Merge pull request #874 from kubernetes-sigs/revert-869-existing-contraints-backport-1.23

Revert "Existing contraints fix backport 1.23"

.gitignore

@@ -3,3 +3,5 @@ _tmp/
 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:1.15.5
+FROM golang:1.17.7
 
 WORKDIR /go/src/sigs.k8s.io/descheduler
 COPY . .
@@ -21,7 +21,9 @@ RUN VERSION=${VERSION} make build.$ARCH
 
 FROM scratch
 
-MAINTAINER Avesh Agarwal <avesh.ncsu@gmail.com>
+MAINTAINER Kubernetes SIG Scheduling <kubernetes-sig-scheduling@googlegroups.com>
+
+USER 1000
 
 COPY --from=0 /go/src/sigs.k8s.io/descheduler/_output/bin/descheduler /bin/descheduler
 

Dockerfile.dev

@@ -13,7 +13,9 @@
 # limitations under the License.
 FROM scratch
 
-MAINTAINER Avesh Agarwal <avesh.ncsu@gmail.com>
+MAINTAINER Kubernetes SIG Scheduling <kubernetes-sig-scheduling@googlegroups.com>
+
+USER 1000
 
 COPY _output/bin/descheduler /bin/descheduler
 

Makefile

@@ -16,14 +16,16 @@
 
 # VERSION is based on a date stamp plus the last commit
 VERSION?=v$(shell date +%Y%m%d)-$(shell git describe --tags --match "v*")
+BRANCH?=$(shell git branch --show-current)
+SHA1?=$(shell git rev-parse HEAD)
 BUILD=$(shell date +%FT%T%z)
-LDFLAG_LOCATION=sigs.k8s.io/descheduler/cmd/descheduler/app
-ARCHS = amd64 arm64
+LDFLAG_LOCATION=sigs.k8s.io/descheduler/pkg/version
+ARCHS = amd64 arm arm64
 
-LDFLAGS=-ldflags "-X ${LDFLAG_LOCATION}.version=${VERSION} -X ${LDFLAG_LOCATION}.buildDate=${BUILD}"
+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 := v1.43.0
+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
@@ -41,7 +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
 
@@ -51,6 +53,9 @@ build:
 build.amd64:
 	CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build ${LDFLAGS} -o _output/bin/descheduler sigs.k8s.io/descheduler/cmd/descheduler
 
+build.arm:
+	CGO_ENABLED=0 GOOS=linux GOARCH=arm GOARM=7 go build ${LDFLAGS} -o _output/bin/descheduler sigs.k8s.io/descheduler/cmd/descheduler
+
 build.arm64:
 	CGO_ENABLED=0 GOOS=linux GOARCH=arm64 go build ${LDFLAGS} -o _output/bin/descheduler sigs.k8s.io/descheduler/cmd/descheduler
 
@@ -63,6 +68,9 @@ image:
 image.amd64:
 	docker build --build-arg VERSION="$(VERSION)" --build-arg ARCH="amd64" -t $(IMAGE)-amd64 .
 
+image.arm:
+	docker build --build-arg VERSION="$(VERSION)" --build-arg ARCH="arm" -t $(IMAGE)-arm .
+
 image.arm64:
 	docker build --build-arg VERSION="$(VERSION)" --build-arg ARCH="arm64" -t $(IMAGE)-arm64 .
 
@@ -71,7 +79,7 @@ push: image
 	docker tag $(IMAGE) $(IMAGE_GCLOUD)
 	docker push $(IMAGE_GCLOUD)
 
-push-all: image.amd64 image.arm64
+push-all: image.amd64 image.arm image.arm64
 	gcloud auth configure-docker
 	for arch in $(ARCHS); do \
 		docker tag $(IMAGE)-$${arch} $(IMAGE_GCLOUD)-$${arch} ;\
@@ -85,8 +93,12 @@ push-all: image.amd64 image.arm64
 
 clean:
 	rm -rf _output
+	rm -rf _tmp
 
-verify: verify-gofmt verify-vendor lint lint-chart verify-spelling
+verify: verify-govet verify-spelling verify-gofmt verify-vendor lint lint-chart verify-toc verify-gen
+
+verify-govet:
+	./hack/verify-govet.sh
 
 verify-spelling:
 	./hack/verify-spelling.sh
@@ -97,6 +109,9 @@ verify-gofmt:
 verify-vendor:
 	./hack/verify-vendor.sh
 
+verify-toc:
+	./hack/verify-toc.sh
+
 test-unit:
 	./test/run-unit-tests.sh
 
@@ -107,15 +122,26 @@ gen:
 	./hack/update-generated-conversions.sh
 	./hack/update-generated-deep-copies.sh
 	./hack/update-generated-defaulters.sh
+	./hack/update-toc.sh
+
+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
+endif

OWNERS

@@ -1,8 +1,7 @@
 approvers:
-- aveshagarwal
-- k82cn
-- ravisantoshgudimetla
 - damemi
+- ingvagabund
+- seanmalloy
 reviewers:
 - aveshagarwal
 - k82cn
@@ -10,3 +9,8 @@ reviewers:
 - damemi
 - seanmalloy
 - ingvagabund
+- lixiang233
+emeritus_approvers:
+- aveshagarwal
+- k82cn
+- ravisantoshgudimetla

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
@@ -24,36 +24,43 @@ but relies on the default scheduler for that.
 
