))
+	for name := range podRequests {
+		resourceNames = append(resourceNames, name)
 	}
-	return false
-}
-
-// PodFitsCurrentNode checks if the given pod fits on the given node if the pod
-// node selector matches the node label.
-func PodFitsCurrentNode(pod *v1.Pod, node *v1.Node) bool {
-	ok, err := utils.PodMatchNodeSelector(pod, node)
 
+	availableResources, err := nodeAvailableResources(nodeIndexer, node, resourceNames)
 	if err != nil {
-		klog.ErrorS(err, "Failed to match node selector")
-		return false
+		return false, []error{err}
 	}
 
-	if !ok {
-		klog.V(2).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-		return false
+	podFitsOnNode := true
+	for _, resource := range resourceNames {
+		podResourceRequest := podRequests[resource]
+		availableResource, ok := availableResources[resource]
+		if !ok || podResourceRequest.MilliValue() > availableResource.MilliValue() {
+			insufficientResources = append(insufficientResources, fmt.Errorf("insufficient %v", resource))
+			podFitsOnNode = false
+		}
+	}
+	return podFitsOnNode, insufficientResources
+}
+
+// nodeAvailableResources returns resources mapped to the quanitity available on the node.
+func nodeAvailableResources(nodeIndexer podutil.GetPodsAssignedToNodeFunc, node *v1.Node, resourceNames []v1.ResourceName) (map[v1.ResourceName]*resource.Quantity, error) {
+	podsOnNode, err := podutil.ListPodsOnANode(node.Name, nodeIndexer, nil)
+	if err != nil {
+		return nil, err
+	}
+	nodeUtilization := NodeUtilization(podsOnNode, resourceNames)
+	remainingResources := map[v1.ResourceName]*resource.Quantity{
+		v1.ResourceCPU:    resource.NewMilliQuantity(node.Status.Allocatable.Cpu().MilliValue()-nodeUtilization[v1.ResourceCPU].MilliValue(), resource.DecimalSI),
+		v1.ResourceMemory: resource.NewQuantity(node.Status.Allocatable.Memory().Value()-nodeUtilization[v1.ResourceMemory].Value(), resource.BinarySI),
+		v1.ResourcePods:   resource.NewQuantity(node.Status.Allocatable.Pods().Value()-nodeUtilization[v1.ResourcePods].Value(), resource.DecimalSI),
+	}
+	for _, name := range resourceNames {
+		if !IsBasicResource(name) {
+			if _, exists := node.Status.Allocatable[name]; exists {
+				allocatableResource := node.Status.Allocatable[name]
+				remainingResources[name] = resource.NewQuantity(allocatableResource.Value()-nodeUtilization[name].Value(), resource.DecimalSI)
+			} else {
+				remainingResources[name] = resource.NewQuantity(0, resource.DecimalSI)
+			}
+		}
+	}
+
+	return remainingResources, nil
+}
+
+// NodeUtilization returns the resources requested by the given pods. Only resources supplied in the resourceNames parameter are calculated.
+func NodeUtilization(pods []*v1.Pod, resourceNames []v1.ResourceName) map[v1.ResourceName]*resource.Quantity {
+	totalReqs := map[v1.ResourceName]*resource.Quantity{
+		v1.ResourceCPU:    resource.NewMilliQuantity(0, resource.DecimalSI),
+		v1.ResourceMemory: resource.NewQuantity(0, resource.BinarySI),
+		v1.ResourcePods:   resource.NewQuantity(int64(len(pods)), resource.DecimalSI),
+	}
+	for _, name := range resourceNames {
+		if !IsBasicResource(name) {
+			totalReqs[name] = resource.NewQuantity(0, resource.DecimalSI)
+		}
+	}
+
+	for _, pod := range pods {
+		req, _ := utils.PodRequestsAndLimits(pod)
+		for _, name := range resourceNames {
+			quantity, ok := req[name]
+			if ok && name != v1.ResourcePods {
+				// As Quantity.Add says: Add adds the provided y quantity to the current value. If the current value is zero,
+				// the format of the quantity will be updated to the format of y.
+				totalReqs[name].Add(quantity)
+			}
+		}
+	}
+
+	return totalReqs
+}
+
+// IsBasicResource checks if resource is basic native.
+func IsBasicResource(name v1.ResourceName) bool {
+	switch name {
+	case v1.ResourceCPU, v1.ResourceMemory, v1.ResourcePods:
+		return true
+	default:
+		return false
 	}
-
-	klog.V(2).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-	return true
 }

pkg/descheduler/node/node_test.go

@@ -21,9 +21,12 @@ import (
 	"testing"
 
 	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/test"
 )
 
@@ -147,13 +150,13 @@ func TestPodFitsCurrentNode(t *testing.T) {
 					},
 				},
 			},
-			node: &v1.Node{
-				ObjectMeta: metav1.ObjectMeta{
-					Labels: map[string]string{
-						nodeLabelKey: nodeLabelValue,
-					},
-				},
-			},
+			node: test.BuildTestNode("node1", 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+				node.ObjectMeta.Labels = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+
+				node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+			}),
 			success: true,
 		},
 		{
@@ -181,27 +184,48 @@ func TestPodFitsCurrentNode(t *testing.T) {
 					},
 				},
 			},
-			node: &v1.Node{
-				ObjectMeta: metav1.ObjectMeta{
-					Labels: map[string]string{
-						nodeLabelKey: "no",
-					},
-				},
-			},
+			node: test.BuildTestNode("node1", 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+				node.ObjectMeta.Labels = map[string]string{
+					nodeLabelKey: "no",
+				}
+
+				node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+			}),
 			success: false,
 		},
 	}
 
 	for _, tc := range tests {
-		actual := PodFitsCurrentNode(tc.pod, tc.node)
-		if actual != tc.success {
-			t.Errorf("Test %#v failed", tc.description)
-		}
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			objs = append(objs, tc.node)
+			objs = append(objs, tc.pod)
+
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			actual := PodFitsCurrentNode(getPodsAssignedToNode, tc.pod, tc.node)
+			if actual != tc.success {
+				t.Errorf("Test %#v failed", tc.description)
+			}
+		})
 	}
 }
 
 func TestPodFitsAnyOtherNode(t *testing.T) {
-
 	nodeLabelKey := "kubernetes.io/desiredNode"
 	nodeLabelValue := "yes"
 	nodeTaintKey := "hardware"
@@ -215,238 +239,527 @@ func TestPodFitsAnyOtherNode(t *testing.T) {
 		pod         *v1.Pod
 		nodes       []*v1.Node
 		success     bool
+		podsOnNodes []*v1.Pod
 	}{
 		{
 			description: "Pod fits another node matching node affinity",
-			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			pod: test.BuildTestPod("p1", 0, 0, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+			}),
 			nodes: []*v1.Node{
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[0],
-						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[1],
-						Labels: map[string]string{
-							nodeLabelKey: "no",
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[2],
-					},
-				},
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[1], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: "no",
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
 			},
-			success: true,
+			podsOnNodes: []*v1.Pod{},
+			success:     true,
 		},
 		{
 			description: "Pod expected to fit one of the nodes",
-			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			pod: test.BuildTestPod("p1", 0, 0, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+			}),
 			nodes: []*v1.Node{
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[0],
-						Labels: map[string]string{
-							nodeLabelKey: "no",
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[1],
-						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[2],
-					},
-				},
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: "no",
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[1], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
 			},
-			success: true,
+			podsOnNodes: []*v1.Pod{},
+			success:     true,
 		},
 		{
 			description: "Pod expected to fit none of the nodes",
-			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			pod: test.BuildTestPod("p1", 0, 0, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+			}),
 			nodes: []*v1.Node{
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[0],
-						Labels: map[string]string{
-							nodeLabelKey: "unfit1",
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[1],
-						Labels: map[string]string{
-							nodeLabelKey: "unfit2",
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[2],
-					},
-				},
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: "unfit1",
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[1], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: "unfit2",
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
 			},
-			success: false,
+			podsOnNodes: []*v1.Pod{},
+			success:     false,
 		},
 		{
 			description: "Nodes are unschedulable but labels match, should fail",
-			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			pod: test.BuildTestPod("p1", 0, 0, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+			}),
 			nodes: []*v1.Node{
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[0],
-						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
-						},
-					},
-					Spec: v1.NodeSpec{
-						Unschedulable: true,
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[1],
-						Labels: map[string]string{
-							nodeLabelKey: "no",
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[2],
-					},
-				},
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Unschedulable = true
+				}),
+				test.BuildTestNode(nodeNames[1], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: "no",
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
 			},
-			success: false,
+			podsOnNodes: []*v1.Pod{},
+			success:     false,
 		},
 		{
-			description: "Two nodes matches node selector, one of them is tained, should pass",
-			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			description: "Both nodes are tained, should fail",
+			pod: test.BuildTestPod("p1", 2000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
 			nodes: []*v1.Node{
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[0],
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode(nodeNames[1], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{},
+			success:     false,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, there is a pod on the available node, and requests are low, should pass",
+			pod: test.BuildTestPod("p1", 2000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode(nodeNames[1], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("test-pod", 12*1000, 20*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
 						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
+							"test": "true",
 						},
-					},
-					Spec: v1.NodeSpec{
-						Taints: []v1.Taint{
-							{
-								Key:    nodeTaintKey,
-								Value:  nodeTaintValue,
-								Effect: v1.TaintEffectNoSchedule,
-							},
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[1],
-						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
-						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[2],
-					},
-				},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(40*1000*1000*1000, resource.DecimalSI)
+				}),
 			},
 			success: true,
 		},
 		{
-			description: "Both nodes are tained, should fail",
-			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			description: "Two nodes matches node selector, one of them is tained, but CPU requests are too big, should fail",
+			pod: test.BuildTestPod("p1", 2000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
 			nodes: []*v1.Node{
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[0],
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				// Notice that this node only has 4 cores, the pod already on the node below requests 3 cores, and the pod above requests 2 cores
+				test.BuildTestNode(nodeNames[1], 4000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("3-core-pod", 3000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
 						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
+							"test": "true",
 						},
-					},
-					Spec: v1.NodeSpec{
-						Taints: []v1.Taint{
-							{
-								Key:    nodeTaintKey,
-								Value:  nodeTaintValue,
-								Effect: v1.TaintEffectNoSchedule,
-							},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+				}),
+			},
+			success: false,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, but memory requests are too big, should fail",
+			pod: test.BuildTestPod("p1", 2000, 5*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
 						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[1],
+					}
+				}),
+				// Notice that this node only has 8GB of memory, the pod already on the node below requests 4GB, and the pod above requests 5GB
+				test.BuildTestNode(nodeNames[1], 10*1000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("4GB-mem-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
 						Labels: map[string]string{
-							nodeLabelKey: nodeLabelValue,
+							"test": "true",
 						},
-					},
-					Spec: v1.NodeSpec{
-						Taints: []v1.Taint{
-							{
-								Key:    nodeTaintKey,
-								Value:  nodeTaintValue,
-								Effect: v1.TaintEffectNoExecute,
-							},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+				}),
+			},
+			success: false,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, but ephemeral storage requests are too big, should fail",
+			pod: test.BuildTestPod("p1", 2000, 4*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
 						},
-					},
-				},
-				{
-					ObjectMeta: metav1.ObjectMeta{
-						Name: nodeNames[2],
-					},
-				},
+					}
+				}),
+				// Notice that this node only has 20GB of storage, the pod already on the node below requests 11GB, and the pod above requests 10GB
+				test.BuildTestNode(nodeNames[1], 10*1000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(20*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("11GB-storage-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(11*1000*1000*1000, resource.DecimalSI)
+				}),
+			},
+			success: false,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, but custom resource requests are too big, should fail",
+			pod: test.BuildTestPod("p1", 2000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+				pod.Spec.Containers[0].Resources.Requests["example.com/custom-resource"] = *resource.NewQuantity(10, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+					node.Status.Allocatable["example.com/custom-resource"] = *resource.NewQuantity(15, resource.DecimalSI)
+				}),
+				// Notice that this node only has 15 of the custom resource, the pod already on the node below requests 10, and the pod above requests 10
+				test.BuildTestNode(nodeNames[1], 10*1000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+					node.Status.Allocatable["example.com/custom-resource"] = *resource.NewQuantity(15, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("10-custom-resource-pod", 0, 0, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests["example.com/custom-resource"] = *resource.NewQuantity(10, resource.DecimalSI)
+				}),
+			},
+			success: false,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, CPU requests will fit, and pod Overhead is low enough, should pass",
+			pod: test.BuildTestPod("p1", 1000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				// Notice that this node has 5 CPU cores, the pod below requests 2 cores, and has CPU overhead of 1 cores, and the pod above requests 1 core
+				test.BuildTestNode(nodeNames[1], 5000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("3-core-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+					pod.Spec.Overhead = createResourceList(1000, 1000*1000*1000, 1000*1000*1000)
+				}),
+			},
+			success: true,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, CPU requests will fit, but pod Overhead is too high, should fail",
+			pod: test.BuildTestPod("p1", 2000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				// Notice that this node only has 5 CPU cores, the pod below requests 2 cores, but has CPU overhead of 2 cores, and the pod above requests 2 cores
+				test.BuildTestNode(nodeNames[1], 5000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("3-core-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+					pod.Spec.Overhead = createResourceList(2000, 1000*1000*1000, 1000*1000*1000)
+				}),
 			},
 			success: false,
 		},
 	}
 
 	for _, tc := range tests {
-		actual := PodFitsAnyOtherNode(tc.pod, tc.nodes)
-		if actual != tc.success {
-			t.Errorf("Test %#v failed", tc.description)
-		}
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range tc.podsOnNodes {
+				objs = append(objs, pod)
+			}
+			objs = append(objs, tc.pod)
+
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			actual := PodFitsAnyOtherNode(getPodsAssignedToNode, tc.pod, tc.nodes)
+			if actual != tc.success {
+				t.Errorf("Test %#v failed", tc.description)
+			}
+		})
 	}
 }
 
-func createPodManifest(nodeName string, nodeSelectorKey string, nodeSelectorValue string) *v1.Pod {
-	return (&v1.Pod{
-		Spec: v1.PodSpec{
-			NodeName: nodeName,
-			Affinity: &v1.Affinity{
-				NodeAffinity: &v1.NodeAffinity{
-					RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
-						NodeSelectorTerms: []v1.NodeSelectorTerm{
-							{
-								MatchExpressions: []v1.NodeSelectorRequirement{
-									{
-										Key:      nodeSelectorKey,
-										Operator: "In",
-										Values: []string{
-											nodeSelectorValue,
-										},
-									},
-								},
-							},
-						},
-					},
-				},
-			},
-		},
-	})
+// createResourceList builds a small resource list of core resources
+func createResourceList(cpu int64, memory int64, ephemeralStorage int64) v1.ResourceList {
+	resourceList := make(map[v1.ResourceName]resource.Quantity)
+	resourceList[v1.ResourceCPU] = *resource.NewMilliQuantity(cpu, resource.DecimalSI)
+	resourceList[v1.ResourceMemory] = *resource.NewQuantity(memory, resource.DecimalSI)
+	resourceList[v1.ResourceEphemeralStorage] = *resource.NewQuantity(ephemeralStorage, resource.DecimalSI)
+	return resourceList
 }

pkg/descheduler/pod/pods.go

@@ -17,139 +17,174 @@ limitations under the License.
 package pod
 
 import (
-	"context"
 	"sort"
 
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	"k8s.io/apimachinery/pkg/fields"
-	clientset "k8s.io/client-go/kubernetes"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/sets"
+	coreinformers "k8s.io/client-go/informers/core/v1"
+	"k8s.io/client-go/tools/cache"
+
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
+const (
+	nodeNameKeyIndex = "spec.nodeName"
+)
+
+// FilterFunc is a filter for a pod.
+type FilterFunc func(*v1.Pod) bool
+
+// GetPodsAssignedToNodeFunc is a function which accept a node name and a pod filter function
+// as input and returns the pods that assigned to the node.
+type GetPodsAssignedToNodeFunc func(string, FilterFunc) ([]*v1.Pod, error)
+
+// WrapFilterFuncs wraps a set of FilterFunc in one.
+func WrapFilterFuncs(filters ...FilterFunc) FilterFunc {
+	return func(pod *v1.Pod) bool {
+		for _, filter := range filters {
+			if filter != nil && !filter(pod) {
+				return false
+			}
+		}
+		return true
+	}
+}
+
 type Options struct {
-	filter             func(pod *v1.Pod) bool
-	includedNamespaces []string
-	excludedNamespaces []string
+	filter             FilterFunc
+	includedNamespaces sets.String
+	excludedNamespaces sets.String
 	labelSelector      *metav1.LabelSelector
 }
 
+// NewOptions returns an empty Options.
+func NewOptions() *Options {
+	return &Options{}
+}
+
 // WithFilter sets a pod filter.
 // The filter function should return true if the pod should be returned from ListPodsOnANode
-func WithFilter(filter func(pod *v1.Pod) bool) func(opts *Options) {
-	return func(opts *Options) {
-		opts.filter = filter
-	}
+func (o *Options) WithFilter(filter FilterFunc) *Options {
+	o.filter = filter
+	return o
 }
 
 // WithNamespaces sets included namespaces
-func WithNamespaces(namespaces []string) func(opts *Options) {
-	return func(opts *Options) {
-		opts.includedNamespaces = namespaces
-	}
+func (o *Options) WithNamespaces(namespaces sets.String) *Options {
+	o.includedNamespaces = namespaces
+	return o
 }
 
 // WithoutNamespaces sets excluded namespaces
-func WithoutNamespaces(namespaces []string) func(opts *Options) {
-	return func(opts *Options) {
-		opts.excludedNamespaces = namespaces
-	}
+func (o *Options) WithoutNamespaces(namespaces sets.String) *Options {
+	o.excludedNamespaces = namespaces
+	return o
 }
 
 // WithLabelSelector sets a pod label selector
-func WithLabelSelector(labelSelector *metav1.LabelSelector) func(opts *Options) {
-	return func(opts *Options) {
-		opts.labelSelector = labelSelector
-	}
+func (o *Options) WithLabelSelector(labelSelector *metav1.LabelSelector) *Options {
+	o.labelSelector = labelSelector
+	return o
 }
 
-// ListPodsOnANode lists all of the pods on a node
-// It also accepts an optional "filter" function which can be used to further limit the pods that are returned.
-// (Usually this is podEvictor.Evictable().IsEvictable, in order to only list the evictable pods on a node, but can
-// be used by strategies to extend it if there are further restrictions, such as with NodeAffinity).
-func ListPodsOnANode(
-	ctx context.Context,
-	client clientset.Interface,
-	node *v1.Node,
-	opts ...func(opts *Options),
-) ([]*v1.Pod, error) {
-	options := &Options{}
-	for _, opt := range opts {
-		opt(options)
-	}
-
-	pods := make([]*v1.Pod, 0)
-
-	fieldSelectorString := "spec.nodeName=" + node.Name + ",status.phase!=" + string(v1.PodSucceeded) + ",status.phase!=" + string(v1.PodFailed)
-
-	labelSelectorString := ""
-	if options.labelSelector != nil {
-		selector, err := metav1.LabelSelectorAsSelector(options.labelSelector)
+// BuildFilterFunc builds a final FilterFunc based on Options.
+func (o *Options) BuildFilterFunc() (FilterFunc, error) {
+	var s labels.Selector
+	var err error
+	if o.labelSelector != nil {
+		s, err = metav1.LabelSelectorAsSelector(o.labelSelector)
 		if err != nil {
-			return []*v1.Pod{}, err
+			return nil, err
 		}
-		labelSelectorString = selector.String()
 	}
-
-	if len(options.includedNamespaces) > 0 {
-		fieldSelector, err := fields.ParseSelector(fieldSelectorString)
-		if err != nil {
-			return []*v1.Pod{}, err
+	return func(pod *v1.Pod) bool {
+		if o.filter != nil && !o.filter(pod) {
+			return false
 		}
+		if len(o.includedNamespaces) > 0 && !o.includedNamespaces.Has(pod.Namespace) {
+			return false
+		}
+		if len(o.excludedNamespaces) > 0 && o.excludedNamespaces.Has(pod.Namespace) {
+			return false
+		}
+		if s != nil && !s.Matches(labels.Set(pod.GetLabels())) {
+			return false
+		}
+		return true
+	}, nil
+}
 
-		for _, namespace := range options.includedNamespaces {
-			podList, err := client.CoreV1().Pods(namespace).List(ctx,
-				metav1.ListOptions{
-					FieldSelector: fieldSelector.String(),
-					LabelSelector: labelSelectorString,
-				})
-			if err != nil {
-				return []*v1.Pod{}, err
+// BuildGetPodsAssignedToNodeFunc establishes an indexer to map the pods and their assigned nodes.
+// It returns a function to help us get all the pods that assigned to a node based on the indexer.
+func BuildGetPodsAssignedToNodeFunc(podInformer coreinformers.PodInformer) (GetPodsAssignedToNodeFunc, error) {
+	// Establish an indexer to map the pods and their assigned nodes.
+	err := podInformer.Informer().AddIndexers(cache.Indexers{
+		nodeNameKeyIndex: func(obj interface{}) ([]string, error) {
+			pod, ok := obj.(*v1.Pod)
+			if !ok {
+				return []string{}, nil
 			}
-			for i := range podList.Items {
-				if options.filter != nil && !options.filter(&podList.Items[i]) {
-					continue
-				}
-				pods = append(pods, &podList.Items[i])
+			if len(pod.Spec.NodeName) == 0 {
+				return []string{}, nil
+			}
+			return []string{pod.Spec.NodeName}, nil
+		},
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	// The indexer helps us get all the pods that assigned to a node.
+	podIndexer := podInformer.Informer().GetIndexer()
+	getPodsAssignedToNode := func(nodeName string, filter FilterFunc) ([]*v1.Pod, error) {
+		objs, err := podIndexer.ByIndex(nodeNameKeyIndex, nodeName)
+		if err != nil {
+			return nil, err
+		}
+		pods := make([]*v1.Pod, 0, len(objs))
+		for _, obj := range objs {
+			pod, ok := obj.(*v1.Pod)
+			if !ok {
+				continue
+			}
+			if filter(pod) {
+				pods = append(pods, pod)
 			}
 		}
 		return pods, nil
 	}
+	return getPodsAssignedToNode, nil
+}
 
-	if len(options.excludedNamespaces) > 0 {
-		for _, namespace := range options.excludedNamespaces {
-			fieldSelectorString += ",metadata.namespace!=" + namespace
-		}
+// ListPodsOnANode lists all pods on a node.
+// It also accepts a "filter" function which can be used to further limit the pods that are returned.
+// (Usually this is podEvictor.Evictable().IsEvictable, in order to only list the evictable pods on a node, but can
+// be used by strategies to extend it if there are further restrictions, such as with NodeAffinity).
+func ListPodsOnANode(
+	nodeName string,
+	getPodsAssignedToNode GetPodsAssignedToNodeFunc,
+	filter FilterFunc,
+) ([]*v1.Pod, error) {
+	// Succeeded and failed pods are not considered because they don't occupy any resource.
+	f := func(pod *v1.Pod) bool {
+		return pod.Status.Phase != v1.PodSucceeded && pod.Status.Phase != v1.PodFailed
 	}
+	return ListAllPodsOnANode(nodeName, getPodsAssignedToNode, WrapFilterFuncs(f, filter))
+}
 
-	fieldSelector, err := fields.ParseSelector(fieldSelectorString)
+// ListAllPodsOnANode lists all the pods on a node no matter what the phase of the pod is.
+func ListAllPodsOnANode(
+	nodeName string,
+	getPodsAssignedToNode GetPodsAssignedToNodeFunc,
+	filter FilterFunc,
+) ([]*v1.Pod, error) {
+	pods, err := getPodsAssignedToNode(nodeName, filter)
 	if err != nil {
 		return []*v1.Pod{}, err
 	}
 
-	// INFO(jchaloup): field selectors do not work properly with listers
-	// Once the descheduler switches to pod listers (through informers),
-	// We need to flip to client-side filtering.
-	podList, err := client.CoreV1().Pods(v1.NamespaceAll).List(ctx,
-		metav1.ListOptions{
-			FieldSelector: fieldSelector.String(),
-			LabelSelector: labelSelectorString,
-		})
-	if err != nil {
-		return []*v1.Pod{}, err
-	}
-
-	for i := range podList.Items {
-		// fake client does not support field selectors
-		// so let's filter based on the node name as well (quite cheap)
-		if podList.Items[i].Spec.NodeName != node.Name {
-			continue
-		}
-		if options.filter != nil && !options.filter(&podList.Items[i]) {
-			continue
-		}
-		pods = append(pods, &podList.Items[i])
-	}
 	return pods, nil
 }
 

pkg/descheduler/pod/pods_test.go

@@ -18,16 +18,15 @@ package pod
 
 import (
 	"context"
-	"fmt"
 	"reflect"
-	"strings"
 	"testing"
 
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/test"
 )
 
@@ -39,19 +38,17 @@ var (
 func TestListPodsOnANode(t *testing.T) {
 	testCases := []struct {
 		name             string
-		pods             map[string][]v1.Pod
+		pods             []*v1.Pod
 		node             *v1.Node
 		labelSelector    *metav1.LabelSelector
 		expectedPodCount int
 	}{
 		{
 			name: "test listing pods on a node",
-			pods: map[string][]v1.Pod{
-				"n1": {
-					*test.BuildTestPod("pod1", 100, 0, "n1", nil),
-					*test.BuildTestPod("pod2", 100, 0, "n1", nil),
-				},
-				"n2": {*test.BuildTestPod("pod3", 100, 0, "n2", nil)},
+			pods: []*v1.Pod{
+				test.BuildTestPod("pod1", 100, 0, "n1", nil),
+				test.BuildTestPod("pod2", 100, 0, "n1", nil),
+				test.BuildTestPod("pod3", 100, 0, "n2", nil),
 			},
 			node:             test.BuildTestNode("n1", 2000, 3000, 10, nil),
 			labelSelector:    nil,
@@ -59,17 +56,15 @@ func TestListPodsOnANode(t *testing.T) {
 		},
 		{
 			name: "test listing pods with label selector",
-			pods: map[string][]v1.Pod{
-				"n1": {
-					*test.BuildTestPod("pod1", 100, 0, "n1", nil),
-					*test.BuildTestPod("pod2", 100, 0, "n1", func(pod *v1.Pod) {
-						pod.Labels = map[string]string{"foo": "bar"}
-					}),
-					*test.BuildTestPod("pod3", 100, 0, "n1", func(pod *v1.Pod) {
-						pod.Labels = map[string]string{"foo": "bar1"}
-					}),
-				},
-				"n2": {*test.BuildTestPod("pod4", 100, 0, "n2", nil)},
+			pods: []*v1.Pod{
+				test.BuildTestPod("pod1", 100, 0, "n1", nil),
+				test.BuildTestPod("pod2", 100, 0, "n1", func(pod *v1.Pod) {
+					pod.Labels = map[string]string{"foo": "bar"}
+				}),
+				test.BuildTestPod("pod3", 100, 0, "n1", func(pod *v1.Pod) {
+					pod.Labels = map[string]string{"foo": "bar1"}
+				}),
+				test.BuildTestPod("pod4", 100, 0, "n2", nil),
 			},
 			node: test.BuildTestNode("n1", 2000, 3000, 10, nil),
 			labelSelector: &metav1.LabelSelector{
@@ -85,21 +80,38 @@ func TestListPodsOnANode(t *testing.T) {
 		},
 	}
 	for _, testCase := range testCases {
-		fakeClient := &fake.Clientset{}
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			list := action.(core.ListAction)
-			fieldString := list.GetListRestrictions().Fields.String()
-			if strings.Contains(fieldString, "n1") {
-				return true, &v1.PodList{Items: testCase.pods["n1"]}, nil
-			} else if strings.Contains(fieldString, "n2") {
-				return true, &v1.PodList{Items: testCase.pods["n2"]}, nil
+		t.Run(testCase.name, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			objs = append(objs, testCase.node)
+			for _, pod := range testCase.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			filter, err := NewOptions().WithLabelSelector(testCase.labelSelector).BuildFilterFunc()
+			if err != nil {
+				t.Errorf("Build filter function error: %v", err)
+			}
+
+			pods, _ := ListPodsOnANode(testCase.node.Name, getPodsAssignedToNode, filter)
+			if len(pods) != testCase.expectedPodCount {
+				t.Errorf("Expected %v pods on node %v, got %+v", testCase.expectedPodCount, testCase.node.Name, len(pods))
 			}
-			return true, nil, fmt.Errorf("Failed to list: %v", list)
 		})
-		pods, _ := ListPodsOnANode(context.TODO(), fakeClient, testCase.node, WithLabelSelector(testCase.labelSelector))
-		if len(pods) != testCase.expectedPodCount {
-			t.Errorf("expected %v pods on node %v, got %+v", testCase.expectedPodCount, testCase.node.Name, len(pods))
-		}
 	}
 }
 

pkg/descheduler/strategies/duplicates.go

@@ -66,6 +66,7 @@ func RemoveDuplicatePods(
 	strategy api.DeschedulerStrategy,
 	nodes []*v1.Node,
 	podEvictor *evictions.PodEvictor,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 ) {
 	if err := validateRemoveDuplicatePodsParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid RemoveDuplicatePods parameters")
@@ -77,10 +78,10 @@ func RemoveDuplicatePods(
 		return
 	}
 
-	var includedNamespaces, excludedNamespaces []string
+	var includedNamespaces, excludedNamespaces sets.String
 	if strategy.Params != nil && strategy.Params.Namespaces != nil {
-		includedNamespaces = strategy.Params.Namespaces.Include
-		excludedNamespaces = strategy.Params.Namespaces.Exclude
+		includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+		excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
 	}
 
 	nodeFit := false
@@ -95,15 +96,19 @@ func RemoveDuplicatePods(
 	nodeCount := 0
 	nodeMap := make(map[string]*v1.Node)
 
+	podFilter, err := podutil.NewOptions().
+		WithFilter(evictable.IsEvictable).
+		WithNamespaces(includedNamespaces).
+		WithoutNamespaces(excludedNamespaces).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+
 	for _, node := range nodes {
 		klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
-		pods, err := podutil.ListPodsOnANode(ctx,
-			client,
-			node,
-			podutil.WithFilter(evictable.IsEvictable),
-			podutil.WithNamespaces(includedNamespaces),
-			podutil.WithoutNamespaces(excludedNamespaces),
-		)
+		pods, err := podutil.ListPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
 		if err != nil {
 			klog.ErrorS(err, "Error listing evictable pods on node", "node", klog.KObj(node))
 			continue

pkg/descheduler/strategies/duplicates_test.go

@@ -25,10 +25,12 @@ import (
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
@@ -43,7 +45,6 @@ func buildTestPodWithImage(podName, node, image string) *v1.Pod {
 }
 
 func TestFindDuplicatePods(t *testing.T) {
-	ctx := context.Background()
 	// first setup pods
 	node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
 	node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
@@ -66,6 +67,7 @@ func TestFindDuplicatePods(t *testing.T) {
 			Unschedulable: true,
 		}
 	})
+	node6 := test.BuildTestNode("n6", 200, 200, 10, nil)
 
 	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
 	p1.Namespace = "dev"
@@ -101,6 +103,14 @@ func TestFindDuplicatePods(t *testing.T) {
 	p18 := test.BuildTestPod("TARGET", 100, 0, node1.Name, nil)
 	p18.Namespace = "node-fit"
 
+	// This pod sits on node6 and is used to take up CPU requests on the node
+	p19 := test.BuildTestPod("CPU-eater", 150, 150, node6.Name, nil)
+	p19.Namespace = "test"
+
+	// Dummy pod for node6 used to do the opposite of p19
+	p20 := test.BuildTestPod("CPU-saver", 100, 150, node6.Name, nil)
+	p20.Namespace = "test"
+
 	// ### Evictable Pods ###
 