 Table of Contents
 =================
-
-  * [Quick Start](#quick-start)
-     * [Run As A Job](#run-as-a-job)
-     * [Run As A CronJob](#run-as-a-cronjob)
-     * [Install Using Helm](#install-using-helm)
-     * [Install Using Kustomize](#install-using-kustomize)
-  * [User Guide](#user-guide)
-  * [Policy and Strategies](#policy-and-strategies)
-     * [RemoveDuplicates](#removeduplicates)
-     * [LowNodeUtilization](#lownodeutilization)
-     * [RemovePodsViolatingInterPodAntiAffinity](#removepodsviolatinginterpodantiaffinity)
-     * [RemovePodsViolatingNodeAffinity](#removepodsviolatingnodeaffinity)
-     * [RemovePodsViolatingNodeTaints](#removepodsviolatingnodetaints)
-     * [RemovePodsViolatingTopologySpreadConstraint](#removepodsviolatingtopologyspreadconstraint)
-     * [RemovePodsHavingTooManyRestarts](#removepodshavingtoomanyrestarts)
-     * [PodLifeTime](#podlifetime)
-  * [Filter Pods](#filter-pods)
-     * [Namespace filtering](#namespace-filtering)
-     * [Priority filtering](#priority-filtering)
-  * [Pod Evictions](#pod-evictions)
-     * [Pod Disruption Budget (PDB)](#pod-disruption-budget-pdb)
-  * [Compatibility Matrix](#compatibility-matrix)
-  * [Getting Involved and Contributing](#getting-involved-and-contributing)
-     * [Communicating With Contributors](#communicating-with-contributors)
-  * [Roadmap](#roadmap)
-     * [Code of conduct](#code-of-conduct)
+<!-- toc -->
+- [Quick Start](#quick-start)
+  - [Run As A Job](#run-as-a-job)
+  - [Run As A CronJob](#run-as-a-cronjob)
+  - [Run As A Deployment](#run-as-a-deployment)
+  - [Install Using Helm](#install-using-helm)
+  - [Install Using Kustomize](#install-using-kustomize)
+- [User Guide](#user-guide)
+- [Policy and Strategies](#policy-and-strategies)
+  - [RemoveDuplicates](#removeduplicates)
+  - [LowNodeUtilization](#lownodeutilization)
+  - [HighNodeUtilization](#highnodeutilization)
+  - [RemovePodsViolatingInterPodAntiAffinity](#removepodsviolatinginterpodantiaffinity)
+  - [RemovePodsViolatingNodeAffinity](#removepodsviolatingnodeaffinity)
+  - [RemovePodsViolatingNodeTaints](#removepodsviolatingnodetaints)
+  - [RemovePodsViolatingTopologySpreadConstraint](#removepodsviolatingtopologyspreadconstraint)
+  - [RemovePodsHavingTooManyRestarts](#removepodshavingtoomanyrestarts)
+  - [PodLifeTime](#podlifetime)
+  - [RemoveFailedPods](#removefailedpods)
+- [Filter Pods](#filter-pods)
+  - [Namespace filtering](#namespace-filtering)
+  - [Priority filtering](#priority-filtering)
+  - [Label filtering](#label-filtering)
+  - [Node Fit filtering](#node-fit-filtering)
+- [Pod Evictions](#pod-evictions)
+  - [Pod Disruption Budget (PDB)](#pod-disruption-budget-pdb)
+- [Metrics](#metrics)
+- [Compatibility Matrix](#compatibility-matrix)
+- [Getting Involved and Contributing](#getting-involved-and-contributing)
+  - [Communicating With Contributors](#communicating-with-contributors)
+- [Roadmap](#roadmap)
+  - [Code of conduct](#code-of-conduct)
+<!-- /toc -->
 
 ## Quick Start
 
-The descheduler can be run as a Job or CronJob inside of a k8s cluster. It has the
+The descheduler can be run as a `Job`, `CronJob`, or `Deployment` inside of a k8s cluster. It has the
 advantage of being able to be run multiple times without needing user intervention.
 The descheduler pod is run as a critical pod in the `kube-system` namespace to avoid
 being evicted by itself or by the kubelet.
@@ -74,6 +81,14 @@ kubectl create -f kubernetes/base/configmap.yaml
 kubectl create -f kubernetes/cronjob/cronjob.yaml
 ```
 
+### Run As A Deployment
+
+```
+kubectl create -f kubernetes/base/rbac.yaml
+kubectl create -f kubernetes/base/configmap.yaml
+kubectl create -f kubernetes/deployment/deployment.yaml
+```
+
 ### Install Using Helm
 
 Starting with release v0.18.0 there is an official helm chart that can be used to install the
@@ -84,16 +99,21 @@ The descheduler helm chart is also listed on the [artifact hub](https://artifact
 ### Install Using Kustomize
 
 You can use kustomize to install descheduler.
-See the [resources | Kustomize](https://kubernetes-sigs.github.io/kustomize/api-reference/kustomization/resources/) for detailed instructions.
+See the [resources | Kustomize](https://kubectl.docs.kubernetes.io/references/kustomize/resource/) for detailed instructions.
 
 Run As A Job
 ```
-kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/job?ref=master' | kubectl apply -f -
+kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/job?ref=v0.22.0' | kubectl apply -f -
 ```
 
 Run As A CronJob
 ```
-kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/cronjob?ref=master' | kubectl apply -f -
+kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/cronjob?ref=v0.22.0' | kubectl apply -f -
+```
+
+Run As A Deployment
+```
+kustomize build 'github.com/kubernetes-sigs/descheduler/kubernetes/deployment?ref=v0.22.0' | kubectl apply -f -
 ```
 
 ## User Guide
@@ -102,38 +122,54 @@ 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.
-Eight strategies `RemoveDuplicates`, `LowNodeUtilization`, `RemovePodsViolatingInterPodAntiAffinity`,
-`RemovePodsViolatingNodeAffinity`, `RemovePodsViolatingNodeTaints`, `RemovePodsViolatingTopologySpreadConstraint`,
-`RemovePodsHavingTooManyRestarts`, and `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 policy also includes common configuration for all the strategies:
-- `nodeSelector` - limiting the nodes which are processed
-- `evictLocalStoragePods` - allowing to evict pods with local storage
-- `maxNoOfPodsToEvictPerNode` - maximum number of pods evicted from each node (summed through all strategies)
+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) |
+| `evictFailedBarePods` | `false` | allow eviction of pods without owner references and in failed phase |
+
+As part of the policy, the parameters associated with each strategy can be configured.
+See each strategy for details on available parameters.
+
+**Policy:**
 