 	// Three Pods in the "default" Namespace, bound to same ReplicaSet. 2 should be evicted.
@@ -173,128 +183,152 @@ func TestFindDuplicatePods(t *testing.T) {
 
 	testCases := []struct {
 		description             string
-		maxPodsToEvictPerNode   int
-		pods                    []v1.Pod
+		pods                    []*v1.Pod
 		nodes                   []*v1.Node
-		expectedEvictedPodCount int
+		expectedEvictedPodCount uint
 		strategy                api.DeschedulerStrategy
 	}{
 		{
 			description:             "Three pods in the `dev` Namespace, bound to same ReplicaSet. 1 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 1,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Three pods in the `dev` Namespace, bound to same ReplicaSet, but ReplicaSet kind is excluded. 0 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{Params: &api.StrategyParameters{RemoveDuplicates: &api.RemoveDuplicates{ExcludeOwnerKinds: []string{"ReplicaSet"}}}},
 		},
 		{
 			description:             "Three Pods in the `test` Namespace, bound to same ReplicaSet. 1 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p8, *p9, *p10},
+			pods:                    []*v1.Pod{p8, p9, p10},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 1,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Three Pods in the `dev` Namespace, three Pods in the `test` Namespace. Bound to ReplicaSet with same name. 4 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2, *p3, *p8, *p9, *p10},
+			pods:                    []*v1.Pod{p1, p2, p3, p8, p9, p10},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 2,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Pods are: part of DaemonSet, with local storage, mirror pod annotation, critical pod annotation - none should be evicted.",
-			maxPodsToEvictPerNode:   2,
-			pods:                    []v1.Pod{*p4, *p5, *p6, *p7},
+			pods:                    []*v1.Pod{p4, p5, p6, p7},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Test all Pods: 4 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2, *p3, *p4, *p5, *p6, *p7, *p8, *p9, *p10},
+			pods:                    []*v1.Pod{p1, p2, p3, p4, p5, p6, p7, p8, p9, p10},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 2,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Pods with the same owner but different images should not be evicted",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p11, *p12},
+			pods:                    []*v1.Pod{p11, p12},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Pods with multiple containers should not match themselves",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p13},
+			pods:                    []*v1.Pod{p13},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Pods with matching ownerrefs and at not all matching image should not trigger an eviction",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p11, *p13},
+			pods:                    []*v1.Pod{p11, p13},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{},
 		},
 		{
 			description:             "Three pods in the `dev` Namespace, bound to same ReplicaSet. Only node available has a taint, and nodeFit set to true. 0 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1, node3},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{Params: &api.StrategyParameters{NodeFit: true}},
 		},
 		{
 			description:             "Three pods in the `node-fit` Namespace, bound to same ReplicaSet, all with a nodeSelector. Only node available has an incorrect node label, and nodeFit set to true. 0 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p15, *p16, *p17},
+			pods:                    []*v1.Pod{p15, p16, p17},
 			nodes:                   []*v1.Node{node1, node4},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{Params: &api.StrategyParameters{NodeFit: true}},
 		},
 		{
 			description:             "Three pods in the `node-fit` Namespace, bound to same ReplicaSet. Only node available is not schedulable, and nodeFit set to true. 0 should be evicted.",
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1, node5},
 			expectedEvictedPodCount: 0,
 			strategy:                api.DeschedulerStrategy{Params: &api.StrategyParameters{NodeFit: true}},
 		},
+		{
+			description:             "Three pods in the `node-fit` Namespace, bound to same ReplicaSet. Only node available does not have enough CPU, and nodeFit set to true. 0 should be evicted.",
+			pods:                    []*v1.Pod{p1, p2, p3, p19},
+			nodes:                   []*v1.Node{node1, node6},
+			expectedEvictedPodCount: 0,
+			strategy:                api.DeschedulerStrategy{Params: &api.StrategyParameters{NodeFit: true}},
+		},
+		{
+			description:             "Three pods in the `node-fit` Namespace, bound to same ReplicaSet. Only node available has enough CPU, and nodeFit set to true. 1 should be evicted.",
+			pods:                    []*v1.Pod{p1, p2, p3, p20},
+			nodes:                   []*v1.Node{node1, node6},
+			expectedEvictedPodCount: 1,
+			strategy:                api.DeschedulerStrategy{Params: &api.StrategyParameters{NodeFit: true}},
+		},
 	}
 
 	for _, testCase := range testCases {
 		t.Run(testCase.description, func(t *testing.T) {
-			fakeClient := &fake.Clientset{}
-			fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-				return true, &v1.PodList{Items: testCase.pods}, nil
-			})
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range testCase.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range testCase.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
 			podEvictor := evictions.NewPodEvictor(
 				fakeClient,
 				"v1",
 				false,
-				testCase.maxPodsToEvictPerNode,
+				nil,
+				nil,
 				testCase.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
 				false,
 				false,
 				false,
 			)
 
-			RemoveDuplicatePods(ctx, fakeClient, testCase.strategy, testCase.nodes, podEvictor)
+			RemoveDuplicatePods(ctx, fakeClient, testCase.strategy, testCase.nodes, podEvictor, getPodsAssignedToNode)
 			podsEvicted := podEvictor.TotalEvicted()
 			if podsEvicted != testCase.expectedEvictedPodCount {
 				t.Errorf("Test error for description: %s. Expected evicted pods count %v, got %v", testCase.description, testCase.expectedEvictedPodCount, podsEvicted)
@@ -305,8 +339,6 @@ func TestFindDuplicatePods(t *testing.T) {
 }
 
 func TestRemoveDuplicatesUniformly(t *testing.T) {
-	ctx := context.Background()
-
 	setRSOwnerRef2 := func(pod *v1.Pod) {
 		pod.ObjectMeta.OwnerReferences = []metav1.OwnerReference{
 			{Kind: "ReplicaSet", APIVersion: "v1", Name: "replicaset-2"},
@@ -346,7 +378,7 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		node.Spec.Taints = []v1.Taint{
 			{
 				Effect: v1.TaintEffectNoSchedule,
-				Key:    "node-role.kubernetes.io/master",
+				Key:    "node-role.kubernetes.io/control-plane",
 			},
 		}
 	}
@@ -355,7 +387,7 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		if node.ObjectMeta.Labels == nil {
 			node.ObjectMeta.Labels = map[string]string{}
 		}
-		node.ObjectMeta.Labels["node-role.kubernetes.io/master"] = ""
+		node.ObjectMeta.Labels["node-role.kubernetes.io/control-plane"] = ""
 	}
 
 	setWorkerLabel := func(node *v1.Node) {
@@ -375,7 +407,7 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 						{
 							MatchExpressions: []v1.NodeSelectorRequirement{
 								{
-									Key:      "node-role.kubernetes.io/master",
+									Key:      "node-role.kubernetes.io/control-plane",
 									Operator: v1.NodeSelectorOpDoesNotExist,
 								},
 								{
@@ -399,7 +431,7 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 						{
 							MatchExpressions: []v1.NodeSelectorRequirement{
 								{
-									Key:      "node-role.kubernetes.io/master",
+									Key:      "node-role.kubernetes.io/control-plane",
 									Operator: v1.NodeSelectorOpDoesNotExist,
 								},
 								{
@@ -432,25 +464,24 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 
 	testCases := []struct {
 		description             string
-		maxPodsToEvictPerNode   int
-		pods                    []v1.Pod
+		pods                    []*v1.Pod
 		nodes                   []*v1.Node
-		expectedEvictedPodCount int
+		expectedEvictedPodCount uint
 		strategy                api.DeschedulerStrategy
 	}{
 		{
 			description: "Evict pods uniformly",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1) -> (3,3,3) -> 2 evictions
-				*test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p2", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p3", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p4", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p5", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p6", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p7", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p8", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p9", 100, 0, "n3", test.SetRSOwnerRef),
+				test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p7", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p8", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p9", 100, 0, "n3", test.SetRSOwnerRef),
 			},
 			expectedEvictedPodCount: 2,
 			nodes: []*v1.Node{
@@ -462,17 +493,17 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly with one node left out",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1) -> (4,4,1) -> 1 eviction
-				*test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p2", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p3", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p4", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p5", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p6", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p7", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p8", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p9", 100, 0, "n3", test.SetRSOwnerRef),
+				test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p7", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p8", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p9", 100, 0, "n3", test.SetRSOwnerRef),
 			},
 			expectedEvictedPodCount: 1,
 			nodes: []*v1.Node{
@@ -483,17 +514,17 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly with two replica sets",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1) -> (3,3,3) -> 2 evictions
-				*test.BuildTestPod("p11", 100, 0, "n1", setTwoRSOwnerRef),
-				*test.BuildTestPod("p12", 100, 0, "n1", setTwoRSOwnerRef),
-				*test.BuildTestPod("p13", 100, 0, "n1", setTwoRSOwnerRef),
-				*test.BuildTestPod("p14", 100, 0, "n1", setTwoRSOwnerRef),
-				*test.BuildTestPod("p15", 100, 0, "n1", setTwoRSOwnerRef),
-				*test.BuildTestPod("p16", 100, 0, "n2", setTwoRSOwnerRef),
-				*test.BuildTestPod("p17", 100, 0, "n2", setTwoRSOwnerRef),
-				*test.BuildTestPod("p18", 100, 0, "n2", setTwoRSOwnerRef),
-				*test.BuildTestPod("p19", 100, 0, "n3", setTwoRSOwnerRef),
+				test.BuildTestPod("p11", 100, 0, "n1", setTwoRSOwnerRef),
+				test.BuildTestPod("p12", 100, 0, "n1", setTwoRSOwnerRef),
+				test.BuildTestPod("p13", 100, 0, "n1", setTwoRSOwnerRef),
+				test.BuildTestPod("p14", 100, 0, "n1", setTwoRSOwnerRef),
+				test.BuildTestPod("p15", 100, 0, "n1", setTwoRSOwnerRef),
+				test.BuildTestPod("p16", 100, 0, "n2", setTwoRSOwnerRef),
+				test.BuildTestPod("p17", 100, 0, "n2", setTwoRSOwnerRef),
+				test.BuildTestPod("p18", 100, 0, "n2", setTwoRSOwnerRef),
+				test.BuildTestPod("p19", 100, 0, "n3", setTwoRSOwnerRef),
 			},
 			expectedEvictedPodCount: 4,
 			nodes: []*v1.Node{
@@ -505,27 +536,27 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly with two owner references",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1) -> (3,3,3) -> 2 evictions
-				*test.BuildTestPod("p11", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p12", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p13", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p14", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p15", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p16", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p17", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p18", 100, 0, "n2", test.SetRSOwnerRef),
-				*test.BuildTestPod("p19", 100, 0, "n3", test.SetRSOwnerRef),
+				test.BuildTestPod("p11", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p12", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p13", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p14", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p15", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p16", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p17", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p18", 100, 0, "n2", test.SetRSOwnerRef),
+				test.BuildTestPod("p19", 100, 0, "n3", test.SetRSOwnerRef),
 				// (1,3,5) -> (3,3,3) -> 2 evictions
-				*test.BuildTestPod("p21", 100, 0, "n1", setRSOwnerRef2),
-				*test.BuildTestPod("p22", 100, 0, "n2", setRSOwnerRef2),
-				*test.BuildTestPod("p23", 100, 0, "n2", setRSOwnerRef2),
-				*test.BuildTestPod("p24", 100, 0, "n2", setRSOwnerRef2),
-				*test.BuildTestPod("p25", 100, 0, "n3", setRSOwnerRef2),
-				*test.BuildTestPod("p26", 100, 0, "n3", setRSOwnerRef2),
-				*test.BuildTestPod("p27", 100, 0, "n3", setRSOwnerRef2),
-				*test.BuildTestPod("p28", 100, 0, "n3", setRSOwnerRef2),
-				*test.BuildTestPod("p29", 100, 0, "n3", setRSOwnerRef2),
+				test.BuildTestPod("p21", 100, 0, "n1", setRSOwnerRef2),
+				test.BuildTestPod("p22", 100, 0, "n2", setRSOwnerRef2),
+				test.BuildTestPod("p23", 100, 0, "n2", setRSOwnerRef2),
+				test.BuildTestPod("p24", 100, 0, "n2", setRSOwnerRef2),
+				test.BuildTestPod("p25", 100, 0, "n3", setRSOwnerRef2),
+				test.BuildTestPod("p26", 100, 0, "n3", setRSOwnerRef2),
+				test.BuildTestPod("p27", 100, 0, "n3", setRSOwnerRef2),
+				test.BuildTestPod("p28", 100, 0, "n3", setRSOwnerRef2),
+				test.BuildTestPod("p29", 100, 0, "n3", setRSOwnerRef2),
 			},
 			expectedEvictedPodCount: 4,
 			nodes: []*v1.Node{
@@ -537,10 +568,10 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods with number of pods less than nodes",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (2,0,0) -> (1,1,0) -> 1 eviction
-				*test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
-				*test.BuildTestPod("p2", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 100, 0, "n1", test.SetRSOwnerRef),
 			},
 			expectedEvictedPodCount: 1,
 			nodes: []*v1.Node{
@@ -552,14 +583,14 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods with number of pods less than nodes, but ignore different pods with the same ownerref",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (1, 0, 0) for "bar","baz" images -> no eviction, even with a matching ownerKey
 				// (2, 0, 0) for "foo" image -> (1,1,0) - 1 eviction
 				// In this case the only "real" duplicates are p1 and p4, so one of those should be evicted
-				*buildTestPodWithImage("p1", "n1", "foo"),
-				*buildTestPodWithImage("p2", "n1", "bar"),
-				*buildTestPodWithImage("p3", "n1", "baz"),
-				*buildTestPodWithImage("p4", "n1", "foo"),
+				buildTestPodWithImage("p1", "n1", "foo"),
+				buildTestPodWithImage("p2", "n1", "bar"),
+				buildTestPodWithImage("p3", "n1", "baz"),
+				buildTestPodWithImage("p4", "n1", "foo"),
 			},
 			expectedEvictedPodCount: 1,
 			nodes: []*v1.Node{
@@ -571,9 +602,9 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods with a single pod with three nodes",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (2,0,0) -> (1,1,0) -> 1 eviction
-				*test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
+				test.BuildTestPod("p1", 100, 0, "n1", test.SetRSOwnerRef),
 			},
 			expectedEvictedPodCount: 0,
 			nodes: []*v1.Node{
@@ -585,17 +616,17 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly respecting taints",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1,0,0,0) -> (3,3,3,0,0,0) -> 2 evictions
-				*test.BuildTestPod("p1", 100, 0, "worker1", setTolerationsK1),
-				*test.BuildTestPod("p2", 100, 0, "worker1", setTolerationsK2),
-				*test.BuildTestPod("p3", 100, 0, "worker1", setTolerationsK1),
-				*test.BuildTestPod("p4", 100, 0, "worker1", setTolerationsK2),
-				*test.BuildTestPod("p5", 100, 0, "worker1", setTolerationsK1),
-				*test.BuildTestPod("p6", 100, 0, "worker2", setTolerationsK2),
-				*test.BuildTestPod("p7", 100, 0, "worker2", setTolerationsK1),
-				*test.BuildTestPod("p8", 100, 0, "worker2", setTolerationsK2),
-				*test.BuildTestPod("p9", 100, 0, "worker3", setTolerationsK1),
+				test.BuildTestPod("p1", 100, 0, "worker1", setTolerationsK1),
+				test.BuildTestPod("p2", 100, 0, "worker1", setTolerationsK2),
+				test.BuildTestPod("p3", 100, 0, "worker1", setTolerationsK1),
+				test.BuildTestPod("p4", 100, 0, "worker1", setTolerationsK2),
+				test.BuildTestPod("p5", 100, 0, "worker1", setTolerationsK1),
+				test.BuildTestPod("p6", 100, 0, "worker2", setTolerationsK2),
+				test.BuildTestPod("p7", 100, 0, "worker2", setTolerationsK1),
+				test.BuildTestPod("p8", 100, 0, "worker2", setTolerationsK2),
+				test.BuildTestPod("p9", 100, 0, "worker3", setTolerationsK1),
 			},
 			expectedEvictedPodCount: 2,
 			nodes: []*v1.Node{
@@ -610,17 +641,17 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly respecting RequiredDuringSchedulingIgnoredDuringExecution node affinity",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1,0,0,0) -> (3,3,3,0,0,0) -> 2 evictions
-				*test.BuildTestPod("p1", 100, 0, "worker1", setNotMasterNodeSelectorK1),
-				*test.BuildTestPod("p2", 100, 0, "worker1", setNotMasterNodeSelectorK2),
-				*test.BuildTestPod("p3", 100, 0, "worker1", setNotMasterNodeSelectorK1),
-				*test.BuildTestPod("p4", 100, 0, "worker1", setNotMasterNodeSelectorK2),
-				*test.BuildTestPod("p5", 100, 0, "worker1", setNotMasterNodeSelectorK1),
-				*test.BuildTestPod("p6", 100, 0, "worker2", setNotMasterNodeSelectorK2),
-				*test.BuildTestPod("p7", 100, 0, "worker2", setNotMasterNodeSelectorK1),
-				*test.BuildTestPod("p8", 100, 0, "worker2", setNotMasterNodeSelectorK2),
-				*test.BuildTestPod("p9", 100, 0, "worker3", setNotMasterNodeSelectorK1),
+				test.BuildTestPod("p1", 100, 0, "worker1", setNotMasterNodeSelectorK1),
+				test.BuildTestPod("p2", 100, 0, "worker1", setNotMasterNodeSelectorK2),
+				test.BuildTestPod("p3", 100, 0, "worker1", setNotMasterNodeSelectorK1),
+				test.BuildTestPod("p4", 100, 0, "worker1", setNotMasterNodeSelectorK2),
+				test.BuildTestPod("p5", 100, 0, "worker1", setNotMasterNodeSelectorK1),
+				test.BuildTestPod("p6", 100, 0, "worker2", setNotMasterNodeSelectorK2),
+				test.BuildTestPod("p7", 100, 0, "worker2", setNotMasterNodeSelectorK1),
+				test.BuildTestPod("p8", 100, 0, "worker2", setNotMasterNodeSelectorK2),
+				test.BuildTestPod("p9", 100, 0, "worker3", setNotMasterNodeSelectorK1),
 			},
 			expectedEvictedPodCount: 2,
 			nodes: []*v1.Node{
@@ -635,17 +666,17 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly respecting node selector",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1,0,0,0) -> (3,3,3,0,0,0) -> 2 evictions
-				*test.BuildTestPod("p1", 100, 0, "worker1", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p2", 100, 0, "worker1", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p3", 100, 0, "worker1", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p4", 100, 0, "worker1", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p5", 100, 0, "worker1", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p6", 100, 0, "worker2", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p7", 100, 0, "worker2", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p8", 100, 0, "worker2", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p9", 100, 0, "worker3", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p1", 100, 0, "worker1", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p2", 100, 0, "worker1", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p3", 100, 0, "worker1", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p4", 100, 0, "worker1", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p5", 100, 0, "worker1", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p6", 100, 0, "worker2", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p7", 100, 0, "worker2", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p8", 100, 0, "worker2", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p9", 100, 0, "worker3", setWorkerLabelSelectorK1),
 			},
 			expectedEvictedPodCount: 2,
 			nodes: []*v1.Node{
@@ -660,17 +691,17 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		},
 		{
 			description: "Evict pods uniformly respecting node selector with zero target nodes",
-			pods: []v1.Pod{
+			pods: []*v1.Pod{
 				// (5,3,1,0,0,0) -> (3,3,3,0,0,0) -> 2 evictions
-				*test.BuildTestPod("p1", 100, 0, "worker1", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p2", 100, 0, "worker1", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p3", 100, 0, "worker1", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p4", 100, 0, "worker1", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p5", 100, 0, "worker1", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p6", 100, 0, "worker2", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p7", 100, 0, "worker2", setWorkerLabelSelectorK1),
-				*test.BuildTestPod("p8", 100, 0, "worker2", setWorkerLabelSelectorK2),
-				*test.BuildTestPod("p9", 100, 0, "worker3", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p1", 100, 0, "worker1", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p2", 100, 0, "worker1", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p3", 100, 0, "worker1", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p4", 100, 0, "worker1", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p5", 100, 0, "worker1", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p6", 100, 0, "worker2", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p7", 100, 0, "worker2", setWorkerLabelSelectorK1),
+				test.BuildTestPod("p8", 100, 0, "worker2", setWorkerLabelSelectorK2),
+				test.BuildTestPod("p9", 100, 0, "worker3", setWorkerLabelSelectorK1),
 			},
 			expectedEvictedPodCount: 0,
 			nodes: []*v1.Node{
@@ -687,22 +718,45 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 
 	for _, testCase := range testCases {
 		t.Run(testCase.description, func(t *testing.T) {
-			fakeClient := &fake.Clientset{}
-			fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-				return true, &v1.PodList{Items: testCase.pods}, nil
-			})
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range testCase.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range testCase.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
 			podEvictor := evictions.NewPodEvictor(
 				fakeClient,
 				policyv1.SchemeGroupVersion.String(),
 				false,
-				testCase.maxPodsToEvictPerNode,
+				nil,
+				nil,
 				testCase.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
 				false,
 				false,
 				false,
 			)
 
-			RemoveDuplicatePods(ctx, fakeClient, testCase.strategy, testCase.nodes, podEvictor)
+			RemoveDuplicatePods(ctx, fakeClient, testCase.strategy, testCase.nodes, podEvictor, getPodsAssignedToNode)
 			podsEvicted := podEvictor.TotalEvicted()
 			if podsEvicted != testCase.expectedEvictedPodCount {
 				t.Errorf("Test error for description: %s. Expected evicted pods count %v, got %v", testCase.description, testCase.expectedEvictedPodCount, podsEvicted)

pkg/descheduler/strategies/failedpods.go

@@ -0,0 +1,179 @@
+package strategies
+
+import (
+	"context"
+	"fmt"
+
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	utilerrors "k8s.io/apimachinery/pkg/util/errors"
+	"k8s.io/apimachinery/pkg/util/sets"
+	clientset "k8s.io/client-go/kubernetes"
+	"k8s.io/klog/v2"
+
+	"sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies/validation"
+)
+
+// validatedFailedPodsStrategyParams contains validated strategy parameters
+type validatedFailedPodsStrategyParams struct {
+	validation.ValidatedStrategyParams
+	includingInitContainers bool
+	reasons                 sets.String
+	excludeOwnerKinds       sets.String
+	minPodLifetimeSeconds   *uint
+}
+
+// RemoveFailedPods removes Pods that are in failed status phase.
+func RemoveFailedPods(
+	ctx context.Context,
+	client clientset.Interface,
+	strategy api.DeschedulerStrategy,
+	nodes []*v1.Node,
+	podEvictor *evictions.PodEvictor,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
+) {
+	strategyParams, err := validateAndParseRemoveFailedPodsParams(ctx, client, strategy.Params)
+	if err != nil {
+		klog.ErrorS(err, "Invalid RemoveFailedPods parameters")
+		return
+	}
+
+	evictable := podEvictor.Evictable(
+		evictions.WithPriorityThreshold(strategyParams.ThresholdPriority),
+		evictions.WithNodeFit(strategyParams.NodeFit),
+		evictions.WithLabelSelector(strategyParams.LabelSelector),
+	)
+
+	var labelSelector *metav1.LabelSelector
+	if strategy.Params != nil {
+		labelSelector = strategy.Params.LabelSelector
+	}
+
+	podFilter, err := podutil.NewOptions().
+		WithFilter(evictable.IsEvictable).
+		WithNamespaces(strategyParams.IncludedNamespaces).
+		WithoutNamespaces(strategyParams.ExcludedNamespaces).
+		WithLabelSelector(labelSelector).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+	// Only list failed pods
+	phaseFilter := func(pod *v1.Pod) bool { return pod.Status.Phase == v1.PodFailed }
+	podFilter = podutil.WrapFilterFuncs(phaseFilter, podFilter)
+
+	for _, node := range nodes {
+		klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
+		pods, err := podutil.ListAllPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
+		if err != nil {
+			klog.ErrorS(err, "Error listing a nodes failed pods", "node", klog.KObj(node))
+			continue
+		}
+
+		for i, pod := range pods {
+			if err = validateFailedPodShouldEvict(pod, *strategyParams); err != nil {
+				klog.V(4).InfoS(fmt.Sprintf("ignoring pod for eviction due to: %s", err.Error()), "pod", klog.KObj(pod))
+				continue
+			}
+
+			if _, err = podEvictor.EvictPod(ctx, pods[i], node, "FailedPod"); err != nil {
+				klog.ErrorS(err, "Error evicting pod", "pod", klog.KObj(pod))
+				break
+			}
+		}
+	}
+}
+
+func validateAndParseRemoveFailedPodsParams(
+	ctx context.Context,
+	client clientset.Interface,
+	params *api.StrategyParameters,
+) (*validatedFailedPodsStrategyParams, error) {
+	if params == nil {
+		return &validatedFailedPodsStrategyParams{
+			ValidatedStrategyParams: validation.DefaultValidatedStrategyParams(),
+		}, nil
+	}
+
+	strategyParams, err := validation.ValidateAndParseStrategyParams(ctx, client, params)
+	if err != nil {
+		return nil, err
+	}
+
+	var reasons, excludeOwnerKinds sets.String
+	var includingInitContainers bool
+	var minPodLifetimeSeconds *uint
+	if params.FailedPods != nil {
+		reasons = sets.NewString(params.FailedPods.Reasons...)
+		includingInitContainers = params.FailedPods.IncludingInitContainers
+		excludeOwnerKinds = sets.NewString(params.FailedPods.ExcludeOwnerKinds...)
+		minPodLifetimeSeconds = params.FailedPods.MinPodLifetimeSeconds
+	}
+
+	return &validatedFailedPodsStrategyParams{
+		ValidatedStrategyParams: *strategyParams,
+		includingInitContainers: includingInitContainers,
+		reasons:                 reasons,
+		excludeOwnerKinds:       excludeOwnerKinds,
+		minPodLifetimeSeconds:   minPodLifetimeSeconds,
+	}, nil
+}
+
+// validateFailedPodShouldEvict looks at strategy params settings to see if the Pod
+// should be evicted given the params in the PodFailed policy.
+func validateFailedPodShouldEvict(pod *v1.Pod, strategyParams validatedFailedPodsStrategyParams) error {
+	var errs []error
+
+	if strategyParams.minPodLifetimeSeconds != nil {
+		podAgeSeconds := uint(metav1.Now().Sub(pod.GetCreationTimestamp().Local()).Seconds())
+		if podAgeSeconds < *strategyParams.minPodLifetimeSeconds {
+			errs = append(errs, fmt.Errorf("pod does not exceed the min age seconds of %d", *strategyParams.minPodLifetimeSeconds))
+		}
+	}
+
+	if len(strategyParams.excludeOwnerKinds) > 0 {
+		ownerRefList := podutil.OwnerRef(pod)
+		for _, owner := range ownerRefList {
+			if strategyParams.excludeOwnerKinds.Has(owner.Kind) {
+				errs = append(errs, fmt.Errorf("pod's owner kind of %s is excluded", owner.Kind))
+			}
+		}
+	}
+
+	if len(strategyParams.reasons) > 0 {
+		reasons := getFailedContainerStatusReasons(pod.Status.ContainerStatuses)
+
+		if pod.Status.Phase == v1.PodFailed && pod.Status.Reason != "" {
+			reasons = append(reasons, pod.Status.Reason)
+		}
+
+		if strategyParams.includingInitContainers {
+			reasons = append(reasons, getFailedContainerStatusReasons(pod.Status.InitContainerStatuses)...)
+		}
+
+		if !strategyParams.reasons.HasAny(reasons...) {
+			errs = append(errs, fmt.Errorf("pod does not match any of the reasons"))
+		}
+	}
+
+	return utilerrors.NewAggregate(errs)
+}
+
+func getFailedContainerStatusReasons(containerStatuses []v1.ContainerStatus) []string {
+	reasons := make([]string, 0)
+
+	for _, containerStatus := range containerStatuses {
+		if containerStatus.State.Waiting != nil && containerStatus.State.Waiting.Reason != "" {
+			reasons = append(reasons, containerStatus.State.Waiting.Reason)
+		}
+		if containerStatus.State.Terminated != nil && containerStatus.State.Terminated.Reason != "" {
+			reasons = append(reasons, containerStatus.State.Terminated.Reason)
+		}
+	}
+
+	return reasons
+}

pkg/descheduler/strategies/failedpods_test.go

@@ -0,0 +1,358 @@
+package strategies
+
+import (
+	"context"
+	"testing"
+
+	v1 "k8s.io/api/core/v1"
+	policyv1 "k8s.io/api/policy/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
+	"k8s.io/client-go/kubernetes/fake"
+
+	"sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
+	"sigs.k8s.io/descheduler/test"
+)
+
+var (
+	OneHourInSeconds uint = 3600
+)
+
+func TestRemoveFailedPods(t *testing.T) {
+	createStrategy := func(enabled, includingInitContainers bool, reasons, excludeKinds []string, minAgeSeconds *uint, nodeFit bool) api.DeschedulerStrategy {
+		return api.DeschedulerStrategy{
+			Enabled: enabled,
+			Params: &api.StrategyParameters{
+				FailedPods: &api.FailedPods{
+					Reasons:                 reasons,
+					IncludingInitContainers: includingInitContainers,
+					ExcludeOwnerKinds:       excludeKinds,
+					MinPodLifetimeSeconds:   minAgeSeconds,
+				},
+				NodeFit: nodeFit,
+			},
+		}
+	}
+
+	tests := []struct {
+		description             string
+		nodes                   []*v1.Node
+		strategy                api.DeschedulerStrategy
+		expectedEvictedPodCount uint
+		pods                    []*v1.Pod
+	}{
+		{
+			description:             "default empty strategy, 0 failures, 0 evictions",
+			strategy:                api.DeschedulerStrategy{},
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods:                    []*v1.Pod{}, // no pods come back with field selector phase=Failed
+		},
+		{
+			description:             "0 failures, 0 evictions",
+			strategy:                createStrategy(true, false, nil, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods:                    []*v1.Pod{}, // no pods come back with field selector phase=Failed
+		},
+		{
+			description:             "1 container terminated with reason NodeAffinity, 1 eviction",
+			strategy:                createStrategy(true, false, nil, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 1,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}), nil),
+			},
+		},
+		{
+			description:             "1 init container terminated with reason NodeAffinity, 1 eviction",
+			strategy:                createStrategy(true, true, nil, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 1,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}, nil), nil),
+			},
+		},
+		{
+			description:             "1 init container waiting with reason CreateContainerConfigError, 1 eviction",
+			strategy:                createStrategy(true, true, nil, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 1,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Waiting: &v1.ContainerStateWaiting{Reason: "CreateContainerConfigError"},
+				}, nil), nil),
+			},
+		},
+		{
+			description: "2 init container waiting with reason CreateContainerConfigError, 2 nodes, 2 evictions",
+			strategy:    createStrategy(true, true, nil, nil, nil, false),
+			nodes: []*v1.Node{
+				test.BuildTestNode("node1", 2000, 3000, 10, nil),
+				test.BuildTestNode("node2", 2000, 3000, 10, nil),
+			},
+			expectedEvictedPodCount: 2,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "CreateContainerConfigError"},
+				}, nil), nil),
+				buildTestPod("p2", "node2", newPodStatus("", "", &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "CreateContainerConfigError"},
+				}, nil), nil),
+			},
+		},
+		{
+			description:             "include reason=CreateContainerConfigError, 1 container terminated with reason CreateContainerConfigError, 1 eviction",
+			strategy:                createStrategy(true, false, []string{"CreateContainerConfigError"}, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 1,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "CreateContainerConfigError"},
+				}), nil),
+			},
+		},
+		{
+			description:             "include reason=CreateContainerConfigError+NodeAffinity, 1 container terminated with reason CreateContainerConfigError, 1 eviction",
+			strategy:                createStrategy(true, false, []string{"CreateContainerConfigError", "NodeAffinity"}, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 1,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "CreateContainerConfigError"},
+				}), nil),
+			},
+		},
+		{
+			description:             "include reason=CreateContainerConfigError, 1 container terminated with reason NodeAffinity, 0 eviction",
+			strategy:                createStrategy(true, false, []string{"CreateContainerConfigError"}, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}), nil),
+			},
+		},
+		{
+			description:             "include init container=false, 1 init container waiting with reason CreateContainerConfigError, 0 eviction",
+			strategy:                createStrategy(true, false, []string{"CreateContainerConfigError"}, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Waiting: &v1.ContainerStateWaiting{Reason: "CreateContainerConfigError"},
+				}, nil), nil),
+			},
+		},
+		{
+			description:             "lifetime 1 hour, 1 container terminated with reason NodeAffinity, 0 eviction",
+			strategy:                createStrategy(true, false, nil, nil, &OneHourInSeconds, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}), nil),
+			},
+		},
+		{
+			description: "nodeFit=true, 1 unschedulable node, 1 container terminated with reason NodeAffinity, 0 eviction",
+			strategy:    createStrategy(true, false, nil, nil, nil, true),
+			nodes: []*v1.Node{
+				test.BuildTestNode("node1", 2000, 3000, 10, nil),
+				test.BuildTestNode("node2", 2000, 2000, 10, func(node *v1.Node) {
+					node.Spec.Unschedulable = true
+				}),
+			},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}), nil),
+			},
+		},
+		{
+			description:             "nodeFit=true, only available node does not have enough resources, 1 container terminated with reason CreateContainerConfigError, 0 eviction",
+			strategy:                createStrategy(true, false, []string{"CreateContainerConfigError"}, nil, nil, true),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 1, 1, 10, nil), test.BuildTestNode("node2", 0, 0, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", nil, &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "CreateContainerConfigError"},
+				}), nil),
+			},
+		},
+		{
+			description:             "excluded owner kind=ReplicaSet, 1 init container terminated with owner kind=ReplicaSet, 0 eviction",
+			strategy:                createStrategy(true, true, nil, []string{"ReplicaSet"}, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}, nil), nil),
+			},
+		},
+		{
+			description:             "excluded owner kind=DaemonSet, 1 init container terminated with owner kind=ReplicaSet, 1 eviction",
+			strategy:                createStrategy(true, true, nil, []string{"DaemonSet"}, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 1,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}, nil), nil),
+			},
+		},
+		{
+			description:             "excluded owner kind=DaemonSet, 1 init container terminated with owner kind=ReplicaSet, 1 pod in termination; nothing should be moved",
+			strategy:                createStrategy(true, true, nil, []string{"DaemonSet"}, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("", "", &v1.ContainerState{
+					Terminated: &v1.ContainerStateTerminated{Reason: "NodeAffinity"},
+				}, nil), &metav1.Time{}),
+			},
+		},
+		{
+			description:             "1 container terminated with reason ShutDown, 0 evictions",
+			strategy:                createStrategy(true, false, nil, nil, nil, true),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 0,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("Shutdown", v1.PodFailed, nil, nil), nil),
+			},
+		},
+		{
+			description:             "include reason=Shutdown, 2 containers terminated with reason ShutDown, 2 evictions",
+			strategy:                createStrategy(true, false, []string{"Shutdown"}, nil, nil, false),
+			nodes:                   []*v1.Node{test.BuildTestNode("node1", 2000, 3000, 10, nil)},
+			expectedEvictedPodCount: 2,
+			pods: []*v1.Pod{
+				buildTestPod("p1", "node1", newPodStatus("Shutdown", v1.PodFailed, nil, nil), nil),
+				buildTestPod("p2", "node1", newPodStatus("Shutdown", v1.PodFailed, nil, nil), nil),
+			},
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range tc.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := evictions.NewPodEvictor(
+				fakeClient,
+				policyv1.SchemeGroupVersion.String(),
+				false,
+				nil,
+				nil,
+				tc.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
+				false,
+				false,
+				false,
+			)
+
+			RemoveFailedPods(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor, getPodsAssignedToNode)
+			actualEvictedPodCount := podEvictor.TotalEvicted()
+			if actualEvictedPodCount != tc.expectedEvictedPodCount {
+				t.Errorf("Test %#v failed, expected %v pod evictions, but got %v pod evictions\n", tc.description, tc.expectedEvictedPodCount, actualEvictedPodCount)
+			}
+		})
+	}
+}
+
+func TestValidRemoveFailedPodsParams(t *testing.T) {
+	ctx := context.Background()
+	fakeClient := &fake.Clientset{}
+	testCases := []struct {
+		name   string
+		params *api.StrategyParameters
+	}{
+		{name: "validate nil params", params: nil},
+		{name: "validate empty params", params: &api.StrategyParameters{}},
+		{name: "validate reasons params", params: &api.StrategyParameters{FailedPods: &api.FailedPods{
+			Reasons: []string{"CreateContainerConfigError"},
+		}}},
+		{name: "validate includingInitContainers params", params: &api.StrategyParameters{FailedPods: &api.FailedPods{
+			IncludingInitContainers: true,
+		}}},
+		{name: "validate excludeOwnerKinds params", params: &api.StrategyParameters{FailedPods: &api.FailedPods{
+			ExcludeOwnerKinds: []string{"Job"},
+		}}},
+		{name: "validate excludeOwnerKinds params", params: &api.StrategyParameters{FailedPods: &api.FailedPods{
+			MinPodLifetimeSeconds: &OneHourInSeconds,
+		}}},
+	}
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			params, err := validateAndParseRemoveFailedPodsParams(ctx, fakeClient, tc.params)
+			if err != nil {
+				t.Errorf("strategy params should be valid but got err: %v", err.Error())
+			}
+			if params == nil {
+				t.Errorf("strategy params should return a ValidatedFailedPodsStrategyParams but got nil")
+			}
+		})
+	}
+}
+
+func newPodStatus(reason string, phase v1.PodPhase, initContainerState, containerState *v1.ContainerState) v1.PodStatus {
+	ps := v1.PodStatus{
+		Reason: reason,
+		Phase:  phase,
+	}
+
+	if initContainerState != nil {
+		ps.InitContainerStatuses = []v1.ContainerStatus{{State: *initContainerState}}
+		ps.Phase = v1.PodFailed
+	}
+
+	if containerState != nil {
+		ps.ContainerStatuses = []v1.ContainerStatus{{State: *containerState}}
+		ps.Phase = v1.PodFailed
+	}
+
+	return ps
+}
+
+func buildTestPod(podName, nodeName string, podStatus v1.PodStatus, deletionTimestamp *metav1.Time) *v1.Pod {
+	pod := test.BuildTestPod(podName, 1, 1, nodeName, func(p *v1.Pod) {
+		p.Status = podStatus
+	})
+	pod.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
+	pod.ObjectMeta.SetCreationTimestamp(metav1.Now())
+	pod.DeletionTimestamp = deletionTimestamp
+	return pod
+}

pkg/descheduler/strategies/node_affinity.go

@@ -21,6 +21,7 @@ import (
 	"fmt"
 