 ```yaml
 apiVersion: "descheduler/v1alpha1"
 kind: "DeschedulerPolicy"
 nodeSelector: prod=dev
+evictFailedBarePods: false
 evictLocalStoragePods: true
+evictSystemCriticalPods: true
 maxNoOfPodsToEvictPerNode: 40
+ignorePvcPods: false
 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 Replica Set (RS),
-Replication Controller (RC), Deployment, or Job running on the same node. If there are more,
+This strategy makes sure that there is only one pod associated with a ReplicaSet (RS),
+ReplicationController (RC), StatefulSet, or Job running on the same node. If there are more,
 those duplicate pods are evicted for better spreading of pods in a cluster. This issue could happen
 if some nodes went down due to whatever reasons, and pods on them were moved to other nodes leading to
 more than one pod associated with a RS or RC, for example, running on the same node. Once the failed nodes
 are ready again, this strategy could be enabled to evict those duplicate pods.
 
-It provides one optional parameter, `ExcludeOwnerKinds`, which is a list of OwnerRef `Kind`s. If a pod
-has any of these `Kind`s listed as an `OwnerRef`, that pod will not be considered for eviction.
+It provides one optional parameter, `excludeOwnerKinds`, which is a list of OwnerRef `Kind`s. If a pod
+has any of these `Kind`s listed as an `OwnerRef`, that pod will not be considered for eviction. Note that
+pods created by Deployments are considered for eviction by this strategy. The `excludeOwnerKinds` parameter
+should include `ReplicaSet` to have pods created by Deployments excluded.
 
 **Parameters:**
 
@@ -143,6 +179,7 @@ has any of these `Kind`s listed as an `OwnerRef`, that pod will not be considere
 |`namespaces`|(see [namespace filtering](#namespace-filtering))|
 |`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
 |`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
+|`nodeFit`|bool (see [node fit filtering](#node-fit-filtering))|
 
 **Example:**
 ```yaml
@@ -164,17 +201,22 @@ in the hope that recreation of evicted pods will be scheduled on these underutil
 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, and number of pods in terms of percentage. If a node's
-usage is below threshold for all (cpu, memory, and number of pods), the node is considered underutilized.
+`thresholds` can be configured for cpu, memory, number of pods, and extended resources in terms of percentage (the percentage is
+calculated as the current resources requested on the node vs [total allocatable](https://kubernetes.io/docs/concepts/architecture/nodes/#capacity).
+For pods, this means the number of pods on the node as a fraction of the pod capacity set for that node).
+
+If a node's usage is below threshold for all (cpu, memory, number of pods and extended resources), the node is considered underutilized.
 Currently, pods request resource requirements are considered for computing node resource utilization.
 
 There is another configurable threshold, `targetThresholds`, that is used to compute those potential nodes
-from where pods could be evicted. If a node's usage is above targetThreshold for any (cpu, memory, or number of pods),
+from where pods could be evicted. If a node's usage is above targetThreshold for any (cpu, memory, number of pods, or extended resources),
 the node is considered over utilized. Any node between the thresholds, `thresholds` and `targetThresholds` is
 considered appropriately utilized and is not considered for eviction. The threshold, `targetThresholds`,
 can be configured for cpu, memory, and number of pods too in terms of percentage.
 
-These thresholds, `thresholds` and `targetThresholds`, could be tuned as per your cluster requirements.
+These thresholds, `thresholds` and `targetThresholds`, could be tuned as per your cluster requirements. Note that this
+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.
 
 **Parameters:**
 
@@ -185,6 +227,7 @@ These thresholds, `thresholds` and `targetThresholds`, could be tuned as per you
 |`numberOfNodes`|int|
 |`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
 |`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
+|`nodeFit`|bool (see [node fit filtering](#node-fit-filtering))|
 
 **Example:**
 
@@ -207,19 +250,78 @@ strategies:
 ```
 
 Policy should pass the following validation checks:
-* Only three types of resources are supported: `cpu`, `memory` and `pods`.
+* Three basic native types of resources are supported: `cpu`, `memory` and `pods`.
+If any of these resource types is not specified, all its thresholds default to 100% to avoid nodes going from underutilized to overutilized.
+* Extended resources are supported. For example, resource type `nvidia.com/gpu` is specified for GPU node utilization. Extended resources are optional,
+and will not be used to compute node's usage if it's not specified in `thresholds` and `targetThresholds` explicitly.
 * `thresholds` or `targetThresholds` can not be nil and they must configure exactly the same types of resources.
 * The valid range of the resource's percentage value is \[0, 100\]
 * Percentage value of `thresholds` can not be greater than `targetThresholds` for the same resource.
 
-If any of the resource types is not specified, all its thresholds default to 100% to avoid nodes going
-from underutilized to overutilized.
-
 There is another parameter associated with the `LowNodeUtilization` strategy, called `numberOfNodes`.
 This parameter can be configured to activate the strategy only when the number of under utilized nodes
 are above the configured value. This could be helpful in large clusters where a few nodes could go
 under utilized frequently or for a short period of time. By default, `numberOfNodes` is set to zero.
 