 	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/util/sets"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
 
@@ -47,7 +48,7 @@ func validatePodsViolatingNodeAffinityParams(params *api.StrategyParameters) err
 }
 
 // RemovePodsViolatingNodeAffinity evicts pods on nodes which violate node affinity
-func RemovePodsViolatingNodeAffinity(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func RemovePodsViolatingNodeAffinity(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	if err := validatePodsViolatingNodeAffinityParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid RemovePodsViolatingNodeAffinity parameters")
 		return
@@ -58,10 +59,10 @@ func RemovePodsViolatingNodeAffinity(ctx context.Context, client clientset.Inter
 		return
 	}
 
-	var includedNamespaces, excludedNamespaces []string
+	var includedNamespaces, excludedNamespaces sets.String
 	if strategy.Params.Namespaces != nil {
-		includedNamespaces = strategy.Params.Namespaces.Include
-		excludedNamespaces = strategy.Params.Namespaces.Exclude
+		includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+		excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
 	}
 
 	nodeFit := false
@@ -71,6 +72,16 @@ func RemovePodsViolatingNodeAffinity(ctx context.Context, client clientset.Inter
 
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
+	podFilter, err := podutil.NewOptions().
+		WithNamespaces(includedNamespaces).
+		WithoutNamespaces(excludedNamespaces).
+		WithLabelSelector(strategy.Params.LabelSelector).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+
 	for _, nodeAffinity := range strategy.Params.NodeAffinityType {
 		klog.V(2).InfoS("Executing for nodeAffinityType", "nodeAffinity", nodeAffinity)
 
@@ -80,17 +91,13 @@ func RemovePodsViolatingNodeAffinity(ctx context.Context, client clientset.Inter
 				klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
 
 				pods, err := podutil.ListPodsOnANode(
-					ctx,
-					client,
-					node,
-					podutil.WithFilter(func(pod *v1.Pod) bool {
+					node.Name,
+					getPodsAssignedToNode,
+					podutil.WrapFilterFuncs(podFilter, func(pod *v1.Pod) bool {
 						return evictable.IsEvictable(pod) &&
-							!nodeutil.PodFitsCurrentNode(pod, node) &&
-							nodeutil.PodFitsAnyNode(pod, nodes)
+							!nodeutil.PodFitsCurrentNode(getPodsAssignedToNode, pod, node) &&
+							nodeutil.PodFitsAnyNode(getPodsAssignedToNode, pod, nodes)
 					}),
-					podutil.WithNamespaces(includedNamespaces),
-					podutil.WithoutNamespaces(excludedNamespaces),
-					podutil.WithLabelSelector(strategy.Params.LabelSelector),
 				)
 				if err != nil {
 					klog.ErrorS(err, "Failed to get pods", "node", klog.KObj(node))

pkg/descheduler/strategies/node_affinity_test.go

@@ -22,16 +22,18 @@ import (
 
 	v1 "k8s.io/api/core/v1"
 	policyv1 "k8s.io/api/policy/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/test"
 )
 
 func TestRemovePodsViolatingNodeAffinity(t *testing.T) {
-	ctx := context.Background()
 	requiredDuringSchedulingIgnoredDuringExecutionStrategy := api.DeschedulerStrategy{
 		Enabled: true,
 		Params: &api.StrategyParameters{
@@ -59,10 +61,10 @@ func TestRemovePodsViolatingNodeAffinity(t *testing.T) {
 	nodeWithoutLabels := test.BuildTestNode("nodeWithoutLabels", 2000, 3000, 10, nil)
 
 	unschedulableNodeWithLabels := test.BuildTestNode("unschedulableNodeWithLabels", 2000, 3000, 10, nil)
-	nodeWithLabels.Labels[nodeLabelKey] = nodeLabelValue
+	unschedulableNodeWithLabels.Labels[nodeLabelKey] = nodeLabelValue
 	unschedulableNodeWithLabels.Spec.Unschedulable = true
 
-	addPodsToNode := func(node *v1.Node) []v1.Pod {
+	addPodsToNode := func(node *v1.Node, deletionTimestamp *metav1.Time) []*v1.Pod {
 		podWithNodeAffinity := test.BuildTestPod("podWithNodeAffinity", 100, 0, node.Name, nil)
 		podWithNodeAffinity.Spec.Affinity = &v1.Affinity{
 			NodeAffinity: &v1.NodeAffinity{
@@ -91,20 +93,25 @@ func TestRemovePodsViolatingNodeAffinity(t *testing.T) {
 		pod1.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 		pod2.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 
-		return []v1.Pod{
-			*podWithNodeAffinity,
-			*pod1,
-			*pod2,
+		pod1.DeletionTimestamp = deletionTimestamp
+		pod2.DeletionTimestamp = deletionTimestamp
+
+		return []*v1.Pod{
+			podWithNodeAffinity,
+			pod1,
+			pod2,
 		}
 	}
 
+	var uint1 uint = 1
 	tests := []struct {
-		description             string
-		nodes                   []*v1.Node
-		pods                    []v1.Pod
-		strategy                api.DeschedulerStrategy
-		expectedEvictedPodCount int
-		maxPodsToEvictPerNode   int
+		description                    string
+		nodes                          []*v1.Node
+		pods                           []*v1.Pod
+		strategy                       api.DeschedulerStrategy
+		expectedEvictedPodCount        uint
+		maxPodsToEvictPerNode          *uint
+		maxNoOfPodsToEvictPerNamespace *uint
 	}{
 		{
 			description: "Invalid strategy type, should not evict any pods",
@@ -117,74 +124,117 @@ func TestRemovePodsViolatingNodeAffinity(t *testing.T) {
 				},
 			},
 			expectedEvictedPodCount: 0,
-			pods:                    addPodsToNode(nodeWithoutLabels),
+			pods:                    addPodsToNode(nodeWithoutLabels, nil),
 			nodes:                   []*v1.Node{nodeWithoutLabels, nodeWithLabels},
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Pod is correctly scheduled on node, no eviction expected",
 			strategy:                requiredDuringSchedulingIgnoredDuringExecutionStrategy,
 			expectedEvictedPodCount: 0,
-			pods:                    addPodsToNode(nodeWithLabels),
+			pods:                    addPodsToNode(nodeWithLabels, nil),
 			nodes:                   []*v1.Node{nodeWithLabels},
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Pod is scheduled on node without matching labels, another schedulable node available, should be evicted",
 			expectedEvictedPodCount: 1,
 			strategy:                requiredDuringSchedulingIgnoredDuringExecutionStrategy,
-			pods:                    addPodsToNode(nodeWithoutLabels),
+			pods:                    addPodsToNode(nodeWithoutLabels, nil),
 			nodes:                   []*v1.Node{nodeWithoutLabels, nodeWithLabels},
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Pod is scheduled on node without matching labels, another schedulable node available, maxPodsToEvictPerNode set to 1, should not be evicted",
 			expectedEvictedPodCount: 1,
 			strategy:                requiredDuringSchedulingIgnoredDuringExecutionStrategy,
-			pods:                    addPodsToNode(nodeWithoutLabels),
+			pods:                    addPodsToNode(nodeWithoutLabels, nil),
 			nodes:                   []*v1.Node{nodeWithoutLabels, nodeWithLabels},
-			maxPodsToEvictPerNode:   1,
+			maxPodsToEvictPerNode:   &uint1,
+		},
+		{
+			description:             "Pod is scheduled on node without matching labels, another schedulable node available, maxPodsToEvictPerNode set to 1, no pod evicted since pod terminting",
+			expectedEvictedPodCount: 1,
+			strategy:                requiredDuringSchedulingIgnoredDuringExecutionStrategy,
+			pods:                    addPodsToNode(nodeWithoutLabels, &metav1.Time{}),
+			nodes:                   []*v1.Node{nodeWithoutLabels, nodeWithLabels},
+			maxPodsToEvictPerNode:   &uint1,
+		},
+		{
+			description:                    "Pod is scheduled on node without matching labels, another schedulable node available, maxNoOfPodsToEvictPerNamespace set to 1, should not be evicted",
+			expectedEvictedPodCount:        1,
+			strategy:                       requiredDuringSchedulingIgnoredDuringExecutionStrategy,
+			pods:                           addPodsToNode(nodeWithoutLabels, nil),
+			nodes:                          []*v1.Node{nodeWithoutLabels, nodeWithLabels},
+			maxNoOfPodsToEvictPerNamespace: &uint1,
+		},
+		{
+			description:                    "Pod is scheduled on node without matching labels, another schedulable node available, maxNoOfPodsToEvictPerNamespace set to 1, no pod evicted since pod terminting",
+			expectedEvictedPodCount:        1,
+			strategy:                       requiredDuringSchedulingIgnoredDuringExecutionStrategy,
+			pods:                           addPodsToNode(nodeWithoutLabels, &metav1.Time{}),
+			nodes:                          []*v1.Node{nodeWithoutLabels, nodeWithLabels},
+			maxNoOfPodsToEvictPerNamespace: &uint1,
 		},
 		{
 			description:             "Pod is scheduled on node without matching labels, but no node where pod fits is available, should not evict",
 			expectedEvictedPodCount: 0,
 			strategy:                requiredDuringSchedulingIgnoredDuringExecutionWithNodeFitStrategy,
-			pods:                    addPodsToNode(nodeWithoutLabels),
+			pods:                    addPodsToNode(nodeWithoutLabels, nil),
 			nodes:                   []*v1.Node{nodeWithoutLabels, unschedulableNodeWithLabels},
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Pod is scheduled on node without matching labels, and node where pod fits is available, should evict",
 			expectedEvictedPodCount: 0,
 			strategy:                requiredDuringSchedulingIgnoredDuringExecutionWithNodeFitStrategy,
-			pods:                    addPodsToNode(nodeWithoutLabels),
+			pods:                    addPodsToNode(nodeWithoutLabels, nil),
 			nodes:                   []*v1.Node{nodeWithLabels, unschedulableNodeWithLabels},
-			maxPodsToEvictPerNode:   1,
+			maxPodsToEvictPerNode:   &uint1,
 		},
 	}
 
 	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
 
-		fakeClient := &fake.Clientset{}
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			return true, &v1.PodList{Items: tc.pods}, nil
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range tc.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := evictions.NewPodEvictor(
+				fakeClient,
+				policyv1.SchemeGroupVersion.String(),
+				false,
+				tc.maxPodsToEvictPerNode,
+				tc.maxNoOfPodsToEvictPerNamespace,
+				tc.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
+				false,
+				false,
+				false,
+			)
+
+			RemovePodsViolatingNodeAffinity(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor, getPodsAssignedToNode)
+			actualEvictedPodCount := podEvictor.TotalEvicted()
+			if actualEvictedPodCount != tc.expectedEvictedPodCount {
+				t.Errorf("Test %#v failed, expected %v pod evictions, but got %v pod evictions\n", tc.description, tc.expectedEvictedPodCount, actualEvictedPodCount)
+			}
 		})
-
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			policyv1.SchemeGroupVersion.String(),
-			false,
-			tc.maxPodsToEvictPerNode,
-			tc.nodes,
-			false,
-			false,
-			false,
-		)
-
-		RemovePodsViolatingNodeAffinity(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor)
-		actualEvictedPodCount := podEvictor.TotalEvicted()
-		if actualEvictedPodCount != tc.expectedEvictedPodCount {
-			t.Errorf("Test %#v failed, expected %v pod evictions, but got %v pod evictions\n", tc.description, tc.expectedEvictedPodCount, actualEvictedPodCount)
-		}
 	}
 }

pkg/descheduler/strategies/node_taint.go

@@ -20,6 +20,7 @@ import (
 	"context"
 	"fmt"
 
+	"k8s.io/apimachinery/pkg/util/sets"
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
@@ -48,18 +49,21 @@ func validateRemovePodsViolatingNodeTaintsParams(params *api.StrategyParameters)
 }
 
 // RemovePodsViolatingNodeTaints evicts pods on the node which violate NoSchedule Taints on nodes
-func RemovePodsViolatingNodeTaints(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func RemovePodsViolatingNodeTaints(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	if err := validateRemovePodsViolatingNodeTaintsParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid RemovePodsViolatingNodeTaints parameters")
 		return
 	}
 
-	var includedNamespaces, excludedNamespaces []string
+	var includedNamespaces, excludedNamespaces, excludedTaints sets.String
 	var labelSelector *metav1.LabelSelector
 	if strategy.Params != nil {
 		if strategy.Params.Namespaces != nil {
-			includedNamespaces = strategy.Params.Namespaces.Include
-			excludedNamespaces = strategy.Params.Namespaces.Exclude
+			includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+			excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
+		}
+		if strategy.Params.ExcludedTaints != nil {
+			excludedTaints = sets.NewString(strategy.Params.ExcludedTaints...)
 		}
 		labelSelector = strategy.Params.LabelSelector
 	}
@@ -77,17 +81,32 @@ func RemovePodsViolatingNodeTaints(ctx context.Context, client clientset.Interfa
 
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
+	podFilter, err := podutil.NewOptions().
+		WithFilter(evictable.IsEvictable).
+		WithNamespaces(includedNamespaces).
+		WithoutNamespaces(excludedNamespaces).
+		WithLabelSelector(labelSelector).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+
+	excludeTaint := func(taint *v1.Taint) bool {
+		// Exclude taints by key *or* key=value
+		return excludedTaints.Has(taint.Key) || (taint.Value != "" && excludedTaints.Has(fmt.Sprintf("%s=%s", taint.Key, taint.Value)))
+	}
+
+	taintFilterFnc := func(taint *v1.Taint) bool { return (taint.Effect == v1.TaintEffectNoSchedule) && !excludeTaint(taint) }
+	if strategy.Params != nil && strategy.Params.IncludePreferNoSchedule {
+		taintFilterFnc = func(taint *v1.Taint) bool {
+			return (taint.Effect == v1.TaintEffectNoSchedule || taint.Effect == v1.TaintEffectPreferNoSchedule) && !excludeTaint(taint)
+		}
+	}
+
 	for _, node := range nodes {
 		klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
-		pods, err := podutil.ListPodsOnANode(
-			ctx,
-			client,
-			node,
-			podutil.WithFilter(evictable.IsEvictable),
-			podutil.WithNamespaces(includedNamespaces),
-			podutil.WithoutNamespaces(excludedNamespaces),
-			podutil.WithLabelSelector(labelSelector),
-		)
+		pods, err := podutil.ListAllPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
 		if err != nil {
 			//no pods evicted as error encountered retrieving evictable Pods
 			return
@@ -97,7 +116,7 @@ func RemovePodsViolatingNodeTaints(ctx context.Context, client clientset.Interfa
 			if !utils.TolerationsTolerateTaintsWithFilter(
 				pods[i].Spec.Tolerations,
 				node.Spec.Taints,
-				func(taint *v1.Taint) bool { return taint.Effect == v1.TaintEffectNoSchedule },
+				taintFilterFnc,
 			) {
 				klog.V(2).InfoS("Not all taints with NoSchedule effect are tolerated after update for pod on node", "pod", klog.KObj(pods[i]), "node", klog.KObj(node))
 				if _, err := podEvictor.EvictPod(ctx, pods[i], node, "NodeTaint"); err != nil {

pkg/descheduler/strategies/node_taint_test.go

@@ -9,10 +9,12 @@ import (
 	policyv1 "k8s.io/api/policy/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
@@ -25,6 +27,14 @@ func createNoScheduleTaint(key, value string, index int) v1.Taint {
 	}
 }
 
+func createPreferNoScheduleTaint(key, value string, index int) v1.Taint {
+	return v1.Taint{
+		Key:    "testTaint" + fmt.Sprintf("%v", index),
+		Value:  "test" + fmt.Sprintf("%v", index),
+		Effect: v1.TaintEffectPreferNoSchedule,
+	}
+}
+
 func addTaintsToNode(node *v1.Node, key, value string, indices []int) *v1.Node {
 	taints := []v1.Taint{}
 	for _, index := range indices {
@@ -34,22 +44,21 @@ func addTaintsToNode(node *v1.Node, key, value string, indices []int) *v1.Node {
 	return node
 }
 
-func addTolerationToPod(pod *v1.Pod, key, value string, index int) *v1.Pod {
+func addTolerationToPod(pod *v1.Pod, key, value string, index int, effect v1.TaintEffect) *v1.Pod {
 	if pod.Annotations == nil {
 		pod.Annotations = map[string]string{}
 	}
 
-	pod.Spec.Tolerations = []v1.Toleration{{Key: key + fmt.Sprintf("%v", index), Value: value + fmt.Sprintf("%v", index), Effect: v1.TaintEffectNoSchedule}}
+	pod.Spec.Tolerations = []v1.Toleration{{Key: key + fmt.Sprintf("%v", index), Value: value + fmt.Sprintf("%v", index), Effect: effect}}
 
 	return pod
 }
 
 func TestDeletePodsViolatingNodeTaints(t *testing.T) {
-	ctx := context.Background()
 	node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
 	node1 = addTaintsToNode(node1, "testTaint", "test", []int{1})
 	node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
-	node1 = addTaintsToNode(node2, "testingTaint", "testing", []int{1})
+	node2 = addTaintsToNode(node2, "testingTaint", "testing", []int{1})
 
 	node3 := test.BuildTestNode("n3", 2000, 3000, 10, func(node *v1.Node) {
 		node.ObjectMeta.Labels = map[string]string{
@@ -62,6 +71,16 @@ func TestDeletePodsViolatingNodeTaints(t *testing.T) {
 		}
 	})
 
+	node5 := test.BuildTestNode("n5", 2000, 3000, 10, nil)
+	node5.Spec.Taints = []v1.Taint{
+		createPreferNoScheduleTaint("testTaint", "test", 1),
+	}
+
+	node6 := test.BuildTestNode("n6", 1, 1, 1, nil)
+	node6.Spec.Taints = []v1.Taint{
+		createPreferNoScheduleTaint("testTaint", "test", 1),
+	}
+
 	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
 	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, nil)
 	p3 := test.BuildTestPod("p3", 100, 0, node1.Name, nil)
@@ -108,152 +127,240 @@ func TestDeletePodsViolatingNodeTaints(t *testing.T) {
 	// A Mirror Pod.
 	p10.Annotations = test.GetMirrorPodAnnotation()
 
-	p1 = addTolerationToPod(p1, "testTaint", "test", 1)
-	p3 = addTolerationToPod(p3, "testTaint", "test", 1)
-	p4 = addTolerationToPod(p4, "testTaintX", "testX", 1)
+	p1 = addTolerationToPod(p1, "testTaint", "test", 1, v1.TaintEffectNoSchedule)
+	p3 = addTolerationToPod(p3, "testTaint", "test", 1, v1.TaintEffectNoSchedule)
+	p4 = addTolerationToPod(p4, "testTaintX", "testX", 1, v1.TaintEffectNoSchedule)
 
 	p12.Spec.NodeSelector = map[string]string{
 		"datacenter": "west",
 	}
 
+	p13 := test.BuildTestPod("p13", 100, 0, node5.Name, nil)
+	p13.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+	// node5 has PreferNoSchedule:testTaint1=test1, so the p13 has to have
+	// PreferNoSchedule:testTaint0=test0 so the pod is not tolarated
+	p13 = addTolerationToPod(p13, "testTaint", "test", 0, v1.TaintEffectPreferNoSchedule)
+
+	var uint1 uint = 1
+
 	tests := []struct {
-		description             string
-		nodes                   []*v1.Node
-		pods                    []v1.Pod
-		evictLocalStoragePods   bool
-		evictSystemCriticalPods bool
-		maxPodsToEvictPerNode   int
-		expectedEvictedPodCount int
-		nodeFit                 bool
+		description                    string
+		nodes                          []*v1.Node
+		pods                           []*v1.Pod
+		evictLocalStoragePods          bool
+		evictSystemCriticalPods        bool
+		maxPodsToEvictPerNode          *uint
+		maxNoOfPodsToEvictPerNamespace *uint
+		expectedEvictedPodCount        uint
+		nodeFit                        bool
+		includePreferNoSchedule        bool
+		excludedTaints                 []string
 	}{
 
 		{
 			description:             "Pods not tolerating node taint should be evicted",
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 1, //p2 gets evicted
 		},
 		{
 			description:             "Pods with tolerations but not tolerating node taint should be evicted",
-			pods:                    []v1.Pod{*p1, *p3, *p4},
+			pods:                    []*v1.Pod{p1, p3, p4},
 			nodes:                   []*v1.Node{node1},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 1, //p4 gets evicted
 		},
 		{
 			description:             "Only <maxPodsToEvictPerNode> number of Pods not tolerating node taint should be evicted",
-			pods:                    []v1.Pod{*p1, *p5, *p6},
+			pods:                    []*v1.Pod{p1, p5, p6},
 			nodes:                   []*v1.Node{node1},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   1,
+			maxPodsToEvictPerNode:   &uint1,
 			expectedEvictedPodCount: 1, //p5 or p6 gets evicted
 		},
+		{
+			description:                    "Only <maxNoOfPodsToEvictPerNamespace> number of Pods not tolerating node taint should be evicted",
+			pods:                           []*v1.Pod{p1, p5, p6},
+			nodes:                          []*v1.Node{node1},
+			evictLocalStoragePods:          false,
+			evictSystemCriticalPods:        false,
+			maxNoOfPodsToEvictPerNamespace: &uint1,
+			expectedEvictedPodCount:        1, //p5 or p6 gets evicted
+		},
 		{
 			description:             "Critical pods not tolerating node taint should not be evicted",
-			pods:                    []v1.Pod{*p7, *p8, *p9, *p10},
+			pods:                    []*v1.Pod{p7, p8, p9, p10},
 			nodes:                   []*v1.Node{node2},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 0, //nothing is evicted
 		},
 		{
 			description:             "Critical pods except storage pods not tolerating node taint should not be evicted",
-			pods:                    []v1.Pod{*p7, *p8, *p9, *p10},
+			pods:                    []*v1.Pod{p7, p8, p9, p10},
 			nodes:                   []*v1.Node{node2},
 			evictLocalStoragePods:   true,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 1, //p9 gets evicted
 		},
 		{
 			description:             "Critical and non critical pods, only non critical pods not tolerating node taint should be evicted",
-			pods:                    []v1.Pod{*p7, *p8, *p10, *p11},
+			pods:                    []*v1.Pod{p7, p8, p10, p11},
 			nodes:                   []*v1.Node{node2},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 1, //p11 gets evicted
 		},
 		{
 			description:             "Critical and non critical pods, pods not tolerating node taint should be evicted even if they are critical",
-			pods:                    []v1.Pod{*p2, *p7, *p9, *p10},
-			nodes:                   []*v1.Node{node2},
+			pods:                    []*v1.Pod{p2, p7, p9, p10},
+			nodes:                   []*v1.Node{node1, node2},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: true,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 2, //p2 and p7 are evicted
 		},
 		{
 			description:             "Pod p2 doesn't tolerate taint on it's node, but also doesn't tolerate taints on other nodes",
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1, node2},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 0, //p2 gets evicted
 			nodeFit:                 true,
 		},
 		{
 			description:             "Pod p12 doesn't tolerate taint on it's node, but other nodes don't match it's selector",
-			pods:                    []v1.Pod{*p1, *p3, *p12},
+			pods:                    []*v1.Pod{p1, p3, p12},
 			nodes:                   []*v1.Node{node1, node3},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 0, //p2 gets evicted
 			nodeFit:                 true,
 		},
 		{
 			description:             "Pod p2 doesn't tolerate taint on it's node, but other nodes are unschedulable",
-			pods:                    []v1.Pod{*p1, *p2, *p3},
+			pods:                    []*v1.Pod{p1, p2, p3},
 			nodes:                   []*v1.Node{node1, node4},
 			evictLocalStoragePods:   false,
 			evictSystemCriticalPods: false,
-			maxPodsToEvictPerNode:   0,
 			expectedEvictedPodCount: 0, //p2 gets evicted
 			nodeFit:                 true,
 		},
+		{
+			description:             "Pods not tolerating PreferNoSchedule node taint should not be evicted when not enabled",
+			pods:                    []*v1.Pod{p13},
+			nodes:                   []*v1.Node{node5},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			expectedEvictedPodCount: 0,
+		},
+		{
+			description:             "Pods not tolerating PreferNoSchedule node taint should be evicted when enabled",
+			pods:                    []*v1.Pod{p13},
+			nodes:                   []*v1.Node{node5},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			includePreferNoSchedule: true,
+			expectedEvictedPodCount: 1, // p13 gets evicted
+		},
+		{
+			description:             "Pods not tolerating excluded node taints (by key) should not be evicted",
+			pods:                    []*v1.Pod{p2},
+			nodes:                   []*v1.Node{node1},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			excludedTaints:          []string{"excludedTaint1", "testTaint1"},
+			expectedEvictedPodCount: 0, // nothing gets evicted, as one of the specified excludedTaints matches the key of node1's taint
+		},
+		{
+			description:             "Pods not tolerating excluded node taints (by key and value) should not be evicted",
+			pods:                    []*v1.Pod{p2},
+			nodes:                   []*v1.Node{node1},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			excludedTaints:          []string{"testTaint1=test1"},
+			expectedEvictedPodCount: 0, // nothing gets evicted, as both the key and value of the excluded taint match node1's taint
+		},
+		{
+			description:             "The excluded taint matches the key of node1's taint, but does not match the value",
+			pods:                    []*v1.Pod{p2},
+			nodes:                   []*v1.Node{node1},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			excludedTaints:          []string{"testTaint1=test2"},
+			expectedEvictedPodCount: 1, // pod gets evicted, as excluded taint value does not match node1's taint value
+		},
+		{
+			description:             "Critical and non critical pods, pods not tolerating node taint can't be evicted because the only available node does not have enough resources.",
+			pods:                    []*v1.Pod{p2, p7, p9, p10},
+			nodes:                   []*v1.Node{node1, node6},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: true,
+			expectedEvictedPodCount: 0, //p2 and p7 can't be evicted
+			nodeFit:                 true,
+		},
 	}
 
 	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
 
-		// create fake client
-		fakeClient := &fake.Clientset{}
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			return true, &v1.PodList{Items: tc.pods}, nil
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range tc.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := evictions.NewPodEvictor(
+				fakeClient,
+				policyv1.SchemeGroupVersion.String(),
+				false,
+				tc.maxPodsToEvictPerNode,
+				tc.maxNoOfPodsToEvictPerNamespace,
+				tc.nodes,
+				getPodsAssignedToNode,
+				tc.evictLocalStoragePods,
+				tc.evictSystemCriticalPods,
+				false,
+				false,
+				false,
+			)
+
+			strategy := api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit:                 tc.nodeFit,
+					IncludePreferNoSchedule: tc.includePreferNoSchedule,
+					ExcludedTaints:          tc.excludedTaints,
+				},
+			}
+
+			RemovePodsViolatingNodeTaints(ctx, fakeClient, strategy, tc.nodes, podEvictor, getPodsAssignedToNode)
+			actualEvictedPodCount := podEvictor.TotalEvicted()
+			if actualEvictedPodCount != tc.expectedEvictedPodCount {
+				t.Errorf("Test %#v failed, Unexpected no of pods evicted: pods evicted: %d, expected: %d", tc.description, actualEvictedPodCount, tc.expectedEvictedPodCount)
+			}
 		})
-
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			policyv1.SchemeGroupVersion.String(),
-			false,
-			tc.maxPodsToEvictPerNode,
-			tc.nodes,
-			tc.evictLocalStoragePods,
-			tc.evictSystemCriticalPods,
-			false,
-		)
-
-		strategy := api.DeschedulerStrategy{
-			Params: &api.StrategyParameters{
-				NodeFit: tc.nodeFit,
-			},
-		}
-
-		RemovePodsViolatingNodeTaints(ctx, fakeClient, strategy, tc.nodes, podEvictor)
-		actualEvictedPodCount := podEvictor.TotalEvicted()
-		if actualEvictedPodCount != tc.expectedEvictedPodCount {
-			t.Errorf("Test %#v failed, Unexpected no of pods evicted: pods evicted: %d, expected: %d", tc.description, actualEvictedPodCount, tc.expectedEvictedPodCount)
-		}
 	}
-
 }
 
 func TestToleratesTaint(t *testing.T) {

pkg/descheduler/strategies/nodeutilization/highnodeutilization.go

@@ -24,15 +24,17 @@ import (
 	"k8s.io/apimachinery/pkg/api/resource"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
 	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
 // HighNodeUtilization evicts pods from under utilized nodes so that scheduler can schedule according to its strategy.
 // Note that CPU/Memory requests are used to calculate nodes' utilization and not the actual resource usage.
-func HighNodeUtilization(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func HighNodeUtilization(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	if err := validateNodeUtilizationParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid HighNodeUtilization parameters")
 		return
@@ -61,16 +63,17 @@ func HighNodeUtilization(ctx context.Context, client clientset.Interface, strate
 	resourceNames := getResourceNames(targetThresholds)
 
 	sourceNodes, highNodes := classifyNodes(
-		getNodeUsage(ctx, client, nodes, thresholds, targetThresholds, resourceNames),
-		func(node *v1.Node, usage NodeUsage) bool {
-			return isNodeWithLowUtilization(usage)
+		getNodeUsage(nodes, resourceNames, getPodsAssignedToNode),
+		getNodeThresholds(nodes, thresholds, targetThresholds, resourceNames, getPodsAssignedToNode, false),
+		func(node *v1.Node, usage NodeUsage, threshold NodeThresholds) bool {
+			return isNodeWithLowUtilization(usage, threshold.lowResourceThreshold)
 		},
-		func(node *v1.Node, usage NodeUsage) bool {
+		func(node *v1.Node, usage NodeUsage, threshold NodeThresholds) bool {
 			if nodeutil.IsNodeUnschedulable(node) {
 				klog.V(2).InfoS("Node is unschedulable", "node", klog.KObj(node))
 				return false
 			}
-			return !isNodeWithLowUtilization(usage)
+			return !isNodeWithLowUtilization(usage, threshold.lowResourceThreshold)
 		})
 
 	// log message in one line
@@ -80,7 +83,7 @@ func HighNodeUtilization(ctx context.Context, client clientset.Interface, strate
 		"Pods", thresholds[v1.ResourcePods],
 	}
 	for name := range thresholds {
-		if !isBasicResource(name) {
+		if !nodeutil.IsBasicResource(name) {
 			keysAndValues = append(keysAndValues, string(name), int64(thresholds[name]))
 		}
 	}
@@ -92,8 +95,8 @@ func HighNodeUtilization(ctx context.Context, client clientset.Interface, strate
 		klog.V(1).InfoS("No node is underutilized, nothing to do here, you might tune your thresholds further")
 		return
 	}
-	if len(sourceNodes) < strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes {
-		klog.V(1).InfoS("Number of nodes underutilized is less than NumberOfNodes, nothing to do here", "underutilizedNodes", len(sourceNodes), "numberOfNodes", strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes)
+	if len(sourceNodes) <= strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes {
+		klog.V(1).InfoS("Number of nodes underutilized is less or equal than NumberOfNodes, nothing to do here", "underutilizedNodes", len(sourceNodes), "numberOfNodes", strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes)
 		return
 	}
 	if len(sourceNodes) == len(nodes) {
@@ -108,7 +111,7 @@ func HighNodeUtilization(ctx context.Context, client clientset.Interface, strate
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
 	// stop if the total available usage has dropped to zero - no more pods can be scheduled
-	continueEvictionCond := func(nodeUsage NodeUsage, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool {
+	continueEvictionCond := func(nodeInfo NodeInfo, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool {
 		for name := range totalAvailableUsage {
 			if totalAvailableUsage[name].CmpInt64(0) < 1 {
 				return false
@@ -117,6 +120,10 @@ func HighNodeUtilization(ctx context.Context, client clientset.Interface, strate
 
 		return true
 	}
+
+	// Sort the nodes by the usage in ascending order
+	sortNodesByUsage(sourceNodes, true)
+
 	evictPodsFromSourceNodes(
 		ctx,
 		sourceNodes,
@@ -157,7 +164,7 @@ func setDefaultForThresholds(thresholds, targetThresholds api.ResourceThresholds
 	targetThresholds[v1.ResourceMemory] = MaxResourcePercentage
 
 	for name := range thresholds {
-		if !isBasicResource(name) {
+		if !nodeutil.IsBasicResource(name) {
 			targetThresholds[name] = MaxResourcePercentage
 		}
 	}

pkg/descheduler/strategies/nodeutilization/highnodeutilization_test.go

@@ -19,23 +19,24 @@ package nodeutilization
 import (
 	"context"
 	"fmt"
-	"strings"
 	"testing"
 
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/api/policy/v1beta1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
 	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
 
 func TestHighNodeUtilization(t *testing.T) {
-	ctx := context.Background()
 	n1NodeName := "n1"
 	n2NodeName := "n2"
 	n3NodeName := "n3"
@@ -44,13 +45,12 @@ func TestHighNodeUtilization(t *testing.T) {
 	nodeSelectorValue := "west"
 
 	testCases := []struct {
-		name                  string
-		thresholds            api.ResourceThresholds
-		nodes                 map[string]*v1.Node
-		pods                  map[string]*v1.PodList
-		maxPodsToEvictPerNode int
-		expectedPodsEvicted   int
-		evictedPods           []string
+		name                string
+		thresholds          api.ResourceThresholds
+		nodes               []*v1.Node
+		pods                []*v1.Pod
+		expectedPodsEvicted uint
+		evictedPods         []string
 	}{
 		{
 			name: "no node below threshold usage",
@@ -58,39 +58,26 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  20,
 				v1.ResourcePods: 20,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 4000, 3000, 10, nil),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, nil),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p7", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p8", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p9", 400, 0, n3NodeName, test.SetRSOwnerRef),
-					},
-				},
+			pods: []*v1.Pod{
+				// These won't be evicted.
+				test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				// These won't be evicted.
+				test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				// These won't be evicted.
+				test.BuildTestPod("p7", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p8", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p9", 400, 0, n3NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   0,
+			expectedPodsEvicted: 0,
 		},
 		{
 			name: "no evictable pods",
@@ -98,57 +85,44 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  40,
 				v1.ResourcePods: 40,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 9, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 4000, 3000, 10, nil),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 9, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, nil),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							// A pod with local storage.
-							test.SetNormalOwnerRef(pod)
-							pod.Spec.Volumes = []v1.Volume{
-								{
-									Name: "sample",
-									VolumeSource: v1.VolumeSource{
-										HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
-										EmptyDir: &v1.EmptyDirVolumeSource{
-											SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
-									},
-								},
-							}
-							// A Mirror Pod.
-							pod.Annotations = test.GetMirrorPodAnnotation()
-						}),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							// A Critical Pod.
-							pod.Namespace = "kube-system"
-							priority := utils.SystemCriticalPriority
-							pod.Spec.Priority = &priority
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetDSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p5", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p6", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p7", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p8", 400, 0, n3NodeName, test.SetRSOwnerRef),
-					},
-				},
+			pods: []*v1.Pod{
+				// These won't be evicted.
+				test.BuildTestPod("p1", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					// A pod with local storage.
+					test.SetNormalOwnerRef(pod)
+					pod.Spec.Volumes = []v1.Volume{
+						{
+							Name: "sample",
+							VolumeSource: v1.VolumeSource{
+								HostPath: &v1.HostPathVolumeSource{Path: "somePath"},
+								EmptyDir: &v1.EmptyDirVolumeSource{
+									SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+							},
+						},
+					}
+					// A Mirror Pod.
+					pod.Annotations = test.GetMirrorPodAnnotation()
+				}),
+				test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					// A Critical Pod.
+					pod.Namespace = "kube-system"
+					priority := utils.SystemCriticalPriority
+					pod.Spec.Priority = &priority
+				}),
+				// These won't be evicted.
+				test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetDSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetDSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p7", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p8", 400, 0, n3NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   0,
+			expectedPodsEvicted: 0,
 		},
 		{
 			name: "no node to schedule evicted pods",
@@ -156,34 +130,21 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  20,
 				v1.ResourcePods: 20,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						// These can't be evicted.
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These can't be evicted.
-						*test.BuildTestPod("p2", 400, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p3", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n3NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 0, n3NodeName, test.SetRSOwnerRef),
-					},
-				},
+			pods: []*v1.Pod{
+				// These can't be evicted.
+				test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				// These can't be evicted.
+				test.BuildTestPod("p2", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n3NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n3NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   0,
+			expectedPodsEvicted: 0,
 		},
 		{
 			name: "without priorities",
@@ -191,42 +152,29 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  30,
 				v1.ResourcePods: 30,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						// These won't be evicted.
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							// A Critical Pod.
-							pod.Namespace = "kube-system"
-							priority := utils.SystemCriticalPriority
-							pod.Spec.Priority = &priority
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p7", 400, 0, n3NodeName, test.SetRSOwnerRef),
-					},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				// These won't be evicted.
+				test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					// A Critical Pod.
+					pod.Namespace = "kube-system"
+					priority := utils.SystemCriticalPriority
+					pod.Spec.Priority = &priority
+				}),
+				// These won't be evicted.
+				test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p7", 400, 0, n3NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   2,
-			evictedPods:           []string{"p1", "p7"},
+			expectedPodsEvicted: 2,
+			evictedPods:         []string{"p1", "p7"},
 		},
 		{
 			name: "without priorities stop when resource capacity is depleted",
@@ -234,117 +182,79 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  30,
 				v1.ResourcePods: 30,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 2000, 3000, 10, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 2000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 2000, 3000, 10, test.SetNodeUnschedulable),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 2000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 2000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 2000, 3000, 10, test.SetNodeUnschedulable),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p2", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p6", 400, 0, n3NodeName, test.SetRSOwnerRef),
-					},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				// These won't be evicted.
+				test.BuildTestPod("p2", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 400, 0, n3NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   1,
+			expectedPodsEvicted: 1,
 		},
 		{
 			name: "with priorities",
 			thresholds: api.ResourceThresholds{
 				v1.ResourceCPU: 30,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 2000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 2000, 3000, 10, test.SetNodeUnschedulable),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 2000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 2000, 3000, 10, test.SetNodeUnschedulable),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodPriority(pod, lowPriority)
-						}),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodPriority(pod, highPriority)
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p7", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p8", 400, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p9", 400, 0, n3NodeName, test.SetDSOwnerRef),
-					},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodPriority(pod, lowPriority)
+				}),
+				test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodPriority(pod, highPriority)
+				}),
+				// These won't be evicted.
+				test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p7", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p8", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				// These won't be evicted.
+				test.BuildTestPod("p9", 400, 0, n3NodeName, test.SetDSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   1,
-			evictedPods:           []string{"p1"},
+			expectedPodsEvicted: 1,
+			evictedPods:         []string{"p1"},
 		},
 		{
 			name: "without priorities evicting best-effort pods only",
 			thresholds: api.ResourceThresholds{
 				v1.ResourceCPU: 30,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 3000, 3000, 10, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 3000, 3000, 5, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 3000, 3000, 10, test.SetNodeUnschedulable),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 3000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 3000, 3000, 5, nil),
+				test.BuildTestNode(n3NodeName, 3000, 3000, 10, test.SetNodeUnschedulable),
 			},
 			// All pods are assumed to be burstable (test.BuildTestNode always sets both cpu/memory resource requests to some value)
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.MakeBestEffortPod(pod)
-						}),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.MakeBestEffortPod(pod)
+				}),
+				test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+				}),
+				// These won't be evicted.
+				test.BuildTestPod("p3", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 400, 0, n2NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   1,
-			evictedPods:           []string{"p1"},
+			expectedPodsEvicted: 1,
+			evictedPods:         []string{"p1"},
 		},
 		{
 			name: "with extended resource",
@@ -352,60 +262,44 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:   20,
 				extendedResource: 40,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 10, func(node *v1.Node) {
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, func(node *v1.Node) {
 					test.SetNodeExtendedResource(node, extendedResource, 8)
 				}),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, func(node *v1.Node) {
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, func(node *v1.Node) {
 					test.SetNodeExtendedResource(node, extendedResource, 8)
 				}),
-				n3NodeName: test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 100, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-						*test.BuildTestPod("p2", 100, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-						// These won't be evicted
-						*test.BuildTestPod("p2", 100, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetDSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p3", 500, 0, n2NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-						*test.BuildTestPod("p4", 500, 0, n2NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-						*test.BuildTestPod("p5", 500, 0, n2NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-						*test.BuildTestPod("p6", 500, 0, n2NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 100, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				test.BuildTestPod("p2", 100, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				// These won't be evicted
+				test.BuildTestPod("p3", 500, 0, n2NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				test.BuildTestPod("p4", 500, 0, n2NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				test.BuildTestPod("p5", 500, 0, n2NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				test.BuildTestPod("p6", 500, 0, n2NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   2,
-			evictedPods:           []string{"p1", "p2"},
+			expectedPodsEvicted: 2,
+			evictedPods:         []string{"p1", "p2"},
 		},
 		{
 			name: "with extended resource in some of nodes",
@@ -413,41 +307,29 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:   40,
 				extendedResource: 40,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 10, func(node *v1.Node) {
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, func(node *v1.Node) {
 					test.SetNodeExtendedResource(node, extendedResource, 8)
 				}),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
-				n3NodeName: test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						//These won't be evicted
-						*test.BuildTestPod("p1", 100, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-						*test.BuildTestPod("p2", 100, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p3", 500, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 500, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 500, 0, n2NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p6", 500, 0, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {
-					Items: []v1.Pod{},
-				},
+			pods: []*v1.Pod{
+				//These won't be evicted
+				test.BuildTestPod("p1", 100, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				test.BuildTestPod("p2", 100, 0, n1NodeName, func(pod *v1.Pod) {
+					test.SetRSOwnerRef(pod)
+					test.SetPodExtendedResourceRequest(pod, extendedResource, 1)
+				}),
+				test.BuildTestPod("p3", 500, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 500, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p5", 500, 0, n2NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p6", 500, 0, n2NodeName, test.SetRSOwnerRef),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   0,
+			expectedPodsEvicted: 0,
 		},
 		{
 			name: "Other node match pod node selector",
@@ -455,37 +337,28 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  30,
 				v1.ResourcePods: 30,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 9, func(node *v1.Node) {
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 9, func(node *v1.Node) {
 					node.ObjectMeta.Labels = map[string]string{
 						nodeSelectorKey: nodeSelectorValue,
 					}
 				}),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n1NodeName, test.SetDSOwnerRef),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, func(pod *v1.Pod) {
-							// A pod selecting nodes in the "west" datacenter
-							test.SetRSOwnerRef(pod)
-							pod.Spec.NodeSelector = map[string]string{
-								nodeSelectorKey: nodeSelectorValue,
-							}
-						}),
-					},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n1NodeName, test.SetDSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n2NodeName, func(pod *v1.Pod) {
+					// A pod selecting nodes in the "west" datacenter
+					test.SetRSOwnerRef(pod)
+					pod.Spec.NodeSelector = map[string]string{
+						nodeSelectorKey: nodeSelectorValue,
+					}
+				}),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   1,
+			expectedPodsEvicted: 1,
 		},
 		{
 			name: "Other node does not match pod node selector",
@@ -493,67 +366,100 @@ func TestHighNodeUtilization(t *testing.T) {
 				v1.ResourceCPU:  30,
 				v1.ResourcePods: 30,
 			},
-			nodes: map[string]*v1.Node{
-				n1NodeName: test.BuildTestNode(n1NodeName, 4000, 3000, 9, nil),
-				n2NodeName: test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 9, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
 			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n1NodeName, test.SetDSOwnerRef),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p5", 400, 0, n2NodeName, func(pod *v1.Pod) {
-							// A pod selecting nodes in the "west" datacenter
-							test.SetRSOwnerRef(pod)
-							pod.Spec.NodeSelector = map[string]string{
-								nodeSelectorKey: nodeSelectorValue,
-							}
-						}),
-					},
-				},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 0, n1NodeName, test.SetDSOwnerRef),
+				test.BuildTestPod("p5", 400, 0, n2NodeName, func(pod *v1.Pod) {
+					// A pod selecting nodes in the "west" datacenter
+					test.SetRSOwnerRef(pod)
+					pod.Spec.NodeSelector = map[string]string{
+						nodeSelectorKey: nodeSelectorValue,
+					}
+				}),
 			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   0,
+			expectedPodsEvicted: 0,
+		},
+		{
+			name: "Other node does not have enough Memory",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  30,
+				v1.ResourcePods: 30,
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 200, 9, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+			},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 50, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 400, 50, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 50, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 50, n1NodeName, test.SetDSOwnerRef),
+				test.BuildTestPod("p5", 400, 100, n2NodeName, func(pod *v1.Pod) {
+					// A pod requesting more memory than is available on node1
+					test.SetRSOwnerRef(pod)
+				}),
+			},
+			expectedPodsEvicted: 0,
+		},
+		{
+			name: "Other node does not have enough Memory",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  30,
+				v1.ResourcePods: 30,
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 200, 9, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+			},
+			pods: []*v1.Pod{
+				test.BuildTestPod("p1", 400, 50, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p2", 400, 50, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p3", 400, 50, n1NodeName, test.SetRSOwnerRef),
+				test.BuildTestPod("p4", 400, 50, n1NodeName, test.SetDSOwnerRef),
+				test.BuildTestPod("p5", 400, 100, n2NodeName, func(pod *v1.Pod) {
+					// A pod requesting more memory than is available on node1
+					test.SetRSOwnerRef(pod)
+				}),
+			},
+			expectedPodsEvicted: 0,
 		},
 	}
 
-	for _, test := range testCases {
-		t.Run(test.name, func(t *testing.T) {
-			fakeClient := &fake.Clientset{}
-			fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-				list := action.(core.ListAction)
-				fieldString := list.GetListRestrictions().Fields.String()
-				if strings.Contains(fieldString, n1NodeName) {
-					return true, test.pods[n1NodeName], nil
-				}
-				if strings.Contains(fieldString, n2NodeName) {
-					return true, test.pods[n2NodeName], nil
-				}
-				if strings.Contains(fieldString, n3NodeName) {
-					return true, test.pods[n3NodeName], nil
-				}
-				return true, nil, fmt.Errorf("Failed to list: %v", list)
-			})
-			fakeClient.Fake.AddReactor("get", "nodes", func(action core.Action) (bool, runtime.Object, error) {
-				getAction := action.(core.GetAction)
-				if node, exists := test.nodes[getAction.GetName()]; exists {
-					return true, node, nil
-				}
-				return true, nil, fmt.Errorf("Wrong node: %v", getAction.GetName())
-			})
+	for _, testCase := range testCases {
+		t.Run(testCase.name, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range testCase.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range testCase.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
 			podsForEviction := make(map[string]struct{})
-			for _, pod := range test.evictedPods {
+			for _, pod := range testCase.evictedPods {
 				podsForEviction[pod] = struct{}{}
 			}
 
 			evictionFailed := false
-			if len(test.evictedPods) > 0 {
+			if len(testCase.evictedPods) > 0 {
 				fakeClient.Fake.AddReactor("create", "pods", func(action core.Action) (bool, runtime.Object, error) {
 					getAction := action.(core.CreateAction)
 					obj := getAction.GetObject()
@@ -568,17 +474,42 @@ func TestHighNodeUtilization(t *testing.T) {
 				})
 			}
 
-			var nodes []*v1.Node
-			for _, node := range test.nodes {
-				nodes = append(nodes, node)
-			}
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			//fakeClient := &fake.Clientset{}
+			//fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
+			//	list := action.(core.ListAction)
+			//	fieldString := list.GetListRestrictions().Fields.String()
+			//	if strings.Contains(fieldString, n1NodeName) {
+			//		return true, test.pods[n1NodeName], nil
+			//	}
+			//	if strings.Contains(fieldString, n2NodeName) {
+			//		return true, test.pods[n2NodeName], nil
+			//	}
+			//	if strings.Contains(fieldString, n3NodeName) {
+			//		return true, test.pods[n3NodeName], nil
+			//	}
+			//	return true, nil, fmt.Errorf("Failed to list: %v", list)
+			//})
+			//fakeClient.Fake.AddReactor("get", "nodes", func(action core.Action) (bool, runtime.Object, error) {
+			//	getAction := action.(core.GetAction)
+			//	if node, exists := testCase.nodes[getAction.GetName()]; exists {
+			//		return true, node, nil
+			//	}
+			//	return true, nil, fmt.Errorf("Wrong node: %v", getAction.GetName())
+			//})
 
 			podEvictor := evictions.NewPodEvictor(
 				fakeClient,
 				"v1",
 				false,
-				test.maxPodsToEvictPerNode,
-				nodes,
+				nil,
+				nil,
+				testCase.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
 				false,
 				false,
 				false,
@@ -588,16 +519,16 @@ func TestHighNodeUtilization(t *testing.T) {
 				Enabled: true,
 				Params: &api.StrategyParameters{
 					NodeResourceUtilizationThresholds: &api.NodeResourceUtilizationThresholds{
-						Thresholds: test.thresholds,
+						Thresholds: testCase.thresholds,
 					},
 					NodeFit: true,
 				},
 			}
-			HighNodeUtilization(ctx, fakeClient, strategy, nodes, podEvictor)
+			HighNodeUtilization(ctx, fakeClient, strategy, testCase.nodes, podEvictor, getPodsAssignedToNode)
 
 			podsEvicted := podEvictor.TotalEvicted()
-			if test.expectedPodsEvicted != podsEvicted {
-				t.Errorf("Expected %#v pods to be evicted but %#v got evicted", test.expectedPodsEvicted, podsEvicted)
+			if testCase.expectedPodsEvicted != podsEvicted {
+				t.Errorf("Expected %v pods to be evicted but %v got evicted", testCase.expectedPodsEvicted, podsEvicted)
 			}
 			if evictionFailed {
 				t.Errorf("Pod evictions failed unexpectedly")
@@ -674,7 +605,6 @@ func TestValidateHighNodeUtilizationStrategyConfig(t *testing.T) {
 }
 
 func TestHighNodeUtilizationWithTaints(t *testing.T) {
-	ctx := context.Background()
 	strategy := api.DeschedulerStrategy{
 		Enabled: true,
 		Params: &api.StrategyParameters{
@@ -710,7 +640,7 @@ func TestHighNodeUtilizationWithTaints(t *testing.T) {
 		name              string
 		nodes             []*v1.Node
 		pods              []*v1.Pod
-		evictionsExpected int
+		evictionsExpected uint
 	}{
 		{
 			name:  "No taints",
@@ -752,6 +682,9 @@ func TestHighNodeUtilizationWithTaints(t *testing.T) {
 
 	for _, item := range tests {
 		t.Run(item.name, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
 			var objs []runtime.Object
 			for _, node := range item.nodes {
 				objs = append(objs, node)
@@ -762,19 +695,33 @@ func TestHighNodeUtilizationWithTaints(t *testing.T) {
 			}
 
 			fakeClient := fake.NewSimpleClientset(objs...)
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
 
 			podEvictor := evictions.NewPodEvictor(
 				fakeClient,
 				"policy/v1",
 				false,
-				item.evictionsExpected,
+				&item.evictionsExpected,
+				nil,
 				item.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
 				false,
 				false,
 				false,
 			)
 
-			HighNodeUtilization(ctx, fakeClient, strategy, item.nodes, podEvictor)
+			HighNodeUtilization(ctx, fakeClient, strategy, item.nodes, podEvictor, getPodsAssignedToNode)
 
 			if item.evictionsExpected != podEvictor.TotalEvicted() {
 				t.Errorf("Expected %v evictions, got %v", item.evictionsExpected, podEvictor.TotalEvicted())

pkg/descheduler/strategies/nodeutilization/lownodeutilization.go

@@ -19,6 +19,7 @@ package nodeutilization
 import (
 	"context"
 	"fmt"
+
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	clientset "k8s.io/client-go/kubernetes"
@@ -27,12 +28,13 @@ import (
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
 	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
 // LowNodeUtilization evicts pods from overutilized nodes to underutilized nodes. Note that CPU/Memory requests are used
 // to calculate nodes' utilization and not the actual resource usage.
-func LowNodeUtilization(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func LowNodeUtilization(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	// TODO: May be create a struct for the strategy as well, so that we don't have to pass along the all the params?
 	if err := validateNodeUtilizationParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid LowNodeUtilization parameters")
@@ -48,40 +50,57 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 	if strategy.Params != nil {
 		nodeFit = strategy.Params.NodeFit
 	}
-
+	useDeviationThresholds := strategy.Params.NodeResourceUtilizationThresholds.UseDeviationThresholds
 	thresholds := strategy.Params.NodeResourceUtilizationThresholds.Thresholds
 	targetThresholds := strategy.Params.NodeResourceUtilizationThresholds.TargetThresholds
-	if err := validateLowUtilizationStrategyConfig(thresholds, targetThresholds); err != nil {
+	if err := validateLowUtilizationStrategyConfig(thresholds, targetThresholds, useDeviationThresholds); err != nil {
 		klog.ErrorS(err, "LowNodeUtilization config is not valid")
 		return
 	}
+
 	// check if Pods/CPU/Mem are set, if not, set them to 100
 	if _, ok := thresholds[v1.ResourcePods]; !ok {
-		thresholds[v1.ResourcePods] = MaxResourcePercentage
-		targetThresholds[v1.ResourcePods] = MaxResourcePercentage
+		if useDeviationThresholds {
+			thresholds[v1.ResourcePods] = MinResourcePercentage
+			targetThresholds[v1.ResourcePods] = MinResourcePercentage
+		} else {
+			thresholds[v1.ResourcePods] = MaxResourcePercentage
+			targetThresholds[v1.ResourcePods] = MaxResourcePercentage
+		}
 	}
 	if _, ok := thresholds[v1.ResourceCPU]; !ok {
-		thresholds[v1.ResourceCPU] = MaxResourcePercentage
-		targetThresholds[v1.ResourceCPU] = MaxResourcePercentage
+		if useDeviationThresholds {
+			thresholds[v1.ResourceCPU] = MinResourcePercentage
+			targetThresholds[v1.ResourceCPU] = MinResourcePercentage
+		} else {
+			thresholds[v1.ResourceCPU] = MaxResourcePercentage
+			targetThresholds[v1.ResourceCPU] = MaxResourcePercentage
+		}
 	}
 	if _, ok := thresholds[v1.ResourceMemory]; !ok {
-		thresholds[v1.ResourceMemory] = MaxResourcePercentage
-		targetThresholds[v1.ResourceMemory] = MaxResourcePercentage
+		if useDeviationThresholds {
+			thresholds[v1.ResourceMemory] = MinResourcePercentage
+			targetThresholds[v1.ResourceMemory] = MinResourcePercentage
+		} else {
+			thresholds[v1.ResourceMemory] = MaxResourcePercentage
+			targetThresholds[v1.ResourceMemory] = MaxResourcePercentage
+		}
 	}
 	resourceNames := getResourceNames(thresholds)
 
 	lowNodes, sourceNodes := classifyNodes(
-		getNodeUsage(ctx, client, nodes, thresholds, targetThresholds, resourceNames),
+		getNodeUsage(nodes, resourceNames, getPodsAssignedToNode),
+		getNodeThresholds(nodes, thresholds, targetThresholds, resourceNames, getPodsAssignedToNode, useDeviationThresholds),
 		// The node has to be schedulable (to be able to move workload there)
-		func(node *v1.Node, usage NodeUsage) bool {
+		func(node *v1.Node, usage NodeUsage, threshold NodeThresholds) bool {
 			if nodeutil.IsNodeUnschedulable(node) {
 				klog.V(2).InfoS("Node is unschedulable, thus not considered as underutilized", "node", klog.KObj(node))
 				return false
 			}
-			return isNodeWithLowUtilization(usage)
+			return isNodeWithLowUtilization(usage, threshold.lowResourceThreshold)
 		},
-		func(node *v1.Node, usage NodeUsage) bool {
-			return isNodeAboveTargetUtilization(usage)
+		func(node *v1.Node, usage NodeUsage, threshold NodeThresholds) bool {
+			return isNodeAboveTargetUtilization(usage, threshold.highResourceThreshold)
 		},
 	)
 
@@ -92,7 +111,7 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 		"Pods", thresholds[v1.ResourcePods],
 	}
 	for name := range thresholds {
-		if !isBasicResource(name) {
+		if !nodeutil.IsBasicResource(name) {
 			keysAndValues = append(keysAndValues, string(name), int64(thresholds[name]))
 		}
 	}
@@ -106,7 +125,7 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 		"Pods", targetThresholds[v1.ResourcePods],
 	}
 	for name := range targetThresholds {
-		if !isBasicResource(name) {
+		if !nodeutil.IsBasicResource(name) {
 			keysAndValues = append(keysAndValues, string(name), int64(targetThresholds[name]))
 		}
 	}
@@ -118,8 +137,8 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 		return
 	}
 
-	if len(lowNodes) < strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes {
-		klog.V(1).InfoS("Number of nodes underutilized is less than NumberOfNodes, nothing to do here", "underutilizedNodes", len(lowNodes), "numberOfNodes", strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes)
+	if len(lowNodes) <= strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes {
+		klog.V(1).InfoS("Number of nodes underutilized is less or equal than NumberOfNodes, nothing to do here", "underutilizedNodes", len(lowNodes), "numberOfNodes", strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes)
 		return
 	}
 
@@ -136,8 +155,8 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
 	// stop if node utilization drops below target threshold or any of required capacity (cpu, memory, pods) is moved
-	continueEvictionCond := func(nodeUsage NodeUsage, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool {
-		if !isNodeAboveTargetUtilization(nodeUsage) {
+	continueEvictionCond := func(nodeInfo NodeInfo, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool {
+		if !isNodeAboveTargetUtilization(nodeInfo.NodeUsage, nodeInfo.thresholds.highResourceThreshold) {
 			return false
 		}
 		for name := range totalAvailableUsage {
@@ -149,6 +168,9 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 		return true
 	}
 
+	// Sort the nodes by the usage in descending order
+	sortNodesByUsage(sourceNodes, false)
+
 	evictPodsFromSourceNodes(
 		ctx,
 		sourceNodes,
@@ -163,7 +185,7 @@ func LowNodeUtilization(ctx context.Context, client clientset.Interface, strateg
 }
 
 // validateLowUtilizationStrategyConfig checks if the strategy's config is valid
-func validateLowUtilizationStrategyConfig(thresholds, targetThresholds api.ResourceThresholds) error {
+func validateLowUtilizationStrategyConfig(thresholds, targetThresholds api.ResourceThresholds, useDeviationThresholds bool) error {
 	// validate thresholds and targetThresholds config
 	if err := validateThresholds(thresholds); err != nil {
 		return fmt.Errorf("thresholds config is not valid: %v", err)
@@ -179,7 +201,7 @@ func validateLowUtilizationStrategyConfig(thresholds, targetThresholds api.Resou
 	for resourceName, value := range thresholds {
 		if targetValue, ok := targetThresholds[resourceName]; !ok {
 			return fmt.Errorf("thresholds and targetThresholds configured different resources")
-		} else if value > targetValue {
+		} else if value > targetValue && !useDeviationThresholds {
 			return fmt.Errorf("thresholds' %v percentage is greater than targetThresholds'", resourceName)
 		}
 	}

pkg/descheduler/strategies/nodeutilization/nodeutilization.go

@@ -19,15 +19,18 @@ package nodeutilization
 import (
 	"context"
 	"fmt"
+	"sort"
+
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
-	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	"sigs.k8s.io/descheduler/pkg/descheduler/node"
+	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
-	"sort"
 )
 
 // NodeUsage stores a node's info, pods on it, thresholds and its resource usage
@@ -35,12 +38,19 @@ type NodeUsage struct {
 	node    *v1.Node
 	usage   map[v1.ResourceName]*resource.Quantity
 	allPods []*v1.Pod
+}
 
+type NodeThresholds struct {
 	lowResourceThreshold  map[v1.ResourceName]*resource.Quantity
 	highResourceThreshold map[v1.ResourceName]*resource.Quantity
 }
 
-type continueEvictionCond func(nodeUsage NodeUsage, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool
+type NodeInfo struct {
+	NodeUsage
+	thresholds NodeThresholds
+}
+
+type continueEvictionCond func(nodeInfo NodeInfo, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool
 
 // NodePodsMap is a set of (node, pods) pairs
 type NodePodsMap map[*v1.Node][]*v1.Pod
@@ -65,7 +75,7 @@ func validateNodeUtilizationParams(params *api.StrategyParameters) error {
 
 // validateThresholds checks if thresholds have valid resource name and resource percentage configured
 func validateThresholds(thresholds api.ResourceThresholds) error {
-	if thresholds == nil || len(thresholds) == 0 {
+	if len(thresholds) == 0 {
 		return fmt.Errorf("no resource threshold is configured")
 	}
 	for name, percent := range thresholds {
@@ -76,64 +86,106 @@ func validateThresholds(thresholds api.ResourceThresholds) error {
 	return nil
 }
 
-func getNodeUsage(
-	ctx context.Context,
-	client clientset.Interface,
+func normalizePercentage(percent api.Percentage) api.Percentage {
+	if percent > MaxResourcePercentage {
+		return MaxResourcePercentage
+	}
+	if percent < MinResourcePercentage {
+		return MinResourcePercentage
+	}
+	return percent
+}
+
+func getNodeThresholds(
 	nodes []*v1.Node,
 	lowThreshold, highThreshold api.ResourceThresholds,
 	resourceNames []v1.ResourceName,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
+	useDeviationThresholds bool,
+) map[string]NodeThresholds {
+	nodeThresholdsMap := map[string]NodeThresholds{}
+
+	averageResourceUsagePercent := api.ResourceThresholds{}
+	if useDeviationThresholds {
+		averageResourceUsagePercent = averageNodeBasicresources(nodes, getPodsAssignedToNode, resourceNames)
+	}
+
+	for _, node := range nodes {
+		nodeCapacity := node.Status.Capacity
+		if len(node.Status.Allocatable) > 0 {
+			nodeCapacity = node.Status.Allocatable
+		}
+
+		nodeThresholdsMap[node.Name] = NodeThresholds{
+			lowResourceThreshold:  map[v1.ResourceName]*resource.Quantity{},
+			highResourceThreshold: map[v1.ResourceName]*resource.Quantity{},
+		}
+
+		for _, resourceName := range resourceNames {
+			if useDeviationThresholds {
+				cap := nodeCapacity[resourceName]
+				if lowThreshold[resourceName] == MinResourcePercentage {
+					nodeThresholdsMap[node.Name].lowResourceThreshold[resourceName] = &cap
+					nodeThresholdsMap[node.Name].highResourceThreshold[resourceName] = &cap
+				} else {
+					nodeThresholdsMap[node.Name].lowResourceThreshold[resourceName] = resourceThreshold(nodeCapacity, resourceName, normalizePercentage(averageResourceUsagePercent[resourceName]-lowThreshold[resourceName]))
+					nodeThresholdsMap[node.Name].highResourceThreshold[resourceName] = resourceThreshold(nodeCapacity, resourceName, normalizePercentage(averageResourceUsagePercent[resourceName]+highThreshold[resourceName]))
+				}
+			} else {
+				nodeThresholdsMap[node.Name].lowResourceThreshold[resourceName] = resourceThreshold(nodeCapacity, resourceName, lowThreshold[resourceName])
+				nodeThresholdsMap[node.Name].highResourceThreshold[resourceName] = resourceThreshold(nodeCapacity, resourceName, highThreshold[resourceName])
+			}
+		}
+
+	}
+	return nodeThresholdsMap
+}
+
+func getNodeUsage(
+	nodes []*v1.Node,
+	resourceNames []v1.ResourceName,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 ) []NodeUsage {
 	var nodeUsageList []NodeUsage
 
 	for _, node := range nodes {
-		pods, err := podutil.ListPodsOnANode(ctx, client, node)
+		pods, err := podutil.ListPodsOnANode(node.Name, getPodsAssignedToNode, nil)
 		if err != nil {
 			klog.V(2).InfoS("Node will not be processed, error accessing its pods", "node", klog.KObj(node), "err", err)
 			continue
 		}
 