+### HighNodeUtilization
+
+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 strategy `MostRequestedPriority`. 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
+calculated as the current resources requested on the node vs [total allocatable](https://kubernetes.io/docs/concepts/architecture/nodes/#capacity).
+For pods, this means the number of pods on the node as a fraction of the pod capacity set for that node.
+
+If a node's usage is below threshold for all (cpu, memory, number of pods and extended resources), the node is considered underutilized.
+Currently, pods request resource requirements are considered for computing node resource utilization.
+Any node above `thresholds` is considered appropriately utilized and is not considered for eviction.
+
+The `thresholds` param could be tuned as per your cluster requirements. Note that this
+strategy evicts pods from `underutilized nodes` (those with usage below `thresholds`)
+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.
+
+**Parameters:**
+
+|Name|Type|
+|---|---|
+|`thresholds`|map(string:int)|
+|`numberOfNodes`|int|
+|`thresholdPriority`|int (see [priority filtering](#priority-filtering))|
+|`thresholdPriorityClassName`|string (see [priority filtering](#priority-filtering))|
+|`nodeFit`|bool (see [node fit filtering](#node-fit-filtering))|
+
+**Example:**
+
+```yaml
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "HighNodeUtilization":
+     enabled: true
+     params:
+       nodeResourceUtilizationThresholds:
+         thresholds:
+           "cpu" : 20
+           "memory": 20
+           "pods": 20
+```
+
+Policy should pass the following validation checks:
+* Three basic native types of resources are supported: `cpu`, `memory` and `pods`. If any of these resource types is not specified, all its thresholds default to 100%.
+* Extended resources are supported. For example, resource type `nvidia.com/gpu` is specified for GPU node utilization. Extended resources are optional, and will not be used to compute node's usage if it's not specified in `thresholds` explicitly.
+* `thresholds` can not be nil.
+* The valid range of the resource's percentage value is \[0, 100\]
+
+There is another parameter associated with the `HighNodeUtilization` strategy, called `numberOfNodes`.
+This parameter can be configured to activate the strategy only when the number of under utilized nodes
+is above the configured value. This could be helpful in large clusters where a few nodes could go
+under utilized frequently or for a short period of time. By default, `numberOfNodes` is set to zero.
+
 ### RemovePodsViolatingInterPodAntiAffinity
 
 This strategy makes sure that pods violating interpod anti-affinity are removed from nodes. For example,
@@ -235,6 +337,8 @@ node.
 |`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:**
 
@@ -272,6 +376,8 @@ podA gets evicted from nodeA.
 |`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:**
 
@@ -300,6 +406,8 @@ and will be evicted.
 |`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:**
 
@@ -317,13 +425,21 @@ This strategy makes sure that pods violating [topology spread constraints](https
 are evicted from nodes. Specifically, it tries to evict the minimum number of pods required to balance topology domains to within each constraint's `maxSkew`.
 This strategy requires k8s version 1.18 at a minimum.
 
+By default, this strategy only deals with hard constraints, setting parameter `includeSoftConstraints` to `true` will
+include soft constraints.
+
+Strategy parameter `labelSelector` is not utilized when balancing topology domains and is only applied during eviction to determine if the pod can be evicted.
+
 **Parameters:**
 
 |Name|Type|
 |---|---|
+|`includeSoftConstraints`|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:**
 