-		// A threshold is in percentages but in <0;100> interval.
-		// Performing `threshold * 0.01` will convert <0;100> interval into <0;1>.
-		// Multiplying it with capacity will give fraction of the capacity corresponding to the given high/low resource threshold in Quantity units.
-		nodeCapacity := node.Status.Capacity
-		if len(node.Status.Allocatable) > 0 {
-			nodeCapacity = node.Status.Allocatable
-		}
-		lowResourceThreshold := map[v1.ResourceName]*resource.Quantity{
-			v1.ResourceCPU:    resource.NewMilliQuantity(int64(float64(lowThreshold[v1.ResourceCPU])*float64(nodeCapacity.Cpu().MilliValue())*0.01), resource.DecimalSI),
-			v1.ResourceMemory: resource.NewQuantity(int64(float64(lowThreshold[v1.ResourceMemory])*float64(nodeCapacity.Memory().Value())*0.01), resource.BinarySI),
-			v1.ResourcePods:   resource.NewQuantity(int64(float64(lowThreshold[v1.ResourcePods])*float64(nodeCapacity.Pods().Value())*0.01), resource.DecimalSI),
-		}
-		for _, name := range resourceNames {
-			if !isBasicResource(name) {
-				cap := nodeCapacity[name]
-				lowResourceThreshold[name] = resource.NewQuantity(int64(float64(lowThreshold[name])*float64(cap.Value())*0.01), resource.DecimalSI)
-			}
-		}
-		highResourceThreshold := map[v1.ResourceName]*resource.Quantity{
-			v1.ResourceCPU:    resource.NewMilliQuantity(int64(float64(highThreshold[v1.ResourceCPU])*float64(nodeCapacity.Cpu().MilliValue())*0.01), resource.DecimalSI),
-			v1.ResourceMemory: resource.NewQuantity(int64(float64(highThreshold[v1.ResourceMemory])*float64(nodeCapacity.Memory().Value())*0.01), resource.BinarySI),
-			v1.ResourcePods:   resource.NewQuantity(int64(float64(highThreshold[v1.ResourcePods])*float64(nodeCapacity.Pods().Value())*0.01), resource.DecimalSI),
-		}
-		for _, name := range resourceNames {
-			if !isBasicResource(name) {
-				cap := nodeCapacity[name]
-				highResourceThreshold[name] = resource.NewQuantity(int64(float64(highThreshold[name])*float64(cap.Value())*0.01), resource.DecimalSI)
-			}
-		}
-
 		nodeUsageList = append(nodeUsageList, NodeUsage{
-			node:                  node,
-			usage:                 nodeUtilization(node, pods, resourceNames),
-			allPods:               pods,
-			lowResourceThreshold:  lowResourceThreshold,
-			highResourceThreshold: highResourceThreshold,
+			node:    node,
+			usage:   nodeutil.NodeUtilization(pods, resourceNames),
+			allPods: pods,
 		})
 	}
 
 	return nodeUsageList
 }
 
+func resourceThreshold(nodeCapacity v1.ResourceList, resourceName v1.ResourceName, threshold api.Percentage) *resource.Quantity {
+	defaultFormat := resource.DecimalSI
+	if resourceName == v1.ResourceMemory {
+		defaultFormat = resource.BinarySI
+	}
+
+	resourceCapacityFraction := func(resourceNodeCapacity int64) int64 {
+		// A threshold is in percentages but in <0;100> interval.
+		// Performing `threshold * 0.01` will convert <0;100> interval into <0;1>.
+		// Multiplying it with capacity will give fraction of the capacity corresponding to the given resource threshold in Quantity units.
+		return int64(float64(threshold) * 0.01 * float64(resourceNodeCapacity))
+	}
+
+	resourceCapacityQuantity := nodeCapacity.Name(resourceName, defaultFormat)
+
+	if resourceName == v1.ResourceCPU {
+		return resource.NewMilliQuantity(resourceCapacityFraction(resourceCapacityQuantity.MilliValue()), defaultFormat)
+	}
+	return resource.NewQuantity(resourceCapacityFraction(resourceCapacityQuantity.Value()), defaultFormat)
+}
+
 func resourceUsagePercentages(nodeUsage NodeUsage) map[v1.ResourceName]float64 {
 	nodeCapacity := nodeUsage.node.Status.Capacity
 	if len(nodeUsage.node.Status.Allocatable) > 0 {
@@ -155,19 +207,24 @@ func resourceUsagePercentages(nodeUsage NodeUsage) map[v1.ResourceName]float64 {
 // low and high thresholds, it is simply ignored.
 func classifyNodes(
 	nodeUsages []NodeUsage,
-	lowThresholdFilter, highThresholdFilter func(node *v1.Node, usage NodeUsage) bool,
-) ([]NodeUsage, []NodeUsage) {
-	lowNodes, highNodes := []NodeUsage{}, []NodeUsage{}
+	nodeThresholds map[string]NodeThresholds,
+	lowThresholdFilter, highThresholdFilter func(node *v1.Node, usage NodeUsage, threshold NodeThresholds) bool,
+) ([]NodeInfo, []NodeInfo) {
+	lowNodes, highNodes := []NodeInfo{}, []NodeInfo{}
 
 	for _, nodeUsage := range nodeUsages {
-		if lowThresholdFilter(nodeUsage.node, nodeUsage) {
-			klog.V(2).InfoS("Node is underutilized", "node", klog.KObj(nodeUsage.node), "usage", nodeUsage.usage, "usagePercentage", resourceUsagePercentages(nodeUsage))
-			lowNodes = append(lowNodes, nodeUsage)
-		} else if highThresholdFilter(nodeUsage.node, nodeUsage) {
-			klog.V(2).InfoS("Node is overutilized", "node", klog.KObj(nodeUsage.node), "usage", nodeUsage.usage, "usagePercentage", resourceUsagePercentages(nodeUsage))
-			highNodes = append(highNodes, nodeUsage)
+		nodeInfo := NodeInfo{
+			NodeUsage:  nodeUsage,
+			thresholds: nodeThresholds[nodeUsage.node.Name],
+		}
+		if lowThresholdFilter(nodeUsage.node, nodeUsage, nodeThresholds[nodeUsage.node.Name]) {
+			klog.InfoS("Node is underutilized", "node", klog.KObj(nodeUsage.node), "usage", nodeUsage.usage, "usagePercentage", resourceUsagePercentages(nodeUsage))
+			lowNodes = append(lowNodes, nodeInfo)
+		} else if highThresholdFilter(nodeUsage.node, nodeUsage, nodeThresholds[nodeUsage.node.Name]) {
+			klog.InfoS("Node is overutilized", "node", klog.KObj(nodeUsage.node), "usage", nodeUsage.usage, "usagePercentage", resourceUsagePercentages(nodeUsage))
+			highNodes = append(highNodes, nodeInfo)
 		} else {
-			klog.V(2).InfoS("Node is appropriately utilized", "node", klog.KObj(nodeUsage.node), "usage", nodeUsage.usage, "usagePercentage", resourceUsagePercentages(nodeUsage))
+			klog.InfoS("Node is appropriately utilized", "node", klog.KObj(nodeUsage.node), "usage", nodeUsage.usage, "usagePercentage", resourceUsagePercentages(nodeUsage))
 		}
 	}
 
@@ -179,16 +236,13 @@ func classifyNodes(
 // TODO: @ravig Break this function into smaller functions.
 func evictPodsFromSourceNodes(
 	ctx context.Context,
-	sourceNodes, destinationNodes []NodeUsage,
+	sourceNodes, destinationNodes []NodeInfo,
 	podEvictor *evictions.PodEvictor,
 	podFilter func(pod *v1.Pod) bool,
 	resourceNames []v1.ResourceName,
 	strategy string,
 	continueEviction continueEvictionCond,
 ) {
-
-	sortNodesByUsage(sourceNodes)
-
 	// upper bound on total number of pods/cpu/memory and optional extended resources to be moved
 	totalAvailableUsage := map[v1.ResourceName]*resource.Quantity{
 		v1.ResourcePods:   {},
@@ -204,7 +258,7 @@ func evictPodsFromSourceNodes(
 			if _, ok := totalAvailableUsage[name]; !ok {
 				totalAvailableUsage[name] = resource.NewQuantity(0, resource.DecimalSI)
 			}
-			totalAvailableUsage[name].Add(*node.highResourceThreshold[name])
+			totalAvailableUsage[name].Add(*node.thresholds.highResourceThreshold[name])
 			totalAvailableUsage[name].Sub(*node.usage[name])
 		}
 	}
@@ -216,7 +270,7 @@ func evictPodsFromSourceNodes(
 		"Pods", totalAvailableUsage[v1.ResourcePods].Value(),
 	}
 	for name := range totalAvailableUsage {
-		if !isBasicResource(name) {
+		if !node.IsBasicResource(name) {
 			keysAndValues = append(keysAndValues, string(name), totalAvailableUsage[name].Value())
 		}
 	}
@@ -244,7 +298,7 @@ func evictPodsFromSourceNodes(
 func evictPods(
 	ctx context.Context,
 	inputPods []*v1.Pod,
-	nodeUsage NodeUsage,
+	nodeInfo NodeInfo,
 	totalAvailableUsage map[v1.ResourceName]*resource.Quantity,
 	taintsOfLowNodes map[string][]v1.Taint,
 	podEvictor *evictions.PodEvictor,
@@ -252,14 +306,14 @@ func evictPods(
 	continueEviction continueEvictionCond,
 ) {
 
-	if continueEviction(nodeUsage, totalAvailableUsage) {
+	if continueEviction(nodeInfo, totalAvailableUsage) {
 		for _, pod := range inputPods {
 			if !utils.PodToleratesTaints(pod, taintsOfLowNodes) {
 				klog.V(3).InfoS("Skipping eviction for pod, doesn't tolerate node taint", "pod", klog.KObj(pod))
 				continue
 			}
 
-			success, err := podEvictor.EvictPod(ctx, pod, nodeUsage.node, strategy)
+			success, err := podEvictor.EvictPod(ctx, pod, nodeInfo.node, strategy)
 			if err != nil {
 				klog.ErrorS(err, "Error evicting pod", "pod", klog.KObj(pod))
 				break
@@ -270,30 +324,30 @@ func evictPods(
 
 				for name := range totalAvailableUsage {
 					if name == v1.ResourcePods {
-						nodeUsage.usage[name].Sub(*resource.NewQuantity(1, resource.DecimalSI))
+						nodeInfo.usage[name].Sub(*resource.NewQuantity(1, resource.DecimalSI))
 						totalAvailableUsage[name].Sub(*resource.NewQuantity(1, resource.DecimalSI))
 					} else {
 						quantity := utils.GetResourceRequestQuantity(pod, name)
-						nodeUsage.usage[name].Sub(quantity)
+						nodeInfo.usage[name].Sub(quantity)
 						totalAvailableUsage[name].Sub(quantity)
 					}
 				}
 
 				keysAndValues := []interface{}{
-					"node", nodeUsage.node.Name,
-					"CPU", nodeUsage.usage[v1.ResourceCPU].MilliValue(),
-					"Mem", nodeUsage.usage[v1.ResourceMemory].Value(),
-					"Pods", nodeUsage.usage[v1.ResourcePods].Value(),
+					"node", nodeInfo.node.Name,
+					"CPU", nodeInfo.usage[v1.ResourceCPU].MilliValue(),
+					"Mem", nodeInfo.usage[v1.ResourceMemory].Value(),
+					"Pods", nodeInfo.usage[v1.ResourcePods].Value(),
 				}
 				for name := range totalAvailableUsage {
-					if !isBasicResource(name) {
+					if !nodeutil.IsBasicResource(name) {
 						keysAndValues = append(keysAndValues, string(name), totalAvailableUsage[name].Value())
 					}
 				}
 
 				klog.V(3).InfoS("Updated node usage", keysAndValues...)
 				// check if pods can be still evicted
-				if !continueEviction(nodeUsage, totalAvailableUsage) {
+				if !continueEviction(nodeInfo, totalAvailableUsage) {
 					break
 				}
 			}
@@ -301,31 +355,36 @@ func evictPods(
 	}
 }
 
-// sortNodesByUsage sorts nodes based on usage in descending order
-func sortNodesByUsage(nodes []NodeUsage) {
+// sortNodesByUsage sorts nodes based on usage according to the given strategy.
+func sortNodesByUsage(nodes []NodeInfo, ascending bool) {
 	sort.Slice(nodes, func(i, j int) bool {
 		ti := nodes[i].usage[v1.ResourceMemory].Value() + nodes[i].usage[v1.ResourceCPU].MilliValue() + nodes[i].usage[v1.ResourcePods].Value()
 		tj := nodes[j].usage[v1.ResourceMemory].Value() + nodes[j].usage[v1.ResourceCPU].MilliValue() + nodes[j].usage[v1.ResourcePods].Value()
 
 		// extended resources
 		for name := range nodes[i].usage {
-			if !isBasicResource(name) {
+			if !nodeutil.IsBasicResource(name) {
 				ti = ti + nodes[i].usage[name].Value()
 				tj = tj + nodes[j].usage[name].Value()
 			}
 		}
 
-		// To return sorted in descending order
+		// Return ascending order for HighNodeUtilization strategy
+		if ascending {
+			return ti < tj
+		}
+
+		// Return descending order for LowNodeUtilization strategy
 		return ti > tj
 	})
 }
 
 // isNodeAboveTargetUtilization checks if a node is overutilized
 // At least one resource has to be above the high threshold
-func isNodeAboveTargetUtilization(usage NodeUsage) bool {
+func isNodeAboveTargetUtilization(usage NodeUsage, threshold map[v1.ResourceName]*resource.Quantity) bool {
 	for name, nodeValue := range usage.usage {
 		// usage.highResourceThreshold[name] < nodeValue
-		if usage.highResourceThreshold[name].Cmp(*nodeValue) == -1 {
+		if threshold[name].Cmp(*nodeValue) == -1 {
 			return true
 		}
 	}
@@ -334,10 +393,10 @@ func isNodeAboveTargetUtilization(usage NodeUsage) bool {
 
 // isNodeWithLowUtilization checks if a node is underutilized
 // All resources have to be below the low threshold
-func isNodeWithLowUtilization(usage NodeUsage) bool {
+func isNodeWithLowUtilization(usage NodeUsage, threshold map[v1.ResourceName]*resource.Quantity) bool {
 	for name, nodeValue := range usage.usage {
 		// usage.lowResourceThreshold[name] < nodeValue
-		if usage.lowResourceThreshold[name].Cmp(*nodeValue) == -1 {
+		if threshold[name].Cmp(*nodeValue) == -1 {
 			return false
 		}
 	}
@@ -354,43 +413,6 @@ func getResourceNames(thresholds api.ResourceThresholds) []v1.ResourceName {
 	return resourceNames
 }
 
-// isBasicResource checks if resource is basic native.
-func isBasicResource(name v1.ResourceName) bool {
-	switch name {
-	case v1.ResourceCPU, v1.ResourceMemory, v1.ResourcePods:
-		return true
-	default:
-		return false
-	}
-}
-
-func nodeUtilization(node *v1.Node, pods []*v1.Pod, resourceNames []v1.ResourceName) map[v1.ResourceName]*resource.Quantity {
-	totalReqs := map[v1.ResourceName]*resource.Quantity{
-		v1.ResourceCPU:    resource.NewMilliQuantity(0, resource.DecimalSI),
-		v1.ResourceMemory: resource.NewQuantity(0, resource.BinarySI),
-		v1.ResourcePods:   resource.NewQuantity(int64(len(pods)), resource.DecimalSI),
-	}
-	for _, name := range resourceNames {
-		if !isBasicResource(name) {
-			totalReqs[name] = resource.NewQuantity(0, resource.DecimalSI)
-		}
-	}
-
-	for _, pod := range pods {
-		req, _ := utils.PodRequestsAndLimits(pod)
-		for _, name := range resourceNames {
-			quantity, ok := req[name]
-			if ok && name != v1.ResourcePods {
-				// As Quantity.Add says: Add adds the provided y quantity to the current value. If the current value is zero,
-				// the format of the quantity will be updated to the format of y.
-				totalReqs[name].Add(quantity)
-			}
-		}
-	}
-
-	return totalReqs
-}
-
 func classifyPods(pods []*v1.Pod, filter func(pod *v1.Pod) bool) ([]*v1.Pod, []*v1.Pod) {
 	var nonRemovablePods, removablePods []*v1.Pod
 
@@ -404,3 +426,34 @@ func classifyPods(pods []*v1.Pod, filter func(pod *v1.Pod) bool) ([]*v1.Pod, []*
 
 	return nonRemovablePods, removablePods
 }
+
+func averageNodeBasicresources(nodes []*v1.Node, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc, resourceNames []v1.ResourceName) api.ResourceThresholds {
+	total := api.ResourceThresholds{}
+	average := api.ResourceThresholds{}
+	numberOfNodes := len(nodes)
+	for _, node := range nodes {
+		pods, err := podutil.ListPodsOnANode(node.Name, getPodsAssignedToNode, nil)
+		if err != nil {
+			numberOfNodes--
+			continue
+		}
+		usage := nodeutil.NodeUtilization(pods, resourceNames)
+		nodeCapacity := node.Status.Capacity
+		if len(node.Status.Allocatable) > 0 {
+			nodeCapacity = node.Status.Allocatable
+		}
+		for resource, value := range usage {
+			nodeCapacityValue := nodeCapacity[resource]
+			if resource == v1.ResourceCPU {
+				total[resource] += api.Percentage(value.MilliValue()) / api.Percentage(nodeCapacityValue.MilliValue()) * 100.0
+			} else {
+				total[resource] += api.Percentage(value.Value()) / api.Percentage(nodeCapacityValue.Value()) * 100.0
+
+			}
+		}
+	}
+	for resource, value := range total {
+		average[resource] = value / api.Percentage(numberOfNodes)
+	}
+	return average
+}

pkg/descheduler/strategies/nodeutilization/nodeutilization_test.go

@@ -18,17 +18,91 @@ package nodeutilization
 
 import (
 	"fmt"
+	"math"
+	"testing"
+
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
-	"math"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"sigs.k8s.io/descheduler/pkg/api"
-	"testing"
 )
 
 var (
 	lowPriority      = int32(0)
 	highPriority     = int32(10000)
 	extendedResource = v1.ResourceName("example.com/foo")
+	testNode1        = NodeInfo{
+		NodeUsage: NodeUsage{
+			node: &v1.Node{
+				Status: v1.NodeStatus{
+					Capacity: v1.ResourceList{
+						v1.ResourceCPU:    *resource.NewMilliQuantity(2000, resource.DecimalSI),
+						v1.ResourceMemory: *resource.NewQuantity(3977868*1024, resource.BinarySI),
+						v1.ResourcePods:   *resource.NewQuantity(29, resource.BinarySI),
+					},
+					Allocatable: v1.ResourceList{
+						v1.ResourceCPU:    *resource.NewMilliQuantity(1930, resource.DecimalSI),
+						v1.ResourceMemory: *resource.NewQuantity(3287692*1024, resource.BinarySI),
+						v1.ResourcePods:   *resource.NewQuantity(29, resource.BinarySI),
+					},
+				},
+				ObjectMeta: metav1.ObjectMeta{Name: "node1"},
+			},
+			usage: map[v1.ResourceName]*resource.Quantity{
+				v1.ResourceCPU:    resource.NewMilliQuantity(1730, resource.DecimalSI),
+				v1.ResourceMemory: resource.NewQuantity(3038982964, resource.BinarySI),
+				v1.ResourcePods:   resource.NewQuantity(25, resource.BinarySI),
+			},
+		},
+	}
+	testNode2 = NodeInfo{
+		NodeUsage: NodeUsage{
+			node: &v1.Node{
+				Status: v1.NodeStatus{
+					Capacity: v1.ResourceList{
+						v1.ResourceCPU:    *resource.NewMilliQuantity(2000, resource.DecimalSI),
+						v1.ResourceMemory: *resource.NewQuantity(3977868*1024, resource.BinarySI),
+						v1.ResourcePods:   *resource.NewQuantity(29, resource.BinarySI),
+					},
+					Allocatable: v1.ResourceList{
+						v1.ResourceCPU:    *resource.NewMilliQuantity(1930, resource.DecimalSI),
+						v1.ResourceMemory: *resource.NewQuantity(3287692*1024, resource.BinarySI),
+						v1.ResourcePods:   *resource.NewQuantity(29, resource.BinarySI),
+					},
+				},
+				ObjectMeta: metav1.ObjectMeta{Name: "node2"},
+			},
+			usage: map[v1.ResourceName]*resource.Quantity{
+				v1.ResourceCPU:    resource.NewMilliQuantity(1220, resource.DecimalSI),
+				v1.ResourceMemory: resource.NewQuantity(3038982964, resource.BinarySI),
+				v1.ResourcePods:   resource.NewQuantity(11, resource.BinarySI),
+			},
+		},
+	}
+	testNode3 = NodeInfo{
+		NodeUsage: NodeUsage{
+			node: &v1.Node{
+				Status: v1.NodeStatus{
+					Capacity: v1.ResourceList{
+						v1.ResourceCPU:    *resource.NewMilliQuantity(2000, resource.DecimalSI),
+						v1.ResourceMemory: *resource.NewQuantity(3977868*1024, resource.BinarySI),
+						v1.ResourcePods:   *resource.NewQuantity(29, resource.BinarySI),
+					},
+					Allocatable: v1.ResourceList{
+						v1.ResourceCPU:    *resource.NewMilliQuantity(1930, resource.DecimalSI),
+						v1.ResourceMemory: *resource.NewQuantity(3287692*1024, resource.BinarySI),
+						v1.ResourcePods:   *resource.NewQuantity(29, resource.BinarySI),
+					},
+				},
+				ObjectMeta: metav1.ObjectMeta{Name: "node3"},
+			},
+			usage: map[v1.ResourceName]*resource.Quantity{
+				v1.ResourceCPU:    resource.NewMilliQuantity(1530, resource.DecimalSI),
+				v1.ResourceMemory: resource.NewQuantity(5038982964, resource.BinarySI),
+				v1.ResourcePods:   resource.NewQuantity(20, resource.BinarySI),
+			},
+		},
+	}
 )
 
 func TestValidateThresholds(t *testing.T) {
@@ -156,3 +230,27 @@ func TestResourceUsagePercentages(t *testing.T) {
 
 	t.Logf("resourceUsagePercentage: %#v\n", resourceUsagePercentage)
 }
+
+func TestSortNodesByUsageDescendingOrder(t *testing.T) {
+	nodeList := []NodeInfo{testNode1, testNode2, testNode3}
+	expectedNodeList := []NodeInfo{testNode3, testNode1, testNode2} // testNode3 has the highest usage
+	sortNodesByUsage(nodeList, false)                               // ascending=false, sort nodes in descending order
+
+	for i := 0; i < len(expectedNodeList); i++ {
+		if nodeList[i].NodeUsage.node.Name != expectedNodeList[i].NodeUsage.node.Name {
+			t.Errorf("Expected %v, got %v", expectedNodeList[i].NodeUsage.node.Name, nodeList[i].NodeUsage.node.Name)
+		}
+	}
+}
+
+func TestSortNodesByUsageAscendingOrder(t *testing.T) {
+	nodeList := []NodeInfo{testNode1, testNode2, testNode3}
+	expectedNodeList := []NodeInfo{testNode2, testNode1, testNode3}
+	sortNodesByUsage(nodeList, true) // ascending=true, sort nodes in ascending order
+
+	for i := 0; i < len(expectedNodeList); i++ {
+		if nodeList[i].NodeUsage.node.Name != expectedNodeList[i].NodeUsage.node.Name {
+			t.Errorf("Expected %v, got %v", expectedNodeList[i].NodeUsage.node.Name, nodeList[i].NodeUsage.node.Name)
+		}
+	}
+}

pkg/descheduler/strategies/pod_antiaffinity.go

@@ -20,6 +20,7 @@ import (
 	"context"
 	"fmt"
 
+	"k8s.io/apimachinery/pkg/util/sets"
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
@@ -48,18 +49,18 @@ func validateRemovePodsViolatingInterPodAntiAffinityParams(params *api.StrategyP
 }
 
 // RemovePodsViolatingInterPodAntiAffinity evicts pods on the node which are having a pod affinity rules.
-func RemovePodsViolatingInterPodAntiAffinity(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func RemovePodsViolatingInterPodAntiAffinity(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	if err := validateRemovePodsViolatingInterPodAntiAffinityParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid RemovePodsViolatingInterPodAntiAffinity parameters")
 		return
 	}
 
-	var includedNamespaces, excludedNamespaces []string
+	var includedNamespaces, excludedNamespaces sets.String
 	var labelSelector *metav1.LabelSelector
 	if strategy.Params != nil {
 		if strategy.Params.Namespaces != nil {
-			includedNamespaces = strategy.Params.Namespaces.Include
-			excludedNamespaces = strategy.Params.Namespaces.Exclude
+			includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+			excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
 		}
 		labelSelector = strategy.Params.LabelSelector
 	}
@@ -77,16 +78,19 @@ func RemovePodsViolatingInterPodAntiAffinity(ctx context.Context, client clients
 
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
+	podFilter, err := podutil.NewOptions().
+		WithNamespaces(includedNamespaces).
+		WithoutNamespaces(excludedNamespaces).
+		WithLabelSelector(labelSelector).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+
 	for _, node := range nodes {
 		klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
-		pods, err := podutil.ListPodsOnANode(
-			ctx,
-			client,
-			node,
-			podutil.WithNamespaces(includedNamespaces),
-			podutil.WithoutNamespaces(excludedNamespaces),
-			podutil.WithLabelSelector(labelSelector),
-		)
+		pods, err := podutil.ListPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
 		if err != nil {
 			return
 		}

pkg/descheduler/strategies/pod_antiaffinity_test.go

@@ -24,16 +24,17 @@ import (
 	policyv1 "k8s.io/api/policy/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
 
 func TestPodAntiAffinity(t *testing.T) {
-	ctx := context.Background()
 	node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
 	node2 := test.BuildTestNode("n2", 2000, 3000, 10, func(node *v1.Node) {
 		node.ObjectMeta.Labels = map[string]string{
@@ -46,6 +47,7 @@ func TestPodAntiAffinity(t *testing.T) {
 			Unschedulable: true,
 		}
 	})
+	node4 := test.BuildTestNode("n4", 2, 2, 1, nil)
 
 	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
 	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, nil)
@@ -55,6 +57,10 @@ func TestPodAntiAffinity(t *testing.T) {
 	p6 := test.BuildTestPod("p6", 100, 0, node1.Name, nil)
 	p7 := test.BuildTestPod("p7", 100, 0, node1.Name, nil)
 	p8 := test.BuildTestPod("p8", 100, 0, node1.Name, nil)
+	p9 := test.BuildTestPod("p9", 100, 0, node1.Name, nil)
+	p10 := test.BuildTestPod("p10", 100, 0, node1.Name, nil)
+	p9.DeletionTimestamp = &metav1.Time{}
+	p10.DeletionTimestamp = &metav1.Time{}
 
 	criticalPriority := utils.SystemCriticalPriority
 	nonEvictablePod := test.BuildTestPod("non-evict", 100, 0, node1.Name, func(pod *v1.Pod) {
@@ -72,6 +78,8 @@ func TestPodAntiAffinity(t *testing.T) {
 	test.SetNormalOwnerRef(p5)
 	test.SetNormalOwnerRef(p6)
 	test.SetNormalOwnerRef(p7)
+	test.SetNormalOwnerRef(p9)
+	test.SetNormalOwnerRef(p10)
 
 	// set pod anti affinity
 	setPodAntiAffinity(p1, "foo", "bar")
@@ -80,6 +88,8 @@ func TestPodAntiAffinity(t *testing.T) {
 	setPodAntiAffinity(p5, "foo1", "bar1")
 	setPodAntiAffinity(p6, "foo1", "bar1")
 	setPodAntiAffinity(p7, "foo", "bar")
+	setPodAntiAffinity(p9, "foo", "bar")
+	setPodAntiAffinity(p10, "foo", "bar")
 
 	// set pod priority
 	test.SetPodPriority(p5, 100)
@@ -91,98 +101,142 @@ func TestPodAntiAffinity(t *testing.T) {
 		"datacenter": "west",
 	}
 
+	var uint1 uint = 1
+	var uint3 uint = 3
+
 	tests := []struct {
-		description             string
-		maxPodsToEvictPerNode   int
-		pods                    []v1.Pod
-		expectedEvictedPodCount int
-		nodeFit                 bool
-		nodes                   []*v1.Node
+		description                    string
+		maxPodsToEvictPerNode          *uint
+		maxNoOfPodsToEvictPerNamespace *uint
+		pods                           []*v1.Pod
+		expectedEvictedPodCount        uint
+		nodeFit                        bool
+		nodes                          []*v1.Node
 	}{
 		{
 			description:             "Maximum pods to evict - 0",
-			maxPodsToEvictPerNode:   0,
-			pods:                    []v1.Pod{*p1, *p2, *p3, *p4},
+			pods:                    []*v1.Pod{p1, p2, p3, p4},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 3,
 		},
 		{
 			description:             "Maximum pods to evict - 3",
-			maxPodsToEvictPerNode:   3,
-			pods:                    []v1.Pod{*p1, *p2, *p3, *p4},
+			maxPodsToEvictPerNode:   &uint3,
+			pods:                    []*v1.Pod{p1, p2, p3, p4},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 3,
 		},
+		{
+			description:                    "Maximum pods to evict (maxPodsToEvictPerNamespace=3) - 3",
+			maxNoOfPodsToEvictPerNamespace: &uint3,
+			pods:                           []*v1.Pod{p1, p2, p3, p4},
+			nodes:                          []*v1.Node{node1},
+			expectedEvictedPodCount:        3,
+		},
 		{
 			description:             "Evict only 1 pod after sorting",
-			maxPodsToEvictPerNode:   0,
-			pods:                    []v1.Pod{*p5, *p6, *p7},
+			pods:                    []*v1.Pod{p5, p6, p7},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
 		{
 			description:             "Evicts pod that conflicts with critical pod (but does not evict critical pod)",
-			maxPodsToEvictPerNode:   1,
-			pods:                    []v1.Pod{*p1, *nonEvictablePod},
+			maxPodsToEvictPerNode:   &uint1,
+			pods:                    []*v1.Pod{p1, nonEvictablePod},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
 		{
 			description:             "Evicts pod that conflicts with critical pod (but does not evict critical pod)",
-			maxPodsToEvictPerNode:   1,
-			pods:                    []v1.Pod{*p1, *nonEvictablePod},
+			maxPodsToEvictPerNode:   &uint1,
+			pods:                    []*v1.Pod{p1, nonEvictablePod},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
 		{
 			description:             "Won't evict pods because node selectors don't match available nodes",
-			maxPodsToEvictPerNode:   1,
-			pods:                    []v1.Pod{*p8, *nonEvictablePod},
+			maxPodsToEvictPerNode:   &uint1,
+			pods:                    []*v1.Pod{p8, nonEvictablePod},
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
 			nodeFit:                 true,
 		},
 		{
 			description:             "Won't evict pods because only other node is not schedulable",
-			maxPodsToEvictPerNode:   1,
-			pods:                    []v1.Pod{*p8, *nonEvictablePod},
+			maxPodsToEvictPerNode:   &uint1,
+			pods:                    []*v1.Pod{p8, nonEvictablePod},
 			nodes:                   []*v1.Node{node1, node3},
 			expectedEvictedPodCount: 0,
 			nodeFit:                 true,
 		},
+		{
+			description:             "No pod to evicted since all pod terminating",
+			pods:                    []*v1.Pod{p9, p10},
+			nodes:                   []*v1.Node{node1},
+			expectedEvictedPodCount: 0,
+		},
+		{
+			description:             "Won't evict pods because only other node doesn't have enough resources",
+			maxPodsToEvictPerNode:   &uint3,
+			pods:                    []*v1.Pod{p1, p2, p3, p4},
+			nodes:                   []*v1.Node{node1, node4},
+			expectedEvictedPodCount: 0,
+			nodeFit:                 true,
+		},
 	}
 
 	for _, test := range tests {
-		// create fake client
-		fakeClient := &fake.Clientset{}
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			return true, &v1.PodList{Items: test.pods}, nil
-		})
-		fakeClient.Fake.AddReactor("get", "nodes", func(action core.Action) (bool, runtime.Object, error) {
-			return true, node1, nil
-		})
+		t.Run(test.description, func(t *testing.T) {
 
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			policyv1.SchemeGroupVersion.String(),
-			false,
-			test.maxPodsToEvictPerNode,
-			test.nodes,
-			false,
-			false,
-			false,
-		)
-		strategy := api.DeschedulerStrategy{
-			Params: &api.StrategyParameters{
-				NodeFit: test.nodeFit,
-			},
-		}
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
 
-		RemovePodsViolatingInterPodAntiAffinity(ctx, fakeClient, strategy, test.nodes, podEvictor)
-		podsEvicted := podEvictor.TotalEvicted()
-		if podsEvicted != test.expectedEvictedPodCount {
-			t.Errorf("Unexpected no of pods evicted: pods evicted: %d, expected: %d", podsEvicted, test.expectedEvictedPodCount)
-		}
+			var objs []runtime.Object
+			for _, node := range test.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range test.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := evictions.NewPodEvictor(
+				fakeClient,
+				policyv1.SchemeGroupVersion.String(),
+				false,
+				test.maxPodsToEvictPerNode,
+				test.maxNoOfPodsToEvictPerNamespace,
+				test.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
+				false,
+				false,
+				false,
+			)
+			strategy := api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: test.nodeFit,
+				},
+			}
+
+			RemovePodsViolatingInterPodAntiAffinity(ctx, fakeClient, strategy, test.nodes, podEvictor, getPodsAssignedToNode)
+			podsEvicted := podEvictor.TotalEvicted()
+			if podsEvicted != test.expectedEvictedPodCount {
+				t.Errorf("Unexpected no of pods evicted: pods evicted: %d, expected: %d", podsEvicted, test.expectedEvictedPodCount)
+			}
+		})
 	}
 }
 

pkg/descheduler/strategies/pod_lifetime.go

@@ -22,6 +22,7 @@ import (
 
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/sets"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
 
@@ -56,7 +57,7 @@ func validatePodLifeTimeParams(params *api.StrategyParameters) error {
 }
 
 // PodLifeTime evicts pods on nodes that were created more than strategy.Params.MaxPodLifeTimeSeconds seconds ago.
-func PodLifeTime(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func PodLifeTime(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	if err := validatePodLifeTimeParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid PodLifeTime parameters")
 		return
@@ -68,10 +69,10 @@ func PodLifeTime(ctx context.Context, client clientset.Interface, strategy api.D
 		return
 	}
 
-	var includedNamespaces, excludedNamespaces []string
+	var includedNamespaces, excludedNamespaces sets.String
 	if strategy.Params.Namespaces != nil {
-		includedNamespaces = strategy.Params.Namespaces.Include
-		excludedNamespaces = strategy.Params.Namespaces.Exclude
+		includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+		excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
 	}
 
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
@@ -88,10 +89,21 @@ func PodLifeTime(ctx context.Context, client clientset.Interface, strategy api.D
 		}
 	}
 
+	podFilter, err := podutil.NewOptions().
+		WithFilter(filter).
+		WithNamespaces(includedNamespaces).
+		WithoutNamespaces(excludedNamespaces).
+		WithLabelSelector(strategy.Params.LabelSelector).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+
 	for _, node := range nodes {
 		klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
 
-		pods := listOldPodsOnNode(ctx, client, node, includedNamespaces, excludedNamespaces, strategy.Params.LabelSelector, *strategy.Params.PodLifeTime.MaxPodLifeTimeSeconds, filter)
+		pods := listOldPodsOnNode(node.Name, getPodsAssignedToNode, podFilter, *strategy.Params.PodLifeTime.MaxPodLifeTimeSeconds)
 		for _, pod := range pods {
 			success, err := podEvictor.EvictPod(ctx, pod, node, "PodLifeTime")
 			if success {
@@ -108,23 +120,12 @@ func PodLifeTime(ctx context.Context, client clientset.Interface, strategy api.D
 }
 
 func listOldPodsOnNode(
-	ctx context.Context,
-	client clientset.Interface,
-	node *v1.Node,
-	includedNamespaces, excludedNamespaces []string,
-	labelSelector *metav1.LabelSelector,
+	nodeName string,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
+	filter podutil.FilterFunc,
 	maxPodLifeTimeSeconds uint,
-	filter func(pod *v1.Pod) bool,
 ) []*v1.Pod {
-	pods, err := podutil.ListPodsOnANode(
-		ctx,
-		client,
-		node,
-		podutil.WithFilter(filter),
-		podutil.WithNamespaces(includedNamespaces),
-		podutil.WithoutNamespaces(excludedNamespaces),
-		podutil.WithLabelSelector(labelSelector),
-	)
+	pods, err := podutil.ListPodsOnANode(nodeName, getPodsAssignedToNode, filter)
 	if err != nil {
 		return nil
 	}

pkg/descheduler/strategies/pod_lifetime_test.go

@@ -25,15 +25,16 @@ import (
 	policyv1 "k8s.io/api/policy/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/test"
 )
 
 func TestPodLifeTime(t *testing.T) {
-	ctx := context.Background()
 	node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
 	olderPodCreationTime := metav1.NewTime(time.Date(2009, time.November, 10, 23, 0, 0, 0, time.UTC))
 	newerPodCreationTime := metav1.NewTime(time.Now())
@@ -125,14 +126,24 @@ func TestPodLifeTime(t *testing.T) {
 	p12.ObjectMeta.OwnerReferences = ownerRef1
 	p13.ObjectMeta.OwnerReferences = ownerRef1
 
+	p14 := test.BuildTestPod("p14", 100, 0, node1.Name, nil)
+	p15 := test.BuildTestPod("p15", 100, 0, node1.Name, nil)
+	p14.Namespace = "dev"
+	p15.Namespace = "dev"
+	p14.ObjectMeta.CreationTimestamp = olderPodCreationTime
+	p15.ObjectMeta.CreationTimestamp = olderPodCreationTime
+	p14.ObjectMeta.OwnerReferences = ownerRef1
+	p15.ObjectMeta.OwnerReferences = ownerRef1
+	p14.DeletionTimestamp = &metav1.Time{}
+	p15.DeletionTimestamp = &metav1.Time{}
+
 	var maxLifeTime uint = 600
 	testCases := []struct {
 		description             string
 		strategy                api.DeschedulerStrategy
-		maxPodsToEvictPerNode   int
-		pods                    []v1.Pod
+		pods                    []*v1.Pod
 		nodes                   []*v1.Node
-		expectedEvictedPodCount int
+		expectedEvictedPodCount uint
 		ignorePvcPods           bool
 	}{
 		{
@@ -143,8 +154,7 @@ func TestPodLifeTime(t *testing.T) {
 					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p1, *p2},
+			pods:                    []*v1.Pod{p1, p2},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
@@ -156,8 +166,7 @@ func TestPodLifeTime(t *testing.T) {
 					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p3, *p4},
+			pods:                    []*v1.Pod{p3, p4},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 0,
 		},
@@ -169,8 +178,7 @@ func TestPodLifeTime(t *testing.T) {
 					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p5, *p6},
+			pods:                    []*v1.Pod{p5, p6},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
@@ -182,8 +190,7 @@ func TestPodLifeTime(t *testing.T) {
 					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p7, *p8},
+			pods:                    []*v1.Pod{p7, p8},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 0,
 		},
@@ -198,8 +205,7 @@ func TestPodLifeTime(t *testing.T) {
 					},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p9, *p10},
+			pods:                    []*v1.Pod{p9, p10},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
@@ -211,8 +217,7 @@ func TestPodLifeTime(t *testing.T) {
 					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p11},
+			pods:                    []*v1.Pod{p11},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 0,
 			ignorePvcPods:           true,
@@ -225,13 +230,12 @@ func TestPodLifeTime(t *testing.T) {
 					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p11},
+			pods:                    []*v1.Pod{p11},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
 		{
-			description: "Two old pods with different labels, 1 selected by labelSelector",
+			description: "No pod to evicted since all pod terminating",
 			strategy: api.DeschedulerStrategy{
 				Enabled: true,
 				Params: &api.StrategyParameters{
@@ -241,36 +245,72 @@ func TestPodLifeTime(t *testing.T) {
 					},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p12, *p13},
+			pods:                    []*v1.Pod{p12, p13},
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
+		{
+			description: "No pod should be evicted since pod terminating",
+			strategy: api.DeschedulerStrategy{
+				Enabled: true,
+				Params: &api.StrategyParameters{
+					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
+					LabelSelector: &metav1.LabelSelector{
+						MatchLabels: map[string]string{"foo": "bar"},
+					},
+				},
+			},
+			pods:                    []*v1.Pod{p14, p15},
+			nodes:                   []*v1.Node{node1},
+			expectedEvictedPodCount: 0,
+		},
 	}
 
 	for _, tc := range testCases {
-		fakeClient := &fake.Clientset{}
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
 
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			return true, &v1.PodList{Items: tc.pods}, nil
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range tc.pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := evictions.NewPodEvictor(
+				fakeClient,
+				policyv1.SchemeGroupVersion.String(),
+				false,
+				nil,
+				nil,
+				tc.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
+				tc.ignorePvcPods,
+				false,
+				false,
+			)
+
+			PodLifeTime(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor, getPodsAssignedToNode)
+			podsEvicted := podEvictor.TotalEvicted()
+			if podsEvicted != tc.expectedEvictedPodCount {
+				t.Errorf("Test error for description: %s. Expected evicted pods count %v, got %v", tc.description, tc.expectedEvictedPodCount, podsEvicted)
+			}
 		})
-
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			policyv1.SchemeGroupVersion.String(),
-			false,
-			tc.maxPodsToEvictPerNode,
-			tc.nodes,
-			false,
-			false,
-			tc.ignorePvcPods,
-		)
-
-		PodLifeTime(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor)
-		podsEvicted := podEvictor.TotalEvicted()
-		if podsEvicted != tc.expectedEvictedPodCount {
-			t.Errorf("Test error for description: %s. Expected evicted pods count %v, got %v", tc.description, tc.expectedEvictedPodCount, podsEvicted)
-		}
 	}
-
 }

pkg/descheduler/strategies/toomanyrestarts.go

@@ -21,6 +21,7 @@ import (
 	"fmt"
 
 	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/util/sets"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
 
@@ -49,7 +50,7 @@ func validateRemovePodsHavingTooManyRestartsParams(params *api.StrategyParameter
 // RemovePodsHavingTooManyRestarts removes the pods that have too many restarts on node.
 // There are too many cases leading this issue: Volume mount failed, app error due to nodes' different settings.
 // As of now, this strategy won't evict daemonsets, mirror pods, critical pods and pods with local storages.
-func RemovePodsHavingTooManyRestarts(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor) {
+func RemovePodsHavingTooManyRestarts(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
 	if err := validateRemovePodsHavingTooManyRestartsParams(strategy.Params); err != nil {
 		klog.ErrorS(err, "Invalid RemovePodsHavingTooManyRestarts parameters")
 		return
@@ -61,10 +62,10 @@ func RemovePodsHavingTooManyRestarts(ctx context.Context, client clientset.Inter
 		return
 	}
 
-	var includedNamespaces, excludedNamespaces []string
+	var includedNamespaces, excludedNamespaces sets.String
 	if strategy.Params.Namespaces != nil {
-		includedNamespaces = strategy.Params.Namespaces.Include
-		excludedNamespaces = strategy.Params.Namespaces.Exclude
+		includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+		excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
 	}
 
 	nodeFit := false
@@ -74,17 +75,20 @@ func RemovePodsHavingTooManyRestarts(ctx context.Context, client clientset.Inter
 
 	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
+	podFilter, err := podutil.NewOptions().
+		WithFilter(evictable.IsEvictable).
+		WithNamespaces(includedNamespaces).
+		WithoutNamespaces(excludedNamespaces).
+		WithLabelSelector(strategy.Params.LabelSelector).
+		BuildFilterFunc()
+	if err != nil {
+		klog.ErrorS(err, "Error initializing pod filter function")
+		return
+	}
+
 	for _, node := range nodes {
 		klog.V(1).InfoS("Processing node", "node", klog.KObj(node))
-		pods, err := podutil.ListPodsOnANode(
-			ctx,
-			client,
-			node,
-			podutil.WithFilter(evictable.IsEvictable),
-			podutil.WithNamespaces(includedNamespaces),
-			podutil.WithoutNamespaces(excludedNamespaces),
-			podutil.WithLabelSelector(strategy.Params.LabelSelector),
-		)
+		pods, err := podutil.ListPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
 		if err != nil {
 			klog.ErrorS(err, "Error listing a nodes pods", "node", klog.KObj(node))
 			continue

pkg/descheduler/strategies/toomanyrestarts_test.go

@@ -18,23 +18,24 @@ package strategies
 
 import (
 	"context"
-	"testing"
-
 	"fmt"
+	"testing"
 
 	v1 "k8s.io/api/core/v1"
 	policyv1 "k8s.io/api/policy/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/test"
 )
 
-func initPods(node *v1.Node) []v1.Pod {
-	pods := make([]v1.Pod, 0)
+func initPods(node *v1.Node) []*v1.Pod {
+	pods := make([]*v1.Pod, 0)
 
 	for i := int32(0); i <= 9; i++ {
 		pod := test.BuildTestPod(fmt.Sprintf("pod-%d", i), 100, 0, node.Name, nil)
@@ -56,7 +57,7 @@ func initPods(node *v1.Node) []v1.Pod {
 				},
 			},
 		}
-		pods = append(pods, *pod)
+		pods = append(pods, pod)
 	}
 
 	// The following 3 pods won't get evicted.
@@ -81,8 +82,6 @@ func initPods(node *v1.Node) []v1.Pod {
 }
 
 func TestRemovePodsHavingTooManyRestarts(t *testing.T) {
-	ctx := context.Background()
-
 	node1 := test.BuildTestNode("node1", 2000, 3000, 10, nil)
 	node2 := test.BuildTestNode("node2", 2000, 3000, 10, func(node *v1.Node) {
 		node.Spec.Taints = []v1.Taint{
@@ -98,8 +97,10 @@ func TestRemovePodsHavingTooManyRestarts(t *testing.T) {
 			Unschedulable: true,
 		}
 	})
+	node4 := test.BuildTestNode("node4", 200, 3000, 10, nil)
+	node5 := test.BuildTestNode("node5", 2000, 3000, 10, nil)
 
-	pods := initPods(node1)
+	pods := append(append(initPods(node1), test.BuildTestPod("CPU-consumer-1", 150, 100, node4.Name, nil)), test.BuildTestPod("CPU-consumer-2", 150, 100, node5.Name, nil))
 
 	createStrategy := func(enabled, includingInitContainers bool, restartThresholds int32, nodeFit bool) api.DeschedulerStrategy {
 		return api.DeschedulerStrategy{
@@ -114,115 +115,154 @@ func TestRemovePodsHavingTooManyRestarts(t *testing.T) {
 		}
 	}
 
+	var uint3 uint = 3
+
 	tests := []struct {
-		description             string
-		nodes                   []*v1.Node
-		strategy                api.DeschedulerStrategy
-		expectedEvictedPodCount int
-		maxPodsToEvictPerNode   int
+		description                    string
+		nodes                          []*v1.Node
+		strategy                       api.DeschedulerStrategy
+		expectedEvictedPodCount        uint
+		maxPodsToEvictPerNode          *uint
+		maxNoOfPodsToEvictPerNamespace *uint
 	}{
 		{
 			description:             "All pods have total restarts under threshold, no pod evictions",
 			strategy:                createStrategy(true, true, 10000, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 0,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Some pods have total restarts bigger than threshold",
 			strategy:                createStrategy(true, true, 1, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 6,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Nine pods have total restarts equals threshold(includingInitContainers=true), 6 pod evictions",
 			strategy:                createStrategy(true, true, 1*25, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 6,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Nine pods have total restarts equals threshold(includingInitContainers=false), 5 pod evictions",
 			strategy:                createStrategy(true, false, 1*25, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 5,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "All pods have total restarts equals threshold(includingInitContainers=true), 6 pod evictions",
 			strategy:                createStrategy(true, true, 1*20, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 6,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Nine pods have total restarts equals threshold(includingInitContainers=false), 6 pod evictions",
 			strategy:                createStrategy(true, false, 1*20, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 6,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Five pods have total restarts bigger than threshold(includingInitContainers=true), but only 1 pod eviction",
 			strategy:                createStrategy(true, true, 5*25+1, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "Five pods have total restarts bigger than threshold(includingInitContainers=false), but only 1 pod eviction",
 			strategy:                createStrategy(true, false, 5*20+1, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
-			maxPodsToEvictPerNode:   0,
 		},
 		{
 			description:             "All pods have total restarts equals threshold(maxPodsToEvictPerNode=3), 3 pod evictions",
 			strategy:                createStrategy(true, true, 1, false),
 			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 3,
-			maxPodsToEvictPerNode:   3,
+			maxPodsToEvictPerNode:   &uint3,
+		},
+		{
+			description:                    "All pods have total restarts equals threshold(maxNoOfPodsToEvictPerNamespace=3), 3 pod evictions",
+			strategy:                       createStrategy(true, true, 1, false),
+			nodes:                          []*v1.Node{node1},
+			expectedEvictedPodCount:        3,
+			maxNoOfPodsToEvictPerNamespace: &uint3,
 		},
 		{
 			description:             "All pods have total restarts equals threshold(maxPodsToEvictPerNode=3) but the only other node is tained, 0 pod evictions",
 			strategy:                createStrategy(true, true, 1, true),
 			nodes:                   []*v1.Node{node1, node2},
 			expectedEvictedPodCount: 0,
-			maxPodsToEvictPerNode:   3,
+			maxPodsToEvictPerNode:   &uint3,
 		},
 		{
 			description:             "All pods have total restarts equals threshold(maxPodsToEvictPerNode=3) but the only other node is not schedulable, 0 pod evictions",
 			strategy:                createStrategy(true, true, 1, true),
 			nodes:                   []*v1.Node{node1, node3},
 			expectedEvictedPodCount: 0,
-			maxPodsToEvictPerNode:   3,
+			maxPodsToEvictPerNode:   &uint3,
+		},
+		{
+			description:             "All pods have total restarts equals threshold(maxPodsToEvictPerNode=3) but the only other node does not have enough CPU, 0 pod evictions",
+			strategy:                createStrategy(true, true, 1, true),
+			nodes:                   []*v1.Node{node1, node4},
+			expectedEvictedPodCount: 0,
+			maxPodsToEvictPerNode:   &uint3,
+		},
+		{
+			description:             "All pods have total restarts equals threshold(maxPodsToEvictPerNode=3) but the only other node has enough CPU, 3 pod evictions",
+			strategy:                createStrategy(true, true, 1, true),
+			nodes:                   []*v1.Node{node1, node5},
+			expectedEvictedPodCount: 3,
+			maxPodsToEvictPerNode:   &uint3,
 		},
 	}
 
 	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
 
-		fakeClient := &fake.Clientset{}
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			return true, &v1.PodList{Items: pods}, nil
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range pods {
+				objs = append(objs, pod)
+			}
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			podEvictor := evictions.NewPodEvictor(
+				fakeClient,
+				policyv1.SchemeGroupVersion.String(),
+				false,
+				tc.maxPodsToEvictPerNode,
+				tc.maxNoOfPodsToEvictPerNamespace,
+				tc.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
+				false,
+				false,
+				false,
+			)
+
+			RemovePodsHavingTooManyRestarts(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor, getPodsAssignedToNode)
+			actualEvictedPodCount := podEvictor.TotalEvicted()
+			if actualEvictedPodCount != tc.expectedEvictedPodCount {
+				t.Errorf("Test %#v failed, expected %v pod evictions, but got %v pod evictions\n", tc.description, tc.expectedEvictedPodCount, actualEvictedPodCount)
+			}
 		})
-
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			policyv1.SchemeGroupVersion.String(),
-			false,
-			tc.maxPodsToEvictPerNode,
-			tc.nodes,
-			false,
-			false,
-			false,
-		)
-
-		RemovePodsHavingTooManyRestarts(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor)
-		actualEvictedPodCount := podEvictor.TotalEvicted()
-		if actualEvictedPodCount != tc.expectedEvictedPodCount {
-			t.Errorf("Test %#v failed, expected %v pod evictions, but got %v pod evictions\n", tc.description, tc.expectedEvictedPodCount, actualEvictedPodCount)
-		}
 	}
-
 }

pkg/descheduler/strategies/topologyspreadconstraint.go

@@ -18,21 +18,20 @@ package strategies
 
 import (
 	"context"
-	"fmt"
 	"math"
 	"sort"
 
-	utilerrors "k8s.io/apimachinery/pkg/util/errors"
-
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/labels"
-	"k8s.io/apimachinery/pkg/util/sets"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
+
 	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
-	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
+	"sigs.k8s.io/descheduler/pkg/descheduler/node"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies/validation"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
@@ -47,70 +46,24 @@ type topology struct {
 	pods []*v1.Pod
 }
 
-// topologySpreadStrategyParams contains validated strategy parameters
-type topologySpreadStrategyParams struct {
-	thresholdPriority  int32
-	includedNamespaces sets.String
-	excludedNamespaces sets.String
-	labelSelector      labels.Selector
-	nodeFit            bool
-}
-
-// validateAndParseTopologySpreadParams will validate parameters to ensure that they do not contain invalid values.
-func validateAndParseTopologySpreadParams(ctx context.Context, client clientset.Interface, params *api.StrategyParameters) (*topologySpreadStrategyParams, error) {
-	var includedNamespaces, excludedNamespaces sets.String
-	if params == nil {
-		return &topologySpreadStrategyParams{includedNamespaces: includedNamespaces, excludedNamespaces: excludedNamespaces}, nil
-	}
-	// At most one of include/exclude can be set
-	if params.Namespaces != nil && len(params.Namespaces.Include) > 0 && len(params.Namespaces.Exclude) > 0 {
-		return nil, fmt.Errorf("only one of Include/Exclude namespaces can be set")
-	}
-	if params.ThresholdPriority != nil && params.ThresholdPriorityClassName != "" {
-		return nil, fmt.Errorf("only one of thresholdPriority and thresholdPriorityClassName can be set")
-	}
-	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, params)
-	if err != nil {
-		return nil, fmt.Errorf("failed to get threshold priority from strategy's params: %+v", err)
-	}
-	if params.Namespaces != nil {
-		includedNamespaces = sets.NewString(params.Namespaces.Include...)
-		excludedNamespaces = sets.NewString(params.Namespaces.Exclude...)
-	}
-	var selector labels.Selector
-	if params.LabelSelector != nil {
-		selector, err = metav1.LabelSelectorAsSelector(params.LabelSelector)
-		if err != nil {
-			return nil, fmt.Errorf("failed to get label selectors from strategy's params: %+v", err)
-		}
-	}
-
-	return &topologySpreadStrategyParams{
-		thresholdPriority:  thresholdPriority,
-		includedNamespaces: includedNamespaces,
-		excludedNamespaces: excludedNamespaces,
-		labelSelector:      selector,
-		nodeFit:            params.NodeFit,
-	}, nil
-}
-
 func RemovePodsViolatingTopologySpreadConstraint(
 	ctx context.Context,
 	client clientset.Interface,
 	strategy api.DeschedulerStrategy,
 	nodes []*v1.Node,
 	podEvictor *evictions.PodEvictor,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 ) {
-	strategyParams, err := validateAndParseTopologySpreadParams(ctx, client, strategy.Params)
+	strategyParams, err := validation.ValidateAndParseStrategyParams(ctx, client, strategy.Params)
 	if err != nil {
 		klog.ErrorS(err, "Invalid RemovePodsViolatingTopologySpreadConstraint parameters")
 		return
 	}
 
 	evictable := podEvictor.Evictable(
-		evictions.WithPriorityThreshold(strategyParams.thresholdPriority),
-		evictions.WithNodeFit(strategyParams.nodeFit),
-		evictions.WithLabelSelector(strategyParams.labelSelector),
+		evictions.WithPriorityThreshold(strategyParams.ThresholdPriority),
+		evictions.WithNodeFit(strategyParams.NodeFit),
+		evictions.WithLabelSelector(strategyParams.LabelSelector),
 	)
 
 	nodeMap := make(map[string]*v1.Node, len(nodes))
@@ -140,8 +93,8 @@ func RemovePodsViolatingTopologySpreadConstraint(
 	podsForEviction := make(map[*v1.Pod]struct{})
 	// 1. for each namespace...
 	for _, namespace := range namespaces.Items {
-		if (len(strategyParams.includedNamespaces) > 0 && !strategyParams.includedNamespaces.Has(namespace.Name)) ||
-			(len(strategyParams.excludedNamespaces) > 0 && strategyParams.excludedNamespaces.Has(namespace.Name)) {
+		if (len(strategyParams.IncludedNamespaces) > 0 && !strategyParams.IncludedNamespaces.Has(namespace.Name)) ||
+			(len(strategyParams.ExcludedNamespaces) > 0 && strategyParams.ExcludedNamespaces.Has(namespace.Name)) {
 			continue
 		}
 		namespacePods, err := client.CoreV1().Pods(namespace.Name).List(ctx, metav1.ListOptions{})
@@ -186,6 +139,10 @@ func RemovePodsViolatingTopologySpreadConstraint(
 			// (this loop is where we count the number of pods per topologyValue that match this constraint's selector)
 			var sumPods float64
 			for i := range namespacePods.Items {
+				// skip pods that are being deleted.
+				if utils.IsPodTerminating(&namespacePods.Items[i]) {
+					continue
+				}
 				// 4. if the pod matches this TopologySpreadConstraint LabelSelector
 				if !selector.Matches(labels.Set(namespacePods.Items[i].Labels)) {
 					continue
@@ -211,7 +168,7 @@ func RemovePodsViolatingTopologySpreadConstraint(
 				klog.V(2).InfoS("Skipping topology constraint because it is already balanced", "constraint", constraint)
 				continue
 			}
-			balanceDomains(podsForEviction, constraint, constraintTopologies, sumPods, evictable.IsEvictable, nodeMap)
+			balanceDomains(client, getPodsAssignedToNode, podsForEviction, constraint, constraintTopologies, sumPods, evictable.IsEvictable, nodes)
 		}
 	}
 
@@ -266,12 +223,14 @@ func topologyIsBalanced(topology map[topologyPair][]*v1.Pod, constraint v1.Topol
 // [5, 5, 5, 5, 5, 5]
 // (assuming even distribution by the scheduler of the evicted pods)
 func balanceDomains(
+	client clientset.Interface,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 	podsForEviction map[*v1.Pod]struct{},
 	constraint v1.TopologySpreadConstraint,
 	constraintTopologies map[topologyPair][]*v1.Pod,
 	sumPods float64,
-	isEvictable func(*v1.Pod) bool,
-	nodeMap map[string]*v1.Node) {
+	isEvictable func(pod *v1.Pod) bool,
+	nodes []*v1.Node) {
 
 	idealAvg := sumPods / float64(len(constraintTopologies))
 	sortedDomains := sortDomains(constraintTopologies, isEvictable)
@@ -281,7 +240,7 @@ func balanceDomains(
 	j := len(sortedDomains) - 1
 	for i < j {
 		// if j has no more to give without falling below the ideal average, move to next aboveAvg
-		if float64(len(sortedDomains[j].pods)) < idealAvg {
+		if float64(len(sortedDomains[j].pods)) <= idealAvg {
 			j--
 		}
 
@@ -314,8 +273,19 @@ func balanceDomains(
 		// also (just for tracking), add them to the list of pods in the lower topology
 		aboveToEvict := sortedDomains[j].pods[len(sortedDomains[j].pods)-movePods:]
 		for k := range aboveToEvict {
-			if err := validatePodFitsOnOtherNodes(aboveToEvict[k], nodeMap); err != nil {
-				klog.V(2).InfoS(fmt.Sprintf("ignoring pod for eviction due to: %s", err.Error()), "pod", klog.KObj(aboveToEvict[k]))
+			// PodFitsAnyOtherNode excludes the current node because, for the sake of domain balancing only, we care about if there is any other
+			// place it could theoretically fit.
+			// If the pod doesn't fit on its current node, that is a job for RemovePodsViolatingNodeAffinity, and irrelevant to Topology Spreading
+			// Also, if the pod has a hard nodeAffinity/nodeSelector/toleration that only matches this node,
+			// don't bother evicting it as it will just end up back on the same node
+			// however we still account for it "being evicted" so the algorithm can complete
+			// TODO(@damemi): Since we don't order pods wrt their affinities, we should refactor this to skip the current pod
+			// but still try to get the required # of movePods (instead of just chopping that value off the slice above).
+			// In other words, PTS can perform suboptimally if some of its chosen pods don't fit on other nodes.
+			// This is because the chosen pods aren't sorted, but immovable pods still count as "evicted" toward the PTS algorithm.
+			// So, a better selection heuristic could improve performance.
+			if !node.PodFitsAnyOtherNode(getPodsAssignedToNode, aboveToEvict[k], nodes) {
+				klog.V(2).InfoS("ignoring pod for eviction as it does not fit on any other node", "pod", klog.KObj(aboveToEvict[k]))
 				continue
 			}
 
@@ -326,56 +296,6 @@ func balanceDomains(
 	}
 }
 
-// validatePodFitsOnOtherNodes performs validation based on scheduling predicates for affinity and toleration.
-// It excludes the current node because, for the sake of domain balancing only, we care about if there is any other
-// place it could theoretically fit.
-// If the pod doesn't fit on its current node, that is a job for RemovePodsViolatingNodeAffinity, and irrelevant to Topology Spreading
-func validatePodFitsOnOtherNodes(pod *v1.Pod, nodeMap map[string]*v1.Node) error {
-	// if the pod has a hard nodeAffinity/nodeSelector/toleration that only matches this node,
-	// don't bother evicting it as it will just end up back on the same node
-	// however we still account for it "being evicted" so the algorithm can complete
-	// TODO(@damemi): Since we don't order pods wrt their affinities, we should refactor this to skip the current pod
-	// but still try to get the required # of movePods (instead of just chopping that value off the slice above)
-	isRequiredDuringSchedulingIgnoredDuringExecution := pod.Spec.Affinity != nil &&
-		pod.Spec.Affinity.NodeAffinity != nil &&
-		pod.Spec.Affinity.NodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution != nil
-
-	hardTaintsFilter := func(taint *v1.Taint) bool {
-		return taint.Effect == v1.TaintEffectNoSchedule || taint.Effect == v1.TaintEffectNoExecute
-	}
-
-	var eligibleNodesCount, ineligibleAffinityNodesCount, ineligibleTaintedNodesCount int
-	for _, node := range nodeMap {
-		if node == nodeMap[pod.Spec.NodeName] {
-			continue
-		}
-		if pod.Spec.NodeSelector != nil || isRequiredDuringSchedulingIgnoredDuringExecution {
-			if !nodeutil.PodFitsCurrentNode(pod, node) {
-				ineligibleAffinityNodesCount++
-				continue
-			}
-		}
-		if !utils.TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, node.Spec.Taints, hardTaintsFilter) {
-			ineligibleTaintedNodesCount++
-			continue
-		}
-
-		eligibleNodesCount++
-	}
-
-	if eligibleNodesCount == 0 {
-		var errs []error
-		if ineligibleAffinityNodesCount > 0 {
-			errs = append(errs, fmt.Errorf("%d nodes with ineligible selector/affinity", ineligibleAffinityNodesCount))
-		}
-		if ineligibleTaintedNodesCount > 0 {
-			errs = append(errs, fmt.Errorf("%d nodes with taints that are not tolerated", ineligibleTaintedNodesCount))
-		}
-		return utilerrors.NewAggregate(errs)
-	}
-	return nil
-}
-
 // sortDomains sorts and splits the list of topology domains based on their size
 // it also sorts the list of pods within the domains based on their node affinity/selector and priority in the following order:
 // 1. non-evictable pods
@@ -383,7 +303,7 @@ func validatePodFitsOnOtherNodes(pod *v1.Pod, nodeMap map[string]*v1.Node) error
 // 3. pods in descending priority
 // 4. all other pods
 // We then pop pods off the back of the list for eviction
-func sortDomains(constraintTopologyPairs map[topologyPair][]*v1.Pod, isEvictable func(*v1.Pod) bool) []topology {
+func sortDomains(constraintTopologyPairs map[topologyPair][]*v1.Pod, isEvictable func(pod *v1.Pod) bool) []topology {
 	sortedTopologies := make([]topology, 0, len(constraintTopologyPairs))
 	// sort the topologies and return 2 lists: those <= the average and those > the average (> list inverted)
 	for pair, list := range constraintTopologyPairs {

pkg/descheduler/strategies/topologyspreadconstraint_test.go

@@ -5,23 +5,23 @@ import (
 	"fmt"
 	"testing"
 