@@ -333,15 +449,18 @@ kind: "DeschedulerPolicy"
 strategies:
   "RemovePodsViolatingTopologySpreadConstraint":
      enabled: true
+     params:
+       includeSoftConstraints: false
 ```
 
 
 ### RemovePodsHavingTooManyRestarts
 
-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.
+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 (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:**
 
@@ -352,6 +471,8 @@ 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:**
 
@@ -383,6 +504,7 @@ to `Running` and `Pending`.
 |`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))|
 
 **Example:**
 
@@ -399,6 +521,47 @@ 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
@@ -411,6 +574,7 @@ The following strategies accept a `namespaces` parameter which allows to specify
 * `RemovePodsViolatingInterPodAntiAffinity`
 * `RemoveDuplicates`
 * `RemovePodsViolatingTopologySpreadConstraint`
+* `RemoveFailedPods`
 
 For example:
 
@@ -457,6 +621,9 @@ All strategies are able to configure a priority threshold, only pods under the t
 specify this threshold by setting `thresholdPriorityClassName`(setting the threshold to the value of the given
 priority class) or `thresholdPriority`(directly setting the threshold) parameters. By default, this threshold
 is set to the value of `system-cluster-critical` priority class.
+
+Note: Setting `evictSystemCriticalPods` to true disables priority filtering entirely.
+
 E.g.
 
 Setting `thresholdPriority`
@@ -488,20 +655,104 @@ strategies:
 Note that you can't configure both `thresholdPriority` and `thresholdPriorityClassName`, if the given priority class
 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/v1.22/#labelselector-v1-meta)
+to filter pods by their labels:
+
+* `PodLifeTime`
+* `RemovePodsHavingTooManyRestarts`
+* `RemovePodsViolatingNodeTaints`
+* `RemovePodsViolatingNodeAffinity`
+* `RemovePodsViolatingInterPodAntiAffinity`
+* `RemovePodsViolatingTopologySpreadConstraint`
+* `RemoveFailedPods`
+
+This allows running strategies among pods the descheduler is interested in.
+
+For example:
+
+```yaml
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "PodLifeTime":
+    enabled: true
+    params:
+      podLifeTime:
+        maxPodLifeTimeSeconds: 86400
+      labelSelector:
+        matchLabels:
+          component: redis
+        matchExpressions:
+          - {key: tier, operator: In, values: [cache]}
+          - {key: environment, operator: NotIn, values: [dev]}
+```
+
+
+### Node Fit filtering
+
+The following strategies accept a `nodeFit` boolean parameter which can optimize descheduling:
+* `RemoveDuplicates`
+* `LowNodeUtilization`
+* `HighNodeUtilization`
+* `RemovePodsViolatingInterPodAntiAffinity`
+* `RemovePodsViolatingNodeAffinity`
+* `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
+- `nodeAffinity` on the pod
+- Whether any of the other nodes are marked as `unschedulable`
+
+E.g.
+
+```yaml
+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
+        nodeFit: true
+```
+
+Note that node fit filtering references the current pod spec, and not that of it's owner.
+Thus, if the pod is owned by a ReplicationController (and that ReplicationController was modified recently),
+the pod may be running with an outdated spec, which the descheduler will reference when determining node fit.
+This is expected behavior as the descheduler is a "best-effort" mechanism.
+
+Using Deployments instead of ReplicationControllers provides an automated rollout of pod spec changes, therefore ensuring that the descheduler has an up-to-date view of the cluster state.
+
 ## Pod Evictions
 
 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.
-* Pods (static or mirrored pods or stand alone pods) not part of an RC, RS, Deployment or Job are
-never evicted because these pods won't be recreated.
+* [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 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 local storage are never evicted (unless `evictLocalStoragePods: true` is set).
+* Pods with PVCs are evicted (unless `ignorePvcPods: true` is set).
 * In `LowNodeUtilization` and `RemovePodsViolatingInterPodAntiAffinity`, pods are evicted by their priority from low to high, and if they have same priority,
 best effort pods are evicted before burstable and guaranteed pods.
-* All types of pods with the annotation descheduler.alpha.kubernetes.io/evict are evicted. This
+* 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.
 
@@ -510,6 +761,16 @@ 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.
 
+## Metrics
+
+| name	| type	| description |
+|-------|-------|----------------|
+| build_info |	gauge |	constant 1 |
+| pods_evicted | CounterVec | total number of pods evicted |
+
+The metrics are served through https://localhost:10258/metrics by default.
+The address and port can be changed by setting `--binding-address` and `--secure-port` flags.
+
 ## Compatibility Matrix
 The below compatibility matrix shows the k8s client package(client-go, apimachinery, etc) versions that descheduler
 is compiled with. At this time descheduler does not have a hard dependency to a specific k8s release. However a
@@ -521,6 +782,8 @@ packages that it is compiled with.
 
 Descheduler  | Supported Kubernetes Version
 -------------|-----------------------------
+v0.22        | v1.22
+v0.21        | v1.21
 v0.20        | v1.20
 v0.19        | v1.19
 v0.18        | v1.18

charts/descheduler/Chart.yaml

@@ -1,7 +1,7 @@
 apiVersion: v1
 name: descheduler
-version: 0.20.0
-appVersion: 0.20.0
+version: 0.23.2
+appVersion: 0.23.1
 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

@@ -45,21 +45,28 @@ The following table lists the configurable parameters of the _descheduler_ chart
 
 | 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`                                 |
 | `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.cpuRequest`        | Descheduler container CPU request                                                                  | `500m`                               |
-| `resources.memoryRequest`        | Descheduler container memory request                                                                  | `256Mi`                               |
+| `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`                                |
+| `tolerations`                  | tolerations to run the descheduler cronjob on specific nodes                                                          | `nil`                                |
+| `suspend`                      | Set spec.suspend in descheduler cronjob                                                                               | `false`                              |
+| `commonLabels`                 | Labels to apply to all resources                                                                                      | `{}`                                 |

charts/descheduler/templates/NOTES.txt

@@ -1 +1 @@
-Descheduler installed as a cron job.
+Descheduler installed as a {{ .Values.kind }} .

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 -}}
 
 {{/*

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"]

charts/descheduler/templates/cronjob.yaml

@@ -1,4 +1,5 @@
-apiVersion: batch/v1beta1
+{{- if eq .Values.kind "CronJob" }}
+apiVersion: {{ .Values.cronJobApiVersion | default "batch/v1" }}
 kind: CronJob
 metadata:
   name: {{ template "descheduler.fullname" . }}
@@ -6,6 +7,9 @@ metadata:
     {{- 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 }}
@@ -27,17 +31,32 @@ 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 }}
         spec:
+          {{- with .Values.nodeSelector }}
+          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) }}"
@@ -53,8 +72,18 @@ spec:
                 - {{ $value | quote }}
                 {{- end }}
                 {{- end }}
+              livenessProbe:
+                {{- toYaml .Values.livenessProbe | nindent 16 }}
               resources:
                 {{- toYaml .Values.resources | nindent 16 }}
+              securityContext:
+                allowPrivilegeEscalation: false
+                capabilities:
+                  drop:
+                    - ALL
+                privileged: false
+                readOnlyRootFilesystem: true
+                runAsNonRoot: true
               volumeMounts:
                 - mountPath: /policy-dir
                   name: policy-volume
@@ -62,3 +91,4 @@ spec:
           - name: policy-volume
             configMap:
               name: {{ template "descheduler.fullname" . }}
+{{- end }}

charts/descheduler/templates/deployment.yaml

@@ -0,0 +1,81 @@
+{{- if eq .Values.kind "Deployment" }}
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ template "descheduler.fullname" . }}
+  labels:
+    {{- include "descheduler.labels" . | nindent 4 }}