-	"sigs.k8s.io/descheduler/pkg/api"
-
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
-	core "k8s.io/client-go/testing"
+
+	"sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/test"
 )
 
 func TestTopologySpreadConstraint(t *testing.T) {
-	ctx := context.Background()
 	testCases := []struct {
 		name                 string
 		pods                 []*v1.Pod
-		expectedEvictedCount int
+		expectedEvictedCount uint
 		nodes                []*v1.Node
 		strategy             api.DeschedulerStrategy
 		namespaces           []string
@@ -483,6 +483,38 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			namespaces: []string{"ns1"},
 		},
+		{
+			name: "2 domains size [2 6], maxSkew=2, can't move any because node1 does not have enough CPU",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 200, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneA" }),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneB" }),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(2),
+				},
+				{
+					count:  1,
+					node:   "n1",
+					labels: map[string]string{"foo": "bar"},
+				},
+				{
+					count:  6,
+					node:   "n2",
+					labels: map[string]string{"foo": "bar"},
+				},
+			}),
+			expectedEvictedCount: 0,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: true,
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
 		{
 			// see https://github.com/kubernetes-sigs/descheduler/issues/564
 			name: "Multiple constraints (6 nodes/2 zones, 4 pods)",
@@ -588,6 +620,62 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			strategy:             api.DeschedulerStrategy{Params: &api.StrategyParameters{IncludeSoftConstraints: true}},
 			namespaces:           []string{"ns1"},
 		},
+		{
+			name: "3 domains size [8 7 0], maxSkew=1, should move 5 to get [5 5 5]",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneA" }),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneB" }),
+				test.BuildTestNode("n3", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneC" }),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:       8,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:       7,
+					node:        "n2",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+			}),
+			expectedEvictedCount: 5,
+			strategy:             api.DeschedulerStrategy{},
+			namespaces:           []string{"ns1"},
+		},
+		{
+			name: "3 domains size [5 5 5], maxSkew=1, should move 0 to retain [5 5 5]",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneA" }),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneB" }),
+				test.BuildTestNode("n3", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneC" }),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:       5,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:       5,
+					node:        "n2",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:       5,
+					node:        "n3",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+			}),
+			expectedEvictedCount: 0,
+			strategy:             api.DeschedulerStrategy{},
+			namespaces:           []string{"ns1"},
+		},
 		{
 			name: "2 domains, sizes [2,0], maxSkew=1, move 1 pod since pod tolerates the node with taint",
 			nodes: []*v1.Node{
@@ -630,7 +718,7 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			namespaces:           []string{"ns1"},
 		},
 		{
-			name: "2 domains, sizes [2,0], maxSkew=1, move 0 pods since pod does not tolerate the tainted node",
+			name: "2 domains, sizes [2,0], maxSkew=1, move 1 pods since pod does not tolerate the tainted node",
 			nodes: []*v1.Node{
 				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneA" }),
 				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) {
@@ -662,6 +750,43 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			strategy:             api.DeschedulerStrategy{},
 			namespaces:           []string{"ns1"},
 		},
+		{
+			name: "2 domains, sizes [2,0], maxSkew=1, move 0 pods since pod does not tolerate the tainted node, and NodeFit is enabled",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneA" }),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels["zone"] = "zoneB"
+					n.Spec.Taints = []v1.Taint{
+						{
+							Key:    "taint-test",
+							Value:  "test",
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					nodeSelector: map[string]string{"zone": "zoneA"},
+				},
+			}),
+			expectedEvictedCount: 0,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: true,
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
 		{
 			name: "2 domains, sizes [2,0], maxSkew=1, move 1 pod for node with PreferNoSchedule Taint",
 			nodes: []*v1.Node{
@@ -776,33 +901,84 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			namespaces: []string{"ns1"},
 		},
+		{
+			name: "2 domains, sizes [4,2], maxSkew=1, 2 pods in termination; nothing should be moved",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneA" }),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneB" }),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:       2,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:             2,
+					node:              "n1",
+					labels:            map[string]string{"foo": "bar"},
+					constraints:       getDefaultTopologyConstraints(1),
+					deletionTimestamp: &metav1.Time{},
+				},
+				{
+					count:       2,
+					node:        "n2",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+			}),
+			expectedEvictedCount: 0,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					LabelSelector: getLabelSelector("foo", []string{"bar"}, metav1.LabelSelectorOpIn),
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
 	}
 
 	for _, tc := range testCases {
 		t.Run(tc.name, func(t *testing.T) {
-			fakeClient := &fake.Clientset{}
-			fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-				podList := make([]v1.Pod, 0, len(tc.pods))
-				for _, pod := range tc.pods {
-					podList = append(podList, *pod)
-				}
-				return true, &v1.PodList{Items: podList}, nil
-			})
-			fakeClient.Fake.AddReactor("list", "namespaces", func(action core.Action) (bool, runtime.Object, error) {
-				return true, &v1.NamespaceList{Items: []v1.Namespace{{ObjectMeta: metav1.ObjectMeta{Name: "ns1", Namespace: "ns1"}}}}, nil
-			})
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			var objs []runtime.Object
+			for _, node := range tc.nodes {
+				objs = append(objs, node)
+			}
+			for _, pod := range tc.pods {
+				objs = append(objs, pod)
+			}
+			objs = append(objs, &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "ns1"}})
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods()
+
+			getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+			if err != nil {
+				t.Errorf("Build get pods assigned to node function error: %v", err)
+			}
+
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
 
 			podEvictor := evictions.NewPodEvictor(
 				fakeClient,
 				"v1",
 				false,
-				100,
+				nil,
+				nil,
 				tc.nodes,
+				getPodsAssignedToNode,
+				false,
+				false,
 				false,
 				false,
 				false,
 			)
-			RemovePodsViolatingTopologySpreadConstraint(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor)
+			RemovePodsViolatingTopologySpreadConstraint(ctx, fakeClient, tc.strategy, tc.nodes, podEvictor, getPodsAssignedToNode)
 			podsEvicted := podEvictor.TotalEvicted()
 			if podsEvicted != tc.expectedEvictedCount {
 				t.Errorf("Test error for description: %s. Expected evicted pods count %v, got %v", tc.name, tc.expectedEvictedCount, podsEvicted)
@@ -812,14 +988,15 @@ func TestTopologySpreadConstraint(t *testing.T) {
 }
 
 type testPodList struct {
-	count        int
-	node         string
-	labels       map[string]string
-	constraints  []v1.TopologySpreadConstraint
-	nodeSelector map[string]string
-	nodeAffinity *v1.Affinity
-	noOwners     bool
-	tolerations  []v1.Toleration
+	count             int
+	node              string
+	labels            map[string]string
+	constraints       []v1.TopologySpreadConstraint
+	nodeSelector      map[string]string
+	nodeAffinity      *v1.Affinity
+	noOwners          bool
+	tolerations       []v1.Toleration
+	deletionTimestamp *metav1.Time
 }
 
 func createTestPods(testPods []testPodList) []*v1.Pod {
@@ -840,6 +1017,7 @@ func createTestPods(testPods []testPodList) []*v1.Pod {
 					p.Spec.NodeSelector = tp.nodeSelector
 					p.Spec.Affinity = tp.nodeAffinity
 					p.Spec.Tolerations = tp.tolerations
+					p.ObjectMeta.DeletionTimestamp = tp.deletionTimestamp
 				}))
 			podNum++
 		}

pkg/descheduler/strategies/validation/strategyparams.go

@@ -0,0 +1,71 @@
+package validation
+
+import (
+	"context"
+	"fmt"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/sets"
+	clientset "k8s.io/client-go/kubernetes"
+
+	"sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/utils"
+)
+
+// ValidatedStrategyParams contains validated common strategy parameters
+type ValidatedStrategyParams struct {
+	ThresholdPriority  int32
+	IncludedNamespaces sets.String
+	ExcludedNamespaces sets.String
+	LabelSelector      labels.Selector
+	NodeFit            bool
+}
+
+func DefaultValidatedStrategyParams() ValidatedStrategyParams {
+	return ValidatedStrategyParams{ThresholdPriority: utils.SystemCriticalPriority}
+}
+
+func ValidateAndParseStrategyParams(
+	ctx context.Context,
+	client clientset.Interface,
+	params *api.StrategyParameters,
+) (*ValidatedStrategyParams, error) {
+	if params == nil {
+		defaultValidatedStrategyParams := DefaultValidatedStrategyParams()
+		return &defaultValidatedStrategyParams, nil
+	}
+
+	// At most one of include/exclude can be set
+	var includedNamespaces, excludedNamespaces sets.String
+	if params.Namespaces != nil && len(params.Namespaces.Include) > 0 && len(params.Namespaces.Exclude) > 0 {
+		return nil, fmt.Errorf("only one of Include/Exclude namespaces can be set")
+	}
+	if params.ThresholdPriority != nil && params.ThresholdPriorityClassName != "" {
+		return nil, fmt.Errorf("only one of ThresholdPriority and thresholdPriorityClassName can be set")
+	}
+
+	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, params)
+	if err != nil {
+		return nil, fmt.Errorf("failed to get threshold priority from strategy's params: %+v", err)
+	}
+	if params.Namespaces != nil {
+		includedNamespaces = sets.NewString(params.Namespaces.Include...)
+		excludedNamespaces = sets.NewString(params.Namespaces.Exclude...)
+	}
+	var selector labels.Selector
+	if params.LabelSelector != nil {
+		selector, err = metav1.LabelSelectorAsSelector(params.LabelSelector)
+		if err != nil {
+			return nil, fmt.Errorf("failed to get label selectors from strategy's params: %+v", err)
+		}
+	}
+
+	return &ValidatedStrategyParams{
+		ThresholdPriority:  thresholdPriority,
+		IncludedNamespaces: includedNamespaces,
+		ExcludedNamespaces: excludedNamespaces,
+		LabelSelector:      selector,
+		NodeFit:            params.NodeFit,
+	}, nil
+}

pkg/descheduler/strategies/validation/strategyparams_test.go

@@ -0,0 +1,79 @@
+package validation
+
+import (
+	"context"
+	"testing"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/client-go/kubernetes/fake"
+
+	"sigs.k8s.io/descheduler/pkg/api"
+)
+
+var (
+	thresholdPriority int32 = 1000
+)
+
+func TestValidStrategyParams(t *testing.T) {
+	ctx := context.Background()
+	fakeClient := &fake.Clientset{}
+	testCases := []struct {
+		name   string
+		params *api.StrategyParameters
+	}{
+		{name: "validate nil params", params: nil},
+		{name: "validate empty params", params: &api.StrategyParameters{}},
+		{name: "validate params with NodeFit", params: &api.StrategyParameters{NodeFit: true}},
+		{name: "validate params with ThresholdPriority", params: &api.StrategyParameters{ThresholdPriority: &thresholdPriority}},
+		{name: "validate params with priorityClassName", params: &api.StrategyParameters{ThresholdPriorityClassName: "high-priority"}},
+		{name: "validate params with excluded namespace", params: &api.StrategyParameters{Namespaces: &api.Namespaces{Exclude: []string{"excluded-ns"}}}},
+		{name: "validate params with included namespace", params: &api.StrategyParameters{Namespaces: &api.Namespaces{Include: []string{"include-ns"}}}},
+		{name: "validate params with empty label selector", params: &api.StrategyParameters{LabelSelector: &metav1.LabelSelector{}}},
+	}
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			params, err := ValidateAndParseStrategyParams(ctx, fakeClient, tc.params)
+			if err != nil {
+				t.Errorf("strategy params should be valid but got err: %v", err.Error())
+			}
+			if params == nil {
+				t.Errorf("strategy params should return a strategyParams but got nil")
+			}
+		})
+	}
+}
+
+func TestInvalidStrategyParams(t *testing.T) {
+	ctx := context.Background()
+	fakeClient := &fake.Clientset{}
+	testCases := []struct {
+		name   string
+		params *api.StrategyParameters
+	}{
+		{
+			name:   "invalid params with both included and excluded namespaces nil params",
+			params: &api.StrategyParameters{Namespaces: &api.Namespaces{Include: []string{"include-ns"}, Exclude: []string{"exclude-ns"}}},
+		},
+		{
+			name:   "invalid params with both threshold priority and priority class name",
+			params: &api.StrategyParameters{ThresholdPriorityClassName: "high-priority", ThresholdPriority: &thresholdPriority},
+		},
+		{
+			name:   "invalid params with bad label selector",
+			params: &api.StrategyParameters{LabelSelector: &metav1.LabelSelector{MatchLabels: map[string]string{"": "missing-label"}}},
+		},
+	}
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			params, err := ValidateAndParseStrategyParams(ctx, fakeClient, tc.params)
+			if err == nil {
+				t.Errorf("strategy params should be invalid but did not get err")
+			}
+			if params != nil {
+				t.Errorf("strategy params should return a nil strategyParams but got %v", params)
+			}
+		})
+	}
+}

pkg/utils/pod.go

@@ -6,25 +6,9 @@ import (
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	utilfeature "k8s.io/apiserver/pkg/util/feature"
-	"k8s.io/component-base/featuregate"
 	"k8s.io/klog/v2"
 )
 
-const (
-	// owner: @jinxu
-	// beta: v1.10
-	//
-	// New local storage types to support local storage capacity isolation
-	LocalStorageCapacityIsolation featuregate.Feature = "LocalStorageCapacityIsolation"
-
-	// owner: @egernst
-	// alpha: v1.16
-	//
-	// Enables PodOverhead, for accounting pod overheads which are specific to a given RuntimeClass
-	PodOverhead featuregate.Feature = "PodOverhead"
-)
-
 // GetResourceRequest finds and returns the request value for a specific resource.
 func GetResourceRequest(pod *v1.Pod, resource v1.ResourceName) int64 {
 	if resource == v1.ResourcePods {
@@ -53,11 +37,6 @@ func GetResourceRequestQuantity(pod *v1.Pod, resourceName v1.ResourceName) resou
 		requestQuantity = resource.Quantity{Format: resource.DecimalSI}
 	}
 
-	if resourceName == v1.ResourceEphemeralStorage && !utilfeature.DefaultFeatureGate.Enabled(LocalStorageCapacityIsolation) {
-		// if the local storage capacity isolation feature gate is disabled, pods request 0 disk
-		return requestQuantity
-	}
-
 	for _, container := range pod.Spec.Containers {
 		if rQuantity, ok := container.Resources.Requests[resourceName]; ok {
 			requestQuantity.Add(rQuantity)
@@ -72,9 +51,9 @@ func GetResourceRequestQuantity(pod *v1.Pod, resourceName v1.ResourceName) resou
 		}
 	}
 
-	// if PodOverhead feature is supported, add overhead for running a pod
-	// to the total requests if the resource total is non-zero
-	if pod.Spec.Overhead != nil && utilfeature.DefaultFeatureGate.Enabled(PodOverhead) {
+	// We assume pod overhead feature gate is enabled.
+	// We can't import the scheduler settings so we will inherit the default.
+	if pod.Spec.Overhead != nil {
 		if podOverhead, ok := pod.Spec.Overhead[resourceName]; ok && !requestQuantity.IsZero() {
 			requestQuantity.Add(podOverhead)
 		}
@@ -89,6 +68,11 @@ func IsMirrorPod(pod *v1.Pod) bool {
 	return ok
 }
 
+// IsPodTerminating returns true if the pod DeletionTimestamp is set.
+func IsPodTerminating(pod *v1.Pod) bool {
+	return pod.DeletionTimestamp != nil
+}
+
 // IsStaticPod returns true if the pod is a static pod.
 func IsStaticPod(pod *v1.Pod) bool {
 	source, err := GetPodSource(pod)
@@ -157,9 +141,9 @@ func PodRequestsAndLimits(pod *v1.Pod) (reqs, limits v1.ResourceList) {
 		maxResourceList(limits, container.Resources.Limits)
 	}
 
-	// if PodOverhead feature is supported, add overhead for running a pod
-	// to the sum of reqeuests and to non-zero limits:
-	if pod.Spec.Overhead != nil && utilfeature.DefaultFeatureGate.Enabled(PodOverhead) {
+	// We assume pod overhead feature gate is enabled.
+	// We can't import the scheduler settings so we will inherit the default.
+	if pod.Spec.Overhead != nil {
 		addResourceList(reqs, pod.Spec.Overhead)
 
 		for name, quantity := range pod.Spec.Overhead {
@@ -202,12 +186,31 @@ func maxResourceList(list, new v1.ResourceList) {
 
 // PodToleratesTaints returns true if a pod tolerates one node's taints
 func PodToleratesTaints(pod *v1.Pod, taintsOfNodes map[string][]v1.Taint) bool {
-	for nodeName, taintsForNode := range taintsOfNodes {
-		if len(pod.Spec.Tolerations) >= len(taintsForNode) && TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, taintsForNode, nil) {
-			return true
-		}
-		klog.V(5).InfoS("Pod doesn't tolerate nodes taint", "pod", klog.KObj(pod), "nodeName", nodeName)
-	}
 
+	for nodeName, taintsForNode := range taintsOfNodes {
+		if len(pod.Spec.Tolerations) >= len(taintsForNode) {
+
+			if TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, taintsForNode, nil) {
+				return true
+			}
+
+			if klog.V(5).Enabled() {
+				for i := range taintsForNode {
+					if !TolerationsTolerateTaint(pod.Spec.Tolerations, &taintsForNode[i]) {
+						klog.V(5).InfoS("Pod doesn't tolerate node taint",
+							"pod", klog.KObj(pod),
+							"nodeName", nodeName,
+							"taint", fmt.Sprintf("%s:%s=%s", taintsForNode[i].Key, taintsForNode[i].Value, taintsForNode[i].Effect),
+						)
+					}
+				}
+			}
+		} else {
+			klog.V(5).InfoS("Pod doesn't tolerate nodes taint, count mismatch",
+				"pod", klog.KObj(pod),
+				"nodeName", nodeName,
+			)
+		}
+	}
 	return false
 }

test/e2e/e2e_duplicatepods_test.go

@@ -0,0 +1,177 @@
+/*
+Copyright 2021 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package e2e
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	appsv1 "k8s.io/api/apps/v1"
+	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	deschedulerapi "sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies"
+)
+
+func TestRemoveDuplicates(t *testing.T) {
+	ctx := context.Background()
+
+	clientSet, _, getPodsAssignedToNode, stopCh := initializeClient(t)
+	defer close(stopCh)
+
+	nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
+	if err != nil {
+		t.Errorf("Error listing node with %v", err)
+	}
+
+	nodes, workerNodes := splitNodesAndWorkerNodes(nodeList.Items)
+
+	t.Log("Creating testing namespace")
+	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
+	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("Unable to create ns %v", testNamespace.Name)
+	}
+	defer clientSet.CoreV1().Namespaces().Delete(ctx, testNamespace.Name, metav1.DeleteOptions{})
+
+	t.Log("Creating duplicates pods")
+
+	deploymentObj := &appsv1.Deployment{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "duplicate-pod",
+			Namespace: testNamespace.Name,
+			Labels:    map[string]string{"app": "test-duplicate", "name": "test-duplicatePods"},
+		},
+		Spec: appsv1.DeploymentSpec{
+			Selector: &metav1.LabelSelector{
+				MatchLabels: map[string]string{"app": "test-duplicate", "name": "test-duplicatePods"},
+			},
+			Template: v1.PodTemplateSpec{
+				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{"app": "test-duplicate", "name": "test-duplicatePods"},
+				},
+				Spec: v1.PodSpec{
+					Containers: []v1.Container{{
+						Name:            "pause",
+						ImagePullPolicy: "Always",
+						Image:           "kubernetes/pause",
+						Ports:           []v1.ContainerPort{{ContainerPort: 80}},
+					}},
+				},
+			},
+		},
+	}
+
+	tests := []struct {
+		description             string
+		replicasNum             int
+		beforeFunc              func(deployment *appsv1.Deployment)
+		expectedEvictedPodCount uint
+	}{
+		{
+			description: "Evict Pod even Pods schedule to specific node",
+			replicasNum: 4,
+			beforeFunc: func(deployment *appsv1.Deployment) {
+				deployment.Spec.Replicas = func(i int32) *int32 { return &i }(4)
+				deployment.Spec.Template.Spec.NodeName = workerNodes[0].Name
+			},
+			expectedEvictedPodCount: 2,
+		},
+		{
+			description: "Evict Pod even Pods with local storage",
+			replicasNum: 5,
+			beforeFunc: func(deployment *appsv1.Deployment) {
+				deployment.Spec.Replicas = func(i int32) *int32 { return &i }(5)
+				deployment.Spec.Template.Spec.Volumes = []v1.Volume{
+					{
+						Name: "sample",
+						VolumeSource: v1.VolumeSource{
+							EmptyDir: &v1.EmptyDirVolumeSource{
+								SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI),
+							},
+						},
+					},
+				}
+			},
+			expectedEvictedPodCount: 2,
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
+			t.Logf("Creating deployment %v in %v namespace", deploymentObj.Name, deploymentObj.Namespace)
+			tc.beforeFunc(deploymentObj)
+
+			_, err = clientSet.AppsV1().Deployments(deploymentObj.Namespace).Create(ctx, deploymentObj, metav1.CreateOptions{})
+			if err != nil {
+				t.Logf("Error creating deployment: %v", err)
+				if err = clientSet.AppsV1().Deployments(deploymentObj.Namespace).DeleteCollection(ctx, metav1.DeleteOptions{}, metav1.ListOptions{
+					LabelSelector: labels.SelectorFromSet(labels.Set(map[string]string{"app": "test-duplicate", "name": "test-duplicatePods"})).String(),
+				}); err != nil {
+					t.Fatalf("Unable to delete deployment: %v", err)
+				}
+				return
+			}
+			defer clientSet.AppsV1().Deployments(deploymentObj.Namespace).Delete(ctx, deploymentObj.Name, metav1.DeleteOptions{})
+			waitForPodsRunning(ctx, t, clientSet, map[string]string{"app": "test-duplicate", "name": "test-duplicatePods"}, tc.replicasNum, testNamespace.Name)
+
+			// Run DeschedulerStrategy strategy
+			evictionPolicyGroupVersion, err := eutils.SupportEviction(clientSet)
+			if err != nil || len(evictionPolicyGroupVersion) == 0 {
+				t.Fatalf("Error creating eviction policy group %v", err)
+			}
+			podEvictor := evictions.NewPodEvictor(
+				clientSet,
+				evictionPolicyGroupVersion,
+				false,
+				nil,
+				nil,
+				nodes,
+				getPodsAssignedToNode,
+				true,
+				false,
+				false,
+				false,
+				false,
+			)
+
+			t.Log("Running DeschedulerStrategy strategy")
+			strategies.RemoveDuplicatePods(
+				ctx,
+				clientSet,
+				deschedulerapi.DeschedulerStrategy{
+					Enabled: true,
+					Params: &deschedulerapi.StrategyParameters{
+						RemoveDuplicates: &deschedulerapi.RemoveDuplicates{},
+					},
+				},
+				workerNodes,
+				podEvictor,
+				getPodsAssignedToNode,
+			)
+
+			waitForTerminatingPodsToDisappear(ctx, t, clientSet, testNamespace.Name)
+			actualEvictedPodCount := podEvictor.TotalEvicted()
+			if actualEvictedPodCount != tc.expectedEvictedPodCount {
+				t.Errorf("Test error for description: %s. Unexpected number of pods have been evicted, got %v, expected %v", tc.description, actualEvictedPodCount, tc.expectedEvictedPodCount)
+			}
+		})
+	}
+}

test/e2e/e2e_failedpods_test.go

@@ -0,0 +1,156 @@
+package e2e
+
+import (
+	"context"
+	"strings"
+	"testing"
+	"time"
+
+	batchv1 "k8s.io/api/batch/v1"
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/wait"
+	clientset "k8s.io/client-go/kubernetes"
+
+	deschedulerapi "sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies"
+)
+
+var oneHourPodLifetimeSeconds uint = 3600
+
+func TestFailedPods(t *testing.T) {
+	ctx := context.Background()
+	clientSet, _, getPodsAssignedToNode, stopCh := initializeClient(t)
+	defer close(stopCh)
+	nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
+	if err != nil {
+		t.Errorf("Error listing node with %v", err)
+	}
+	nodes, _ := splitNodesAndWorkerNodes(nodeList.Items)
+	t.Log("Creating testing namespace")
+	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
+	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("Unable to create ns %v", testNamespace.Name)
+	}
+	defer clientSet.CoreV1().Namespaces().Delete(ctx, testNamespace.Name, metav1.DeleteOptions{})
+	testCases := map[string]struct {
+		expectedEvictedCount uint
+		strategyParams       *deschedulerapi.StrategyParameters
+	}{
+		"test-failed-pods-nil-strategy": {
+			expectedEvictedCount: 1,
+			strategyParams:       nil,
+		},
+		"test-failed-pods-default-strategy": {
+			expectedEvictedCount: 1,
+			strategyParams:       &deschedulerapi.StrategyParameters{},
+		},
+		"test-failed-pods-default-failed-pods": {
+			expectedEvictedCount: 1,
+			strategyParams: &deschedulerapi.StrategyParameters{
+				FailedPods: &deschedulerapi.FailedPods{},
+			},
+		},
+		"test-failed-pods-reason-unmatched": {
+			expectedEvictedCount: 0,
+			strategyParams: &deschedulerapi.StrategyParameters{
+				FailedPods: &deschedulerapi.FailedPods{Reasons: []string{"ReasonDoesNotMatch"}},
+			},
+		},
+		"test-failed-pods-min-age-unmet": {
+			expectedEvictedCount: 0,
+			strategyParams: &deschedulerapi.StrategyParameters{
+				FailedPods: &deschedulerapi.FailedPods{MinPodLifetimeSeconds: &oneHourPodLifetimeSeconds},
+			},
+		},
+		"test-failed-pods-exclude-job-kind": {
+			expectedEvictedCount: 0,
+			strategyParams: &deschedulerapi.StrategyParameters{
+				FailedPods: &deschedulerapi.FailedPods{ExcludeOwnerKinds: []string{"Job"}},
+			},
+		},
+	}
+	for name, tc := range testCases {
+		t.Run(name, func(t *testing.T) {
+			job := initFailedJob(name, testNamespace.Namespace)
+			t.Logf("Creating job %s in %s namespace", job.Name, job.Namespace)
+			jobClient := clientSet.BatchV1().Jobs(testNamespace.Name)
+			if _, err := jobClient.Create(ctx, job, metav1.CreateOptions{}); err != nil {
+				t.Fatalf("Error creating Job %s: %v", name, err)
+			}
+			deletePropagationPolicy := metav1.DeletePropagationForeground
+			defer jobClient.Delete(ctx, job.Name, metav1.DeleteOptions{PropagationPolicy: &deletePropagationPolicy})
+			waitForJobPodPhase(ctx, t, clientSet, job, v1.PodFailed)
+
+			podEvictor := initPodEvictorOrFail(t, clientSet, getPodsAssignedToNode, nodes)
+
+			t.Logf("Running RemoveFailedPods strategy for %s", name)
+			strategies.RemoveFailedPods(
+				ctx,
+				clientSet,
+				deschedulerapi.DeschedulerStrategy{
+					Enabled: true,
+					Params:  tc.strategyParams,
+				},
+				nodes,
+				podEvictor,
+				getPodsAssignedToNode,
+			)
+			t.Logf("Finished RemoveFailedPods strategy for %s", name)
+
+			if actualEvictedCount := podEvictor.TotalEvicted(); actualEvictedCount == tc.expectedEvictedCount {
+				t.Logf("Total of %d Pods were evicted for %s", actualEvictedCount, name)
+			} else {
+				t.Errorf("Unexpected number of pods have been evicted, got %v, expected %v", actualEvictedCount, tc.expectedEvictedCount)
+			}
+		})
+	}
+}
+
+func initFailedJob(name, namespace string) *batchv1.Job {
+	podSpec := MakePodSpec("", nil)
+	podSpec.Containers[0].Command = []string{"/bin/false"}
+	podSpec.RestartPolicy = v1.RestartPolicyNever
+	labelsSet := labels.Set{"test": name, "name": name}
+	jobBackoffLimit := int32(0)
+	return &batchv1.Job{
+		ObjectMeta: metav1.ObjectMeta{
+			Labels:    labelsSet,
+			Name:      name,
+			Namespace: namespace,
+		},
+		Spec: batchv1.JobSpec{
+			Template: v1.PodTemplateSpec{
+				Spec:       podSpec,
+				ObjectMeta: metav1.ObjectMeta{Labels: labelsSet},
+			},
+			BackoffLimit: &jobBackoffLimit,
+		},
+	}
+}
+
+func waitForJobPodPhase(ctx context.Context, t *testing.T, clientSet clientset.Interface, job *batchv1.Job, phase v1.PodPhase) {
+	podClient := clientSet.CoreV1().Pods(job.Namespace)
+	if err := wait.PollImmediate(5*time.Second, 30*time.Second, func() (bool, error) {
+		t.Log(labels.FormatLabels(job.Labels))
+		if podList, err := podClient.List(ctx, metav1.ListOptions{LabelSelector: labels.FormatLabels(job.Labels)}); err != nil {
+			return false, err
+		} else {
+			if len(podList.Items) == 0 {
+				t.Logf("Job controller has not created Pod for job %s yet", job.Name)
+				return false, nil
+			}
+			for _, pod := range podList.Items {
+				if pod.Status.Phase != phase {
+					t.Logf("Pod %v not in %s phase yet, is %v instead", pod.Name, phase, pod.Status.Phase)
+					return false, nil
+				}
+			}
+			t.Logf("Job %v Pod is in %s phase now", job.Name, phase)
+			return true, nil
+		}
+	}); err != nil {
+		t.Fatalf("Error waiting for pods in %s phase: %v", phase, err)
+	}
+}

test/e2e/e2e_leaderelection_test.go

@@ -0,0 +1,201 @@
+/*
+Copyright 2022 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package e2e
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"reflect"
+	"strings"
+	"testing"
+	"time"
+
+	appsv1 "k8s.io/api/apps/v1"
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	clientset "k8s.io/client-go/kubernetes"
+
+	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
+	"sigs.k8s.io/descheduler/pkg/descheduler"
+)
+
+func TestLeaderElection(t *testing.T) {
+	ctx := context.Background()
+
+	clientSet, _, _, stopCh := initializeClient(t)
+	defer close(stopCh)
+
+	ns1 := "e2e-" + strings.ToLower(t.Name()+"-a")
+	ns2 := "e2e-" + strings.ToLower(t.Name()+"-b")
+
+	t.Logf("Creating testing namespace %v", ns1)
+	testNamespace1 := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: ns1}}
+	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace1, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("Unable to create ns %v", testNamespace1.Name)
+	}
+	defer clientSet.CoreV1().Namespaces().Delete(ctx, testNamespace1.Name, metav1.DeleteOptions{})
+
+	t.Logf("Creating testing namespace %v", ns2)
+	testNamespace2 := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: ns2}}
+	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace2, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("Unable to create ns %v", testNamespace2.Name)
+	}
+	defer clientSet.CoreV1().Namespaces().Delete(ctx, testNamespace2.Name, metav1.DeleteOptions{})
+
+	deployment1, err := createDeployment(ctx, clientSet, ns1, 5, t)
+	if err != nil {
+		t.Fatalf("create deployment 1: %v", err)
+	}
+	defer clientSet.AppsV1().Deployments(deployment1.Namespace).Delete(ctx, deployment1.Name, metav1.DeleteOptions{})
+
+	deployment2, err := createDeployment(ctx, clientSet, ns2, 5, t)
+	if err != nil {
+		t.Fatalf("create deployment 2: %v", err)
+	}
+	defer clientSet.AppsV1().Deployments(deployment2.Namespace).Delete(ctx, deployment2.Name, metav1.DeleteOptions{})
+
+	waitForPodsRunning(ctx, t, clientSet, map[string]string{"test": "leaderelection", "name": "test-leaderelection"}, 5, ns1)
+
+	podListAOrg := getPodNameList(ctx, clientSet, ns1, t)
+
+	waitForPodsRunning(ctx, t, clientSet, map[string]string{"test": "leaderelection", "name": "test-leaderelection"}, 5, ns2)
+
+	podListBOrg := getPodNameList(ctx, clientSet, ns2, t)
+
+	s1, err := options.NewDeschedulerServer()
+	if err != nil {
+		t.Fatalf("unable to initialize server: %v", err)
+	}
+	s1.Client = clientSet
+	s1.DeschedulingInterval = 5 * time.Second
+	s1.LeaderElection.LeaderElect = true
+	s1.KubeconfigFile = os.Getenv("KUBECONFIG")
+	s1.PolicyConfigFile = "./policy_leaderelection_a.yaml"
+
+	s2, err := options.NewDeschedulerServer()
+	if err != nil {
+		t.Fatalf("unable to initialize server: %v", err)
+	}
+	s2.Client = clientSet
+	s2.DeschedulingInterval = 5 * time.Second
+	s2.LeaderElection.LeaderElect = true
+	s2.KubeconfigFile = os.Getenv("KUBECONFIG")
+	s2.PolicyConfigFile = "./policy_leaderelection_b.yaml"
+
+	t.Log("starting deschedulers")
+
+	go func() {
+		err := descheduler.Run(ctx, s1)
+		if err != nil {
+			t.Errorf("unable to start descheduler: %v", err)
+			return
+		}
+	}()
+
+	time.Sleep(1 * time.Second)
+
+	go func() {
+		err := descheduler.Run(ctx, s2)
+		if err != nil {
+			t.Errorf("unable to start descheduler: %v", err)
+			return
+		}
+	}()
+
+	defer clientSet.CoordinationV1().Leases(s1.LeaderElection.ResourceNamespace).Delete(ctx, s1.LeaderElection.ResourceName, metav1.DeleteOptions{})
+	defer clientSet.CoordinationV1().Leases(s2.LeaderElection.ResourceNamespace).Delete(ctx, s2.LeaderElection.ResourceName, metav1.DeleteOptions{})
+
+	// wait for a while so all the pods are 5 seconds older
+	time.Sleep(7 * time.Second)
+
+	// validate only pods from e2e-testleaderelection-a namespace are evicted.
+	podListA := getPodNameList(ctx, clientSet, ns1, t)
+
+	podListB := getPodNameList(ctx, clientSet, ns2, t)
+
+	left := reflect.DeepEqual(podListAOrg, podListA)
+	right := reflect.DeepEqual(podListBOrg, podListB)
+
+	singleNamespaceEvicted := (left && !right) || (!left && right)
+
+	if singleNamespaceEvicted {
+		if !left {
+			t.Logf("Only the pods in %s namespace are evicted. Pods before: %s, Pods after %s", ns1, podListAOrg, podListA)
+		} else {
+			t.Logf("Only the pods in %s namespace are evicted. Pods before: %s, Pods after %s", ns2, podListBOrg, podListB)
+		}
+	} else {
+		t.Fatalf("Pods are evicted in both namespaces. For %s namespace Pods before: %s, Pods after %s. And, for %s namespace Pods before: %s, Pods after: %s", ns1, podListAOrg, podListA, ns2, podListBOrg, podListB)
+	}
+}
+
+func createDeployment(ctx context.Context, clientSet clientset.Interface, namespace string, replicas int32, t *testing.T) (*appsv1.Deployment, error) {
+	deployment := &appsv1.Deployment{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "leaderelection",
+			Namespace: namespace,
+			Labels:    map[string]string{"test": "leaderelection", "name": "test-leaderelection"},
+		},
+		Spec: appsv1.DeploymentSpec{
+			Replicas: func(i int32) *int32 { return &i }(replicas),
+			Selector: &metav1.LabelSelector{
+				MatchLabels: map[string]string{"test": "leaderelection", "name": "test-leaderelection"},
+			},
+			Template: v1.PodTemplateSpec{
+				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{"test": "leaderelection", "name": "test-leaderelection"},
+				},
+				Spec: v1.PodSpec{
+					Containers: []v1.Container{{
+						Name:            "pause",
+						ImagePullPolicy: "Always",
+						Image:           "kubernetes/pause",
+						Ports:           []v1.ContainerPort{{ContainerPort: 80}},
+					}},
+				},
+			},
+		},
+	}
+
+	t.Logf("Creating deployment %v for namespace %s", deployment.Name, deployment.Namespace)
+	deployment, err := clientSet.AppsV1().Deployments(deployment.Namespace).Create(ctx, deployment, metav1.CreateOptions{})
+	if err != nil {
+		t.Logf("Error creating deployment: %v", err)
+		if err = clientSet.AppsV1().Deployments(deployment.Namespace).DeleteCollection(ctx, metav1.DeleteOptions{}, metav1.ListOptions{
+			LabelSelector: labels.SelectorFromSet(labels.Set(map[string]string{"test": "leaderelection", "name": "test-leaderelection"})).String(),
+		}); err != nil {
+			t.Fatalf("Unable to delete deployment: %v", err)
+		}
+		return nil, fmt.Errorf("create deployment %v", err)
+	}
+	return deployment, nil
+}
+
+func getPodNameList(ctx context.Context, clientSet clientset.Interface, namespace string, t *testing.T) []string {
+	podList, err := clientSet.CoreV1().Pods(namespace).List(
+		ctx, metav1.ListOptions{LabelSelector: labels.SelectorFromSet(labels.Set(map[string]string{"test": "leaderelection", "name": "test-leaderelection"})).String()})
+	if err != nil {
+		t.Fatalf("Unable to list pods from ns: %s: %v", namespace, err)
+	}
+	podNames := make([]string, len(podList.Items))
+	for i, pod := range podList.Items {
+		podNames[i] = pod.Name
+	}
+	return podNames
+}

test/e2e/e2e_test.go

@@ -39,7 +39,6 @@ import (
 	v1qos "k8s.io/kubectl/pkg/util/qos"
 
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
-	"sigs.k8s.io/descheduler/pkg/api"
 	deschedulerapi "sigs.k8s.io/descheduler/pkg/api"
 	"sigs.k8s.io/descheduler/pkg/descheduler"
 	"sigs.k8s.io/descheduler/pkg/descheduler/client"
@@ -62,11 +61,11 @@ func MakePodSpec(priorityClassName string, gracePeriod *int64) v1.PodSpec {
 			Resources: v1.ResourceRequirements{
 				Limits: v1.ResourceList{
 					v1.ResourceCPU:    resource.MustParse("100m"),
-					v1.ResourceMemory: resource.MustParse("1000Mi"),
+					v1.ResourceMemory: resource.MustParse("200Mi"),
 				},
 				Requests: v1.ResourceList{
 					v1.ResourceCPU:    resource.MustParse("100m"),
-					v1.ResourceMemory: resource.MustParse("800Mi"),
+					v1.ResourceMemory: resource.MustParse("100Mi"),
 				},
 			},
 		}},
@@ -75,7 +74,7 @@ func MakePodSpec(priorityClassName string, gracePeriod *int64) v1.PodSpec {
 	}
 }
 
-// RcByNameContainer returns a ReplicationControoler with specified name and container
+// RcByNameContainer returns a ReplicationController with specified name and container
 func RcByNameContainer(name, namespace string, replicas int32, labels map[string]string, gracePeriod *int64, priorityClassName string) *v1.ReplicationController {
 	zeroGracePeriod := int64(0)
 
@@ -108,7 +107,7 @@ func RcByNameContainer(name, namespace string, replicas int32, labels map[string
 	}
 }
 
-func initializeClient(t *testing.T) (clientset.Interface, coreinformers.NodeInformer, chan struct{}) {
+func initializeClient(t *testing.T) (clientset.Interface, coreinformers.NodeInformer, podutil.GetPodsAssignedToNodeFunc, chan struct{}) {
 	clientSet, err := client.CreateClient(os.Getenv("KUBECONFIG"))
 	if err != nil {
 		t.Errorf("Error during client creation with %v", err)
@@ -117,12 +116,41 @@ func initializeClient(t *testing.T) (clientset.Interface, coreinformers.NodeInfo
 	stopChannel := make(chan struct{})
 
 	sharedInformerFactory := informers.NewSharedInformerFactory(clientSet, 0)
+	nodeInformer := sharedInformerFactory.Core().V1().Nodes()
+	podInformer := sharedInformerFactory.Core().V1().Pods()
+
+	getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+	if err != nil {
+		t.Errorf("build get pods assigned to node function error: %v", err)
+	}
+
 	sharedInformerFactory.Start(stopChannel)
 	sharedInformerFactory.WaitForCacheSync(stopChannel)
 
-	nodeInformer := sharedInformerFactory.Core().V1().Nodes()
+	waitForNodesReady(context.Background(), t, clientSet, nodeInformer)
+	return clientSet, nodeInformer, getPodsAssignedToNode, stopChannel
+}
 
-	return clientSet, nodeInformer, stopChannel
+func waitForNodesReady(ctx context.Context, t *testing.T, clientSet clientset.Interface, nodeInformer coreinformers.NodeInformer) {
+	if err := wait.PollImmediate(5*time.Second, 30*time.Second, func() (bool, error) {
+		nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
+		if err != nil {
+			return false, err
+		}
+
+		readyNodes, err := nodeutil.ReadyNodes(ctx, clientSet, nodeInformer, "")
+		if err != nil {
+			return false, err
+		}
+		if len(nodeList.Items) != len(readyNodes) {
+			t.Logf("%v/%v nodes are ready. Waiting for all nodes to be ready...", len(readyNodes), len(nodeList.Items))
+			return false, nil
+		}
+
+		return true, nil
+	}); err != nil {
+		t.Fatalf("Error waiting for nodes to be ready: %v", err)
+	}
 }
 
 func runPodLifetimeStrategy(
@@ -135,6 +163,7 @@ func runPodLifetimeStrategy(
 	priority *int32,
 	evictCritical bool,
 	labelSelector *metav1.LabelSelector,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 ) {
 	// Run descheduler.
 	evictionPolicyGroupVersion, err := eutils.SupportEviction(clientset)
@@ -166,12 +195,17 @@ func runPodLifetimeStrategy(
 			clientset,
 			evictionPolicyGroupVersion,
 			false,
-			0,
+			nil,
+			nil,
 			nodes,
+			getPodsAssignedToNode,
 			false,
 			evictCritical,
 			false,
+			false,
+			false,
 		),
+		getPodsAssignedToNode,
 	)
 }
 
@@ -201,7 +235,7 @@ func intersectStrings(lista, listb []string) []string {
 func TestLowNodeUtilization(t *testing.T) {
 	ctx := context.Background()
 
-	clientSet, _, stopCh := initializeClient(t)
+	clientSet, _, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
 	nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
@@ -209,15 +243,7 @@ func TestLowNodeUtilization(t *testing.T) {
 		t.Errorf("Error listing node with %v", err)
 	}
 
-	var nodes []*v1.Node
-	var workerNodes []*v1.Node
-	for i := range nodeList.Items {
-		node := nodeList.Items[i]
-		nodes = append(nodes, &node)
-		if _, exists := node.Labels["node-role.kubernetes.io/master"]; !exists {
-			workerNodes = append(workerNodes, &node)
-		}
-	}
+	nodes, workerNodes := splitNodesAndWorkerNodes(nodeList.Items)
 
 	t.Log("Creating testing namespace")
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
@@ -298,22 +324,14 @@ func TestLowNodeUtilization(t *testing.T) {
 	waitForRCPodsRunning(ctx, t, clientSet, rc)
 
 	// Run LowNodeUtilization strategy
-	evictionPolicyGroupVersion, err := eutils.SupportEviction(clientSet)
-	if err != nil || len(evictionPolicyGroupVersion) == 0 {
-		t.Fatalf("%v", err)
-	}
-	podEvictor := evictions.NewPodEvictor(
-		clientSet,
-		evictionPolicyGroupVersion,
-		false,
-		0,
-		nodes,
-		true,
-		false,
-		false,
-	)
+	podEvictor := initPodEvictorOrFail(t, clientSet, getPodsAssignedToNode, nodes)
 
-	podsOnMosttUtilizedNode, err := podutil.ListPodsOnANode(ctx, clientSet, workerNodes[0], podutil.WithFilter(podEvictor.Evictable().IsEvictable))
+	podFilter, err := podutil.NewOptions().WithFilter(podEvictor.Evictable().IsEvictable).BuildFilterFunc()
+	if err != nil {
+		t.Errorf("Error initializing pod filter function, %v", err)
+	}
+
+	podsOnMosttUtilizedNode, err := podutil.ListPodsOnANode(workerNodes[0].Name, getPodsAssignedToNode, podFilter)
 	if err != nil {
 		t.Errorf("Error listing pods on a node %v", err)
 	}
@@ -338,11 +356,17 @@ func TestLowNodeUtilization(t *testing.T) {
 		},
 		workerNodes,
 		podEvictor,
+		getPodsAssignedToNode,
 	)
 
 	waitForTerminatingPodsToDisappear(ctx, t, clientSet, rc.Namespace)
 
-	podsOnMosttUtilizedNode, err = podutil.ListPodsOnANode(ctx, clientSet, workerNodes[0], podutil.WithFilter(podEvictor.Evictable().IsEvictable))
+	podFilter, err = podutil.NewOptions().WithFilter(podEvictor.Evictable().IsEvictable).BuildFilterFunc()
+	if err != nil {
+		t.Errorf("Error initializing pod filter function, %v", err)
+	}
+
+	podsOnMosttUtilizedNode, err = podutil.ListPodsOnANode(workerNodes[0].Name, getPodsAssignedToNode, podFilter)
 	if err != nil {
 		t.Errorf("Error listing pods on a node %v", err)
 	}
@@ -359,7 +383,7 @@ func TestLowNodeUtilization(t *testing.T) {
 func TestNamespaceConstraintsInclude(t *testing.T) {
 	ctx := context.Background()
 
-	clientSet, nodeInformer, stopCh := initializeClient(t)
+	clientSet, nodeInformer, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
@@ -394,7 +418,7 @@ func TestNamespaceConstraintsInclude(t *testing.T) {
 	t.Logf("set the strategy to delete pods from %v namespace", rc.Namespace)
 	runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, &deschedulerapi.Namespaces{
 		Include: []string{rc.Namespace},
-	}, "", nil, false, nil)
+	}, "", nil, false, nil, getPodsAssignedToNode)
 
 	// All pods are supposed to be deleted, wait until all the old pods are deleted
 	if err := wait.PollImmediate(time.Second, 20*time.Second, func() (bool, error) {
@@ -430,7 +454,7 @@ func TestNamespaceConstraintsInclude(t *testing.T) {
 func TestNamespaceConstraintsExclude(t *testing.T) {
 	ctx := context.Background()
 
-	clientSet, nodeInformer, stopCh := initializeClient(t)
+	clientSet, nodeInformer, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
@@ -465,7 +489,7 @@ func TestNamespaceConstraintsExclude(t *testing.T) {
 	t.Logf("set the strategy to delete pods from namespaces except the %v namespace", rc.Namespace)
 	runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, &deschedulerapi.Namespaces{
 		Exclude: []string{rc.Namespace},
-	}, "", nil, false, nil)
+	}, "", nil, false, nil, getPodsAssignedToNode)
 
 	t.Logf("Waiting 10s")
 	time.Sleep(10 * time.Second)
@@ -493,11 +517,11 @@ func TestEvictSystemCriticalPriorityClass(t *testing.T) {
 }
 
 func testEvictSystemCritical(t *testing.T, isPriorityClass bool) {
-	var highPriority = int32(1000)
-	var lowPriority = int32(500)
+	highPriority := int32(1000)
+	lowPriority := int32(500)
 	ctx := context.Background()
 
-	clientSet, nodeInformer, stopCh := initializeClient(t)
+	clientSet, nodeInformer, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
@@ -578,9 +602,9 @@ func testEvictSystemCritical(t *testing.T, isPriorityClass bool) {
 	t.Logf("Existing pods: %v", initialPodNames)
 
 	if isPriorityClass {
-		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, highPriorityClass.Name, nil, true, nil)
+		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, highPriorityClass.Name, nil, true, nil, getPodsAssignedToNode)
 	} else {
-		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", &highPriority, true, nil)
+		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", &highPriority, true, nil, getPodsAssignedToNode)
 	}
 
 	// All pods are supposed to be deleted, wait until all pods in the test namespace are terminating
@@ -623,11 +647,11 @@ func TestThresholdPriorityClass(t *testing.T) {
 }
 
 func testPriority(t *testing.T, isPriorityClass bool) {
-	var highPriority = int32(1000)
-	var lowPriority = int32(500)
+	highPriority := int32(1000)
+	lowPriority := int32(500)
 	ctx := context.Background()
 
-	clientSet, nodeInformer, stopCh := initializeClient(t)
+	clientSet, nodeInformer, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
@@ -697,10 +721,10 @@ func testPriority(t *testing.T, isPriorityClass bool) {
 
 	if isPriorityClass {
 		t.Logf("set the strategy to delete pods with priority lower than priority class %s", highPriorityClass.Name)
-		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, highPriorityClass.Name, nil, false, nil)
+		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, highPriorityClass.Name, nil, false, nil, getPodsAssignedToNode)
 	} else {
 		t.Logf("set the strategy to delete pods with priority lower than %d", highPriority)
-		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", &highPriority, false, nil)
+		runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", &highPriority, false, nil, getPodsAssignedToNode)
 	}
 
 	t.Logf("Waiting 10s")
@@ -721,7 +745,7 @@ func testPriority(t *testing.T, isPriorityClass bool) {
 		t.Fatalf("None of %v high priority pods are expected to be deleted", expectReservePodNames)
 	}
 
-	//check if all pods with low priority class are evicted
+	// check if all pods with low priority class are evicted
 	if err := wait.PollImmediate(time.Second, 30*time.Second, func() (bool, error) {
 		podListLowPriority, err := clientSet.CoreV1().Pods(rcLowPriority.Namespace).List(
 			ctx, metav1.ListOptions{LabelSelector: labels.SelectorFromSet(rcLowPriority.Spec.Template.Labels).String()})
@@ -756,7 +780,7 @@ func testPriority(t *testing.T, isPriorityClass bool) {
 func TestPodLabelSelector(t *testing.T) {
 	ctx := context.Background()
 
-	clientSet, nodeInformer, stopCh := initializeClient(t)
+	clientSet, nodeInformer, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
@@ -804,7 +828,7 @@ func TestPodLabelSelector(t *testing.T) {
 	t.Logf("Pods not expected to be evicted: %v, pods expected to be evicted: %v", expectReservePodNames, expectEvictPodNames)
 
 	t.Logf("set the strategy to delete pods with label test:podlifetime-evict")
-	runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", nil, false, &metav1.LabelSelector{MatchLabels: map[string]string{"test": "podlifetime-evict"}})
+	runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", nil, false, &metav1.LabelSelector{MatchLabels: map[string]string{"test": "podlifetime-evict"}}, getPodsAssignedToNode)
 
 	t.Logf("Waiting 10s")
 	time.Sleep(10 * time.Second)
@@ -824,7 +848,7 @@ func TestPodLabelSelector(t *testing.T) {
 		t.Fatalf("None of %v unevictable pods are expected to be deleted", expectReservePodNames)
 	}
 
-	//check if all selected pods are evicted
+	// check if all selected pods are evicted
 	if err := wait.PollImmediate(time.Second, 30*time.Second, func() (bool, error) {
 		podListEvict, err := clientSet.CoreV1().Pods(rcEvict.Namespace).List(
 			ctx, metav1.ListOptions{LabelSelector: labels.SelectorFromSet(rcEvict.Spec.Template.Labels).String()})
@@ -859,20 +883,9 @@ func TestPodLabelSelector(t *testing.T) {
 func TestEvictAnnotation(t *testing.T) {
 	ctx := context.Background()
 
-	clientSet, nodeInformer, stopCh := initializeClient(t)
+	clientSet, nodeInformer, getPodsAssignedToNode, stopCh := initializeClient(t)
 	defer close(stopCh)
 
-	nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
-	if err != nil {
-		t.Errorf("Error listing node with %v", err)
-	}
-
-	var nodes []*v1.Node
-	for i := range nodeList.Items {
-		node := nodeList.Items[i]
-		nodes = append(nodes, &node)
-	}
-
 	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
 	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace, metav1.CreateOptions{}); err != nil {
 		t.Fatalf("Unable to create ns %v", testNamespace.Name)
@@ -887,7 +900,8 @@ func TestEvictAnnotation(t *testing.T) {
 			Name: "sample",
 			VolumeSource: v1.VolumeSource{
 				EmptyDir: &v1.EmptyDirVolumeSource{
-					SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI)},
+					SizeLimit: resource.NewQuantity(int64(10), resource.BinarySI),
+				},
 			},
 		},
 	}
@@ -914,7 +928,7 @@ func TestEvictAnnotation(t *testing.T) {
 	t.Logf("Existing pods: %v", initialPodNames)
 
 	t.Log("Running PodLifetime strategy")
-	runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", nil, false, nil)
+	runPodLifetimeStrategy(ctx, t, clientSet, nodeInformer, nil, "", nil, false, nil, getPodsAssignedToNode)
 
 	if err := wait.PollImmediate(5*time.Second, time.Minute, func() (bool, error) {
 		podList, err = clientSet.CoreV1().Pods(rc.Namespace).List(ctx, metav1.ListOptions{LabelSelector: labels.SelectorFromSet(rc.Spec.Template.Labels).String()})
@@ -952,7 +966,7 @@ func TestDeschedulingInterval(t *testing.T) {
 	}
 	s.Client = clientSet
 
-	deschedulerPolicy := &api.DeschedulerPolicy{}
+	deschedulerPolicy := &deschedulerapi.DeschedulerPolicy{}
 
 	c := make(chan bool, 1)
 	go func() {
@@ -960,9 +974,7 @@ func TestDeschedulingInterval(t *testing.T) {
 		if err != nil || len(evictionPolicyGroupVersion) == 0 {
 			t.Errorf("Error when checking support for eviction: %v", err)
 		}
-
-		stopChannel := make(chan struct{})
-		if err := descheduler.RunDeschedulerStrategies(ctx, s, deschedulerPolicy, evictionPolicyGroupVersion, stopChannel); err != nil {
+		if err := descheduler.RunDeschedulerStrategies(ctx, s, deschedulerPolicy, evictionPolicyGroupVersion); err != nil {
 			t.Errorf("Error running descheduler strategies: %+v", err)
 		}
 		c <- true
@@ -1121,8 +1133,8 @@ func createBalancedPodForNodes(
 
 	// find the max, if the node has the max,use the one, if not,use the ratio parameter
 	var maxCPUFraction, maxMemFraction float64 = ratio, ratio
-	var cpuFractionMap = make(map[string]float64)
-	var memFractionMap = make(map[string]float64)
+	cpuFractionMap := make(map[string]float64)
+	memFractionMap := make(map[string]float64)
 
 	for _, node := range nodes {
 		cpuFraction, memFraction, _, _ := computeCPUMemFraction(t, cs, node, podRequestedResource)
@@ -1159,7 +1171,7 @@ func createBalancedPodForNodes(
 		// add crioMinMemLimit to ensure that all pods are setting at least that much for a limit, while keeping the same ratios
 		needCreateResource[v1.ResourceMemory] = *resource.NewQuantity(int64((ratio-memFraction)*float64(memAllocatableVal)+float64(crioMinMemLimit)), resource.BinarySI)
 
-		var gracePeriod = int64(1)
+		gracePeriod := int64(1)
 		// Don't set OwnerReferences to avoid pod eviction
 		pod := &v1.Pod{
 			ObjectMeta: metav1.ObjectMeta{
@@ -1288,3 +1300,61 @@ func waitForPodRunning(ctx context.Context, t *testing.T, clientSet clientset.In
 		t.Fatalf("Error waiting for pod running: %v", err)
 	}
 }
+
+func waitForPodsRunning(ctx context.Context, t *testing.T, clientSet clientset.Interface, labelMap map[string]string, desireRunningPodNum int, namespace string) {
+	if err := wait.PollImmediate(10*time.Second, 60*time.Second, func() (bool, error) {
+		podList, err := clientSet.CoreV1().Pods(namespace).List(ctx, metav1.ListOptions{
+			LabelSelector: labels.SelectorFromSet(labelMap).String(),
+		})
+		if err != nil {
+			return false, err
+		}
+		if len(podList.Items) != desireRunningPodNum {
+			t.Logf("Waiting for %v pods to be running, got %v instead", desireRunningPodNum, len(podList.Items))
+			return false, nil
+		}
+		for _, pod := range podList.Items {
+			if pod.Status.Phase != v1.PodRunning {
+				t.Logf("Pod %v not running yet, is %v instead", pod.Name, pod.Status.Phase)
+				return false, nil
+			}
+		}
+		return true, nil
+	}); err != nil {
+		t.Fatalf("Error waiting for pods running: %v", err)
+	}
+}
+
+func splitNodesAndWorkerNodes(nodes []v1.Node) ([]*v1.Node, []*v1.Node) {
+	var allNodes []*v1.Node
+	var workerNodes []*v1.Node
+	for i := range nodes {
+		node := nodes[i]
+		allNodes = append(allNodes, &node)
+		if _, exists := node.Labels["node-role.kubernetes.io/control-plane"]; !exists {
+			workerNodes = append(workerNodes, &node)
+		}
+	}
+	return allNodes, workerNodes
+}
+
+func initPodEvictorOrFail(t *testing.T, clientSet clientset.Interface, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc, nodes []*v1.Node) *evictions.PodEvictor {
+	evictionPolicyGroupVersion, err := eutils.SupportEviction(clientSet)
+	if err != nil || len(evictionPolicyGroupVersion) == 0 {
+		t.Fatalf("Error creating eviction policy group: %v", err)
+	}
+	return evictions.NewPodEvictor(
+		clientSet,
+		evictionPolicyGroupVersion,
+		false,
+		nil,
+		nil,
+		nodes,
+		getPodsAssignedToNode,
+		true,
+		false,
+		false,
+		false,
+		false,
+	)
+}

test/e2e/e2e_toomanyrestarts_test.go

@@ -0,0 +1,208 @@
+/*
+Copyright 2021 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package e2e
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"testing"
+	"time"
+
+	appsv1 "k8s.io/api/apps/v1"
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	clientset "k8s.io/client-go/kubernetes"
+	deschedulerapi "sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
+	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies"
+)
+
+func TestTooManyRestarts(t *testing.T) {
+	ctx := context.Background()
+
+	clientSet, _, getPodsAssignedToNode, stopCh := initializeClient(t)
+	defer close(stopCh)
+
+	nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
+	if err != nil {
+		t.Errorf("Error listing node with %v", err)
+	}
+
+	nodes, workerNodes := splitNodesAndWorkerNodes(nodeList.Items)
+
+	t.Logf("Creating testing namespace %v", t.Name())
+	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
+	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("Unable to create ns %v", testNamespace.Name)
+	}
+	defer clientSet.CoreV1().Namespaces().Delete(ctx, testNamespace.Name, metav1.DeleteOptions{})
+
+	deploymentObj := &appsv1.Deployment{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "restart-pod",
+			Namespace: testNamespace.Name,
+			Labels:    map[string]string{"test": "restart-pod", "name": "test-toomanyrestarts"},
+		},
+		Spec: appsv1.DeploymentSpec{
+			Replicas: func(i int32) *int32 { return &i }(4),
+			Selector: &metav1.LabelSelector{
+				MatchLabels: map[string]string{"test": "restart-pod", "name": "test-toomanyrestarts"},
+			},
+			Template: v1.PodTemplateSpec{
+				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{"test": "restart-pod", "name": "test-toomanyrestarts"},
+				},
+				Spec: v1.PodSpec{
+					Containers: []v1.Container{{
+						Name:            "pause",
+						ImagePullPolicy: "Always",
+						Image:           "kubernetes/pause",
+						Command:         []string{"/bin/sh"},
+						Args:            []string{"-c", "sleep 1s && exit 1"},
+						Ports:           []v1.ContainerPort{{ContainerPort: 80}},
+					}},
+				},
+			},
+		},
+	}
+
+	t.Logf("Creating deployment %v", deploymentObj.Name)
+	_, err = clientSet.AppsV1().Deployments(deploymentObj.Namespace).Create(ctx, deploymentObj, metav1.CreateOptions{})
+	if err != nil {
+		t.Logf("Error creating deployment: %v", err)
+		if err = clientSet.AppsV1().Deployments(deploymentObj.Namespace).DeleteCollection(ctx, metav1.DeleteOptions{}, metav1.ListOptions{
+			LabelSelector: labels.SelectorFromSet(labels.Set(map[string]string{"test": "restart-pod", "name": "test-toomanyrestarts"})).String(),
+		}); err != nil {
+			t.Fatalf("Unable to delete deployment: %v", err)
+		}
+		return
+	}
+	defer clientSet.AppsV1().Deployments(deploymentObj.Namespace).Delete(ctx, deploymentObj.Name, metav1.DeleteOptions{})
+
+	// Need to wait restartCount more than 4
+	result, err := waitPodRestartCount(ctx, clientSet, testNamespace.Name, t)
+	if err != nil {
+		t.Fatalf("Unexpected error: %v", err)
+	}
+	if !result {
+		t.Fatal("Pod restart count not as expected")
+	}
+
+	tests := []struct {
+		name                    string
+		podRestartThreshold     int32
+		includingInitContainers bool
+		expectedEvictedPodCount uint
+	}{
+		{
+			name:                    "test-no-evictions",
+			podRestartThreshold:     int32(10000),
+			includingInitContainers: true,
+			expectedEvictedPodCount: 0,
+		},
+		{
+			name:                    "test-one-evictions",
+			podRestartThreshold:     int32(4),
+			includingInitContainers: true,
+			expectedEvictedPodCount: 4,
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			evictionPolicyGroupVersion, err := eutils.SupportEviction(clientSet)
+			if err != nil || len(evictionPolicyGroupVersion) == 0 {
+				t.Fatalf("Error creating eviction policy group: %v", err)
+			}
+			podEvictor := evictions.NewPodEvictor(
+				clientSet,
+				evictionPolicyGroupVersion,
+				false,
+				nil,
+				nil,
+				nodes,
+				getPodsAssignedToNode,
+				true,
+				false,
+				false,
+				false,
+				false,
+			)
+			// Run RemovePodsHavingTooManyRestarts strategy
+			t.Log("Running RemovePodsHavingTooManyRestarts strategy")
+			strategies.RemovePodsHavingTooManyRestarts(
+				ctx,
+				clientSet,
+				deschedulerapi.DeschedulerStrategy{
+					Enabled: true,
+					Params: &deschedulerapi.StrategyParameters{
+						PodsHavingTooManyRestarts: &deschedulerapi.PodsHavingTooManyRestarts{
+							PodRestartThreshold:     tc.podRestartThreshold,
+							IncludingInitContainers: tc.includingInitContainers,
+						},
+					},
+				},
+				workerNodes,
+				podEvictor,
+				getPodsAssignedToNode,
+			)
+
+			waitForTerminatingPodsToDisappear(ctx, t, clientSet, testNamespace.Name)
+			actualEvictedPodCount := podEvictor.TotalEvicted()
+			if actualEvictedPodCount < tc.expectedEvictedPodCount {
+				t.Errorf("Test error for description: %s. Unexpected number of pods have been evicted, got %v, expected %v", tc.name, actualEvictedPodCount, tc.expectedEvictedPodCount)
+			}
+		})
+	}
+}
+
+func waitPodRestartCount(ctx context.Context, clientSet clientset.Interface, namespace string, t *testing.T) (bool, error) {
+	timeout := time.After(5 * time.Minute)
+	tick := time.Tick(5 * time.Second)
+	for {
+		select {
+		case <-timeout:
+			t.Log("Timeout, still restart count not as expected")
+			return false, fmt.Errorf("timeout Error")
+		case <-tick:
+			podList, err := clientSet.CoreV1().Pods(namespace).List(ctx, metav1.ListOptions{
+				LabelSelector: labels.SelectorFromSet(labels.Set(map[string]string{"test": "restart-pod", "name": "test-toomanyrestarts"})).String(),
+			})
+			if err != nil {
+				t.Fatalf("Unexpected err: %v", err)
+				return false, err
+			}
+			if len(podList.Items) < 4 {
+				t.Log("Waiting for 4 pods")
+				return false, nil
+			}
+			for i := 0; i < 4; i++ {
+				if len(podList.Items[0].Status.ContainerStatuses) < 1 {
+					t.Logf("Waiting for podList.Items[%v].Status.ContainerStatuses to be populated", i)
+					return false, nil
+				}
+			}
+
+			if podList.Items[0].Status.ContainerStatuses[0].RestartCount >= 4 && podList.Items[1].Status.ContainerStatuses[0].RestartCount >= 4 && podList.Items[2].Status.ContainerStatuses[0].RestartCount >= 4 && podList.Items[3].Status.ContainerStatuses[0].RestartCount >= 4 {
+				t.Log("Pod restartCount as expected")
+				return true, nil
+			}
+		}
+	}
+}

test/e2e/e2e_topologyspreadconstraint_test.go

@@ -0,0 +1,159 @@
+package e2e
+
+import (
+	"context"
+	"fmt"
+	"math"
+	"strings"
+	"testing"
+
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+	deschedulerapi "sigs.k8s.io/descheduler/pkg/api"
+	"sigs.k8s.io/descheduler/pkg/descheduler/strategies"
+)
+
+const zoneTopologyKey string = "topology.kubernetes.io/zone"
+
+func TestTopologySpreadConstraint(t *testing.T) {
+	ctx := context.Background()
+	clientSet, _, getPodsAssignedToNode, stopCh := initializeClient(t)
+	defer close(stopCh)
+	nodeList, err := clientSet.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
+	if err != nil {
+		t.Errorf("Error listing node with %v", err)
+	}
+	nodes, workerNodes := splitNodesAndWorkerNodes(nodeList.Items)
+	t.Log("Creating testing namespace")
+	testNamespace := &v1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "e2e-" + strings.ToLower(t.Name())}}
+	if _, err := clientSet.CoreV1().Namespaces().Create(ctx, testNamespace, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("Unable to create ns %v", testNamespace.Name)
+	}
+	defer clientSet.CoreV1().Namespaces().Delete(ctx, testNamespace.Name, metav1.DeleteOptions{})
+
+	testCases := map[string]struct {
+		replicaCount int
+		maxSkew      int
+		labelKey     string
+		labelValue   string
+		constraint   v1.UnsatisfiableConstraintAction
+	}{
+		"test-rc-topology-spread-hard-constraint": {
+			replicaCount: 4,
+			maxSkew:      1,
+			labelKey:     "test",
+			labelValue:   "topology-spread-hard-constraint",
+			constraint:   v1.DoNotSchedule,
+		},
+		"test-rc-topology-spread-soft-constraint": {
+			replicaCount: 4,
+			maxSkew:      1,
+			labelKey:     "test",
+			labelValue:   "topology-spread-soft-constraint",
+			constraint:   v1.ScheduleAnyway,
+		},
+	}
+	for name, tc := range testCases {
+		t.Run(name, func(t *testing.T) {
+			t.Logf("Creating RC %s with %d replicas", name, tc.replicaCount)
+			rc := RcByNameContainer(name, testNamespace.Name, int32(tc.replicaCount), map[string]string{tc.labelKey: tc.labelValue}, nil, "")
+			rc.Spec.Template.Spec.TopologySpreadConstraints = makeTopologySpreadConstraints(tc.maxSkew, tc.labelKey, tc.labelValue, tc.constraint)
+			if _, err := clientSet.CoreV1().ReplicationControllers(rc.Namespace).Create(ctx, rc, metav1.CreateOptions{}); err != nil {
+				t.Fatalf("Error creating RC %s %v", name, err)
+			}
+			defer deleteRC(ctx, t, clientSet, rc)
+			waitForRCPodsRunning(ctx, t, clientSet, rc)
+
+			// Create a "Violator" RC that has the same label and is forced to be on the same node using a nodeSelector
+			violatorRcName := name + "-violator"
+			violatorCount := tc.maxSkew + 1
+			violatorRc := RcByNameContainer(violatorRcName, testNamespace.Name, int32(violatorCount), map[string]string{tc.labelKey: tc.labelValue}, nil, "")
+			violatorRc.Spec.Template.Spec.NodeSelector = map[string]string{zoneTopologyKey: workerNodes[0].Labels[zoneTopologyKey]}
+			rc.Spec.Template.Spec.TopologySpreadConstraints = makeTopologySpreadConstraints(tc.maxSkew, tc.labelKey, tc.labelValue, tc.constraint)
+			if _, err := clientSet.CoreV1().ReplicationControllers(rc.Namespace).Create(ctx, violatorRc, metav1.CreateOptions{}); err != nil {
+				t.Fatalf("Error creating RC %s: %v", violatorRcName, err)
+			}
+			defer deleteRC(ctx, t, clientSet, violatorRc)
+			waitForRCPodsRunning(ctx, t, clientSet, violatorRc)
+
+			podEvictor := initPodEvictorOrFail(t, clientSet, getPodsAssignedToNode, nodes)
+
+			// Run TopologySpreadConstraint strategy
+			t.Logf("Running RemovePodsViolatingTopologySpreadConstraint strategy for %s", name)
+			strategies.RemovePodsViolatingTopologySpreadConstraint(
+				ctx,
+				clientSet,
+				deschedulerapi.DeschedulerStrategy{
+					Enabled: true,
+					Params: &deschedulerapi.StrategyParameters{
+						IncludeSoftConstraints: tc.constraint != v1.DoNotSchedule,
+					},
+				},
+				nodes,
+				podEvictor,
+				getPodsAssignedToNode,
+			)
+			t.Logf("Finished RemovePodsViolatingTopologySpreadConstraint strategy for %s", name)
+
+			t.Logf("Wait for terminating pods of %s to disappear", name)
+			waitForTerminatingPodsToDisappear(ctx, t, clientSet, rc.Namespace)
+
+			if totalEvicted := podEvictor.TotalEvicted(); totalEvicted > 0 {
+				t.Logf("Total of %d Pods were evicted for %s", totalEvicted, name)
+			} else {
+				t.Fatalf("Pods were not evicted for %s TopologySpreadConstraint", name)
+			}
+
+			// Ensure recently evicted Pod are rescheduled and running before asserting for a balanced topology spread
+			waitForRCPodsRunning(ctx, t, clientSet, rc)
+
+			pods, err := clientSet.CoreV1().Pods(testNamespace.Name).List(ctx, metav1.ListOptions{LabelSelector: fmt.Sprintf("%s=%s", tc.labelKey, tc.labelValue)})
+			if err != nil {
+				t.Errorf("Error listing pods for %s: %v", name, err)
+			}
+
+			nodePodCountMap := make(map[string]int)
+			for _, pod := range pods.Items {
+				nodePodCountMap[pod.Spec.NodeName]++
+			}
+
+			if len(nodePodCountMap) != len(workerNodes) {
+				t.Errorf("%s Pods were scheduled on only '%d' nodes and were not properly distributed on the nodes", name, len(nodePodCountMap))
+			}
+
+			min, max := getMinAndMaxPodDistribution(nodePodCountMap)
+			if max-min > tc.maxSkew {
+				t.Errorf("Pod distribution for %s is still violating the max skew of %d as it is %d", name, tc.maxSkew, max-min)
+			}
+
+			t.Logf("Pods for %s were distributed in line with max skew of %d", name, tc.maxSkew)
+		})
+	}
+}
+
+func makeTopologySpreadConstraints(maxSkew int, labelKey, labelValue string, constraint v1.UnsatisfiableConstraintAction) []v1.TopologySpreadConstraint {
+	return []v1.TopologySpreadConstraint{
+		{
+			MaxSkew:           int32(maxSkew),
+			TopologyKey:       zoneTopologyKey,
+			WhenUnsatisfiable: constraint,
+			LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{labelKey: labelValue}},
+		},
+	}
+}
+
+func getMinAndMaxPodDistribution(nodePodCountMap map[string]int) (int, int) {
+	min := math.MaxInt32
+	max := math.MinInt32
+	for _, podCount := range nodePodCountMap {
+		if podCount < min {
+			min = podCount
+		}
+		if podCount > max {
+			max = podCount
+		}
+	}
+
+	return min, max
+}