+spec:
+  replicas: 1
+  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" . }}
+      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 }}
+          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/serviceaccount.yaml

@@ -5,4 +5,7 @@ metadata:
   name: {{ template "descheduler.serviceAccountName" . }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
+{{- if .Values.serviceAccount.annotations }}
+  annotations: {{ toYaml .Values.serviceAccount.annotations | nindent 4 }}
+{{- end }}
 {{- end -}}

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

@@ -0,0 +1,29 @@
+apiVersion: v1
+kind: Pod
+metadata:
+    name: descheduler-test-pod
+    annotations:
+      "helm.sh/hook": test
+spec:
+    restartPolicy: Never
+    serviceAccountName: descheduler-ci
+    containers:
+      - name: descheduler-test-container
+        image: alpine:latest
+        imagePullPolicy: IfNotPresent
+        securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop:
+                - All
+            privileged: false
+            runAsNonRoot: false 
+        command: ["/bin/ash"]
+        args:
+          - -c
+          - >-
+            apk --no-cache add curl &&
+            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"

charts/descheduler/values.yaml

@@ -2,12 +2,17 @@
 # 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: []
+
 resources:
   requests:
     cpu: 500m
@@ -19,11 +24,19 @@ 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
+
 cmdOptions:
   v: 3
   # evict-local-storage-pods:
@@ -39,8 +52,8 @@ deschedulerPolicy:
     RemovePodsViolatingNodeAffinity:
       enabled: true
       params:
-         nodeAffinityType:
-         - requiredDuringSchedulingIgnoredDuringExecution
+        nodeAffinityType:
+        - requiredDuringSchedulingIgnoredDuringExecution
     RemovePodsViolatingInterPodAntiAffinity:
       enabled: true
     LowNodeUtilization:
@@ -58,6 +71,26 @@ deschedulerPolicy:
 
 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
+
+tolerations: []
+# - key: 'management'
+#   operator: 'Equal'
+#   value: 'tool'
+#   effect: 'NoSchedule'
+
 rbac:
   # Specifies whether RBAC resources should be created
   create: true
@@ -72,3 +105,15 @@ 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
+

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

@@ -19,20 +19,26 @@ package options
 
 import (
 	"github.com/spf13/pflag"
-	utilerrors "k8s.io/apimachinery/pkg/util/errors"
+
+	apiserveroptions "k8s.io/apiserver/pkg/server/options"
 	clientset "k8s.io/client-go/kubernetes"
 
-	"k8s.io/component-base/logs"
 	"sigs.k8s.io/descheduler/pkg/apis/componentconfig"
 	"sigs.k8s.io/descheduler/pkg/apis/componentconfig/v1alpha1"
 	deschedulerscheme "sigs.k8s.io/descheduler/pkg/descheduler/scheme"
 )
 
+const (
+	DefaultDeschedulerPort = 10258
+)
+
 // DeschedulerServer configuration
 type DeschedulerServer struct {
 	componentconfig.DeschedulerConfiguration
-	Client clientset.Interface
-	Logs   *logs.Options
+
+	Client         clientset.Interface
+	SecureServing  *apiserveroptions.SecureServingOptionsWithLoopback
+	DisableMetrics bool
 }
 
 // NewDeschedulerServer creates a new DeschedulerServer with default parameters
@@ -41,19 +47,16 @@ func NewDeschedulerServer() (*DeschedulerServer, error) {
 	if err != nil {
 		return nil, err
 	}
+
+	secureServing := apiserveroptions.NewSecureServingOptions().WithLoopback()
+	secureServing.BindPort = DefaultDeschedulerPort
+
 	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{}
 	deschedulerscheme.Scheme.Default(&versionedCfg)
@@ -66,15 +69,12 @@ func newDefaultComponentConfig() (*componentconfig.DeschedulerConfiguration, err
 
 // AddFlags adds flags for a specific SchedulerServer to the specified FlagSet
 func (rs *DeschedulerServer) AddFlags(fs *pflag.FlagSet) {
-	fs.StringVar(&rs.Logging.Format, "logging-format", rs.Logging.Format, `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.`)
+	fs.StringVar(&rs.Logging.Format, "logging-format", "text", `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.`)
 	fs.DurationVar(&rs.DeschedulingInterval, "descheduling-interval", rs.DeschedulingInterval, "Time interval between two consecutive descheduler executions. Setting this value instructs the descheduler to run in a continuous loop at the interval specified.")
 	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.")
+
+	rs.SecureServing.AddFlags(fs)
 }

cmd/descheduler/app/server.go

@@ -18,15 +18,25 @@ limitations under the License.
 package app
 
 import (
-	"flag"
+	"context"
 	"io"
+	"os/signal"
+	"syscall"
+
+	"k8s.io/apiserver/pkg/server/healthz"
 
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
 	"sigs.k8s.io/descheduler/pkg/descheduler"
 
 	"github.com/spf13/cobra"
 
-	aflag "k8s.io/component-base/cli/flag"
+	apiserver "k8s.io/apiserver/pkg/server"
+	"k8s.io/apiserver/pkg/server/mux"
+	restclient "k8s.io/client-go/rest"
+	"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"
 )
 
@@ -43,26 +53,56 @@ 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
 
-			if err := s.Validate(); err != nil {
-				klog.ErrorS(err, "failed to validate server configuration")
+			// LoopbackClientConfig is a config for a privileged loopback connection
+			var LoopbackClientConfig *restclient.Config
+			var SecureServing *apiserver.SecureServingInfo
+			if err := s.SecureServing.ApplyTo(&SecureServing, &LoopbackClientConfig); err != nil {
+				klog.ErrorS(err, "failed to apply secure server configuration")
+				return
 			}
-			err := Run(s)
+
+			factory, _ := registry.LogRegistry.Get(s.Logging.Format)
+			if factory == nil {
+				klog.ClearLogger()
+			} else {
+				log, logrFlush := factory.Create(config.FormatOptions{})
+				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
+			}
+
+			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/app/version.go

@@ -18,77 +18,19 @@ package app
 
 import (
 	"fmt"
-	"regexp"
-	"runtime"
-	"strings"
 
 	"github.com/spf13/cobra"
+	"sigs.k8s.io/descheduler/pkg/version"
 )
 
-var (
-	// version is a constant representing the version tag that
-	// generated this build. It should be set during build via -ldflags.
-	version string
-	// buildDate in ISO8601 format, output of $(date -u +'%Y-%m-%dT%H:%M:%SZ')
-	//It should be set during build via -ldflags.
-	buildDate string
-)
-
-// Info holds the information related to descheduler app version.
-type Info struct {
-	Major      string `json:"major"`
-	Minor      string `json:"minor"`
-	GitVersion string `json:"gitVersion"`
-	BuildDate  string `json:"buildDate"`
-	GoVersion  string `json:"goVersion"`
-	Compiler   string `json:"compiler"`
-	Platform   string `json:"platform"`
-}
-
-// Get returns the overall codebase version. It's for detecting
-// what code a binary was built from.
-func Get() Info {
-	majorVersion, minorVersion := splitVersion(version)
-	return Info{
-		Major:      majorVersion,
-		Minor:      minorVersion,
-		GitVersion: version,
-		BuildDate:  buildDate,
-		GoVersion:  runtime.Version(),
-		Compiler:   runtime.Compiler,
-		Platform:   fmt.Sprintf("%s/%s", runtime.GOOS, runtime.GOARCH),
-	}
-}
-
 func NewVersionCommand() *cobra.Command {
 	var versionCmd = &cobra.Command{
 		Use:   "version",
 		Short: "Version of descheduler",
 		Long:  `Prints the version of descheduler.`,
 		Run: func(cmd *cobra.Command, args []string) {
-			fmt.Printf("Descheduler version %+v\n", Get())
+			fmt.Printf("Descheduler version %+v\n", version.Get())
 		},
 	}
 	return versionCmd
 }
-
-// splitVersion splits the git version to generate major and minor versions needed.
-func splitVersion(version string) (string, string) {
-	if version == "" {
-		return "", ""
-	}
-
-	// Version from an automated container build environment for a tag. For example v20200521-v0.18.0.
-	m1, _ := regexp.MatchString(`^v\d{8}-v\d+\.\d+\.\d+$`, version)
-
-	// Version from an automated container build environment(not a tag) or a local build. For example v20201009-v0.18.0-46-g939c1c0.
-	m2, _ := regexp.MatchString(`^v\d{8}-v\d+\.\d+\.\d+-\w+-\w+$`, version)
-
-	if m1 || m2 {
-		semVer := strings.Split(version, "-")[1]
-		return strings.Trim(strings.Split(semVer, ".")[0], "v"), strings.Split(semVer, ".")[1] + "+"
-	}
-
-	// Something went wrong
-	return "", ""
-}

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/contributor-guide.md

@@ -3,10 +3,10 @@
 ## Required Tools
 
 - [Git](https://git-scm.com/downloads)
-- [Go 1.15+](https://golang.org/dl/)
+- [Go 1.16+](https://golang.org/dl/)
 - [Docker](https://docs.docker.com/install/)
 - [kubectl](https://kubernetes.io/docs/tasks/tools/install-kubectl)
-- [kind v0.9.0+](https://kind.sigs.k8s.io/)
+- [kind v0.10.0+](https://kind.sigs.k8s.io/)
 
 ## Build and Run
 
@@ -39,5 +39,18 @@ make test-unit
 make test-e2e
 ```
 
+## Run Helm Tests
+Run the helm test for a particular descheduler release by setting below variables,
+```
+HELM_IMAGE_REPO="descheduler"
+HELM_IMAGE_TAG="helm-test"
+HELM_CHART_LOCATION="./charts/descheduler"
+```
+The helm tests runs as part of descheduler CI. But, to run it manually from the descheduler root,
+
+```
+make test-helm
+```
+
 ### Miscellaneous
 See the [hack directory](https://github.com/kubernetes-sigs/descheduler/tree/master/hack) for additional tools and scripts used for developing the descheduler.

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

@@ -1,11 +1,19 @@
 # User Guide
 
 Starting with descheduler release v0.10.0 container images are available in the official k8s container registry.
-* `k8s.gcr.io/descheduler/descheduler`
 
-Also, starting with descheduler release v0.20.0 multi-arch container images are provided. Currently AMD64 and ARM64
-container images are provided. 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
+Descheduler Version | Container Image                            | Architectures           |
+------------------- |--------------------------------------------|-------------------------|
+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
 image into a kind cluster.
 ```
 kind create cluster
@@ -86,12 +94,12 @@ strategies:
 ```
 
 ### Balance Cluster By Node Memory Utilization
-
 If your cluster has been running for a long period of time, you may find that the resource utilization is not very
-balanced. The `LowNodeUtilization` strategy can be used to rebalance your cluster based on `cpu`, `memory`
+balanced. The following two strategies can be used to rebalance your cluster based on `cpu`, `memory` 
 or `number of pods`.
 
-Using the following policy configuration file, descheduler will rebalance the cluster based on memory by evicting pods
+#### Balance high utilization nodes
+Using `LowNodeUtilization`, descheduler will rebalance the cluster based on memory by evicting pods
 from nodes with memory utilization over 70% to nodes with memory utilization below 20%.
 
 ```
@@ -108,6 +116,23 @@ strategies:
           "memory": 70
 ```
 
+#### 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`.
+The evicted pods will be compacted into minimal set of nodes.
+
+```
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "HighNodeUtilization":
+    enabled: true
+    params:
+      nodeResourceUtilizationThresholds:
+        thresholds:
+          "memory": 20
+```
+
 ### Autoheal Node Problems
 Descheduler's `RemovePodsViolatingNodeTaints` strategy can be combined with
 [Node Problem Detector](https://github.com/kubernetes/node-problem-detector/) and

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

@@ -0,0 +1,9 @@
+apiVersion: "descheduler/v1alpha1"
+kind: "DeschedulerPolicy"
+strategies:
+  "HighNodeUtilization":
+    enabled: true
+    params:
+      nodeResourceUtilizationThresholds:
+        thresholds:
+          "memory": 20

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

@@ -1,17 +1,117 @@
 module sigs.k8s.io/descheduler
 
-go 1.15
+go 1.17
 
 require (
 	github.com/client9/misspell v0.3.4
-	github.com/spf13/cobra v0.0.5
+	github.com/spf13/cobra v1.2.1
 	github.com/spf13/pflag v1.0.5
-	k8s.io/api v0.20.0
-	k8s.io/apimachinery v0.20.0
-	k8s.io/apiserver v0.20.0
-	k8s.io/client-go v0.20.0
-	k8s.io/code-generator v0.20.0
-	k8s.io/component-base v0.20.0
-	k8s.io/component-helpers v0.20.0
-	k8s.io/klog/v2 v2.4.0
+	k8s.io/api v0.23.0
+	k8s.io/apimachinery v0.23.0
+	k8s.io/apiserver v0.23.0
+	k8s.io/client-go v0.23.0
+	k8s.io/code-generator v0.23.0
+	k8s.io/component-base v0.23.0
+	k8s.io/component-helpers v0.23.0
+	k8s.io/klog/v2 v2.30.0
+	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 v3.5.1+incompatible // indirect
+	github.com/cespare/xxhash/v2 v2.1.1 // 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/go-cmp v0.5.5 // indirect
+	github.com/google/gofuzz v1.1.0 // indirect
+	github.com/google/uuid v1.1.2 // indirect
+	github.com/googleapis/gnostic v0.5.5 // 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 v1.11.0 // indirect
+	github.com/prometheus/client_model v0.2.0 // indirect
+	github.com/prometheus/common v0.28.0 // indirect
+	github.com/prometheus/procfs v0.6.0 // indirect
+	go.etcd.io/etcd/api/v3 v3.5.0 // indirect
+	go.etcd.io/etcd/client/pkg/v3 v3.5.0 // indirect
+	go.etcd.io/etcd/client/v3 v3.5.0 // 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 v0.0.0-20210817164053-32db794688a5 // indirect
+	golang.org/x/mod v0.4.2 // indirect
+	golang.org/x/net v0.0.0-20210825183410-e898025ed96a // indirect
+	golang.org/x/oauth2 v0.0.0-20210819190943-2bc19b11175f // indirect
+	golang.org/x/sync v0.0.0-20210220032951-036812b2e83c // indirect
+	golang.org/x/sys v0.0.0-20210831042530-f4d43177bf5e // indirect
+	golang.org/x/term v0.0.0-20210615171337-6886f2dfbf5b // indirect
+	golang.org/x/text v0.3.7 // indirect
+	golang.org/x/time v0.0.0-20210723032227-1f47c861a9ac // indirect
+	golang.org/x/tools v0.1.6-0.20210820212750-d4cc65f0b2ff // indirect
+	golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1 // indirect
+	google.golang.org/appengine v1.6.7 // indirect
+	google.golang.org/genproto v0.0.0-20210831024726-fe130286e0e2 // 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 v0.0.0-20210813121822-485abfe95c7c // indirect
+	k8s.io/kube-openapi v0.0.0-20211115234752-e816edb12b65 // indirect
+	k8s.io/utils v0.0.0-20210930125809-cb0fa318a74b // indirect
+	sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.0.25 // indirect
+	sigs.k8s.io/json v0.0.0-20211020170558-c049b76a60c6 // indirect
+	sigs.k8s.io/structured-merge-diff/v4 v4.2.0 // 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
 
 /*
@@ -22,4 +23,5 @@ package tools
 import (
 	_ "github.com/client9/misspell/cmd/misspell"
 	_ "k8s.io/code-generator"
+	_ "sigs.k8s.io/mdtoc"
 )

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 'go1.13|go1.14|go1.15') ]]; 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/update-toc.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+
+# 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.
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+
+go build -o "${OS_OUTPUT_BINPATH}/mdtoc" "sigs.k8s.io/mdtoc"
+
+${OS_OUTPUT_BINPATH}/mdtoc --inplace README.md

hack/verify-conversions.sh

@@ -0,0 +1,36 @@
+#!/bin/bash
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
+mkdir -p "${DESCHEDULER_ROOT}/_tmp"
+_tmpdir="$(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}/conversion-gen" "k8s.io/code-generator/cmd/conversion-gen"
+
+${OS_OUTPUT_BINPATH}/conversion-gen \
+		--go-header-file "hack/boilerplate/boilerplate.go.txt" \
+		--input-dirs "./pkg/apis/componentconfig/v1alpha1,./pkg/api/v1alpha1" \
+		--output-file-base zz_generated.conversion
+popd > /dev/null 2>&1
+
+pushd "${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"
+    exit 1
+fi
+popd > /dev/null 2>&1
+
+echo "Generated conversions verified."

hack/verify-deep-copies.sh

@@ -0,0 +1,36 @@
+#!/bin/bash
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
+mkdir -p "${DESCHEDULER_ROOT}/_tmp"
+_tmpdir="$(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}/deepcopy-gen" "k8s.io/code-generator/cmd/deepcopy-gen"
+
+${OS_OUTPUT_BINPATH}/deepcopy-gen \
+                --go-header-file "hack/boilerplate/boilerplate.go.txt" \
+                --input-dirs "./pkg/apis/componentconfig,./pkg/apis/componentconfig/v1alpha1,./pkg/api,./pkg/api/v1alpha1" \
+                --output-file-base zz_generated.deepcopy
+popd > /dev/null 2>&1
+
+pushd "${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"
+    exit 1
+fi
+popd > /dev/null 2>&1
+
+echo "Generated deep-copies verified."

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 'go1.13|go1.14|go1.15') ]]; 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-govet.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+
+# 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.
+
+source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+
+go vet ${OS_ROOT}/...

hack/verify-toc.sh

@@ -0,0 +1,29 @@
+#!/bin/bash
+
+# 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.
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+
+go build -o "${OS_OUTPUT_BINPATH}/mdtoc" "sigs.k8s.io/mdtoc"
+
+if ! ${OS_OUTPUT_BINPATH}/mdtoc --inplace --dryrun README.md
+then
+  echo "ERROR: Changes detected to table of contents. Run ./hack/update-toc.sh" >&2
+  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"]
@@ -33,7 +33,6 @@ apiVersion: rbac.authorization.k8s.io/v1
 kind: ClusterRoleBinding
 metadata:
   name: descheduler-cluster-role-binding
-  namespace: kube-system
 roleRef:
   apiGroup: rbac.authorization.k8s.io
   kind: ClusterRole

kubernetes/cronjob/cronjob.yaml

@@ -1,5 +1,5 @@
 ---
-apiVersion: batch/v1beta1
+apiVersion: batch/v1  # for k8s version < 1.21.0, use batch/v1beta1
 kind: CronJob
 metadata:
   name: descheduler-cronjob
@@ -16,7 +16,7 @@ spec:
           priorityClassName: system-cluster-critical
           containers:
           - name: descheduler
-            image: k8s.gcr.io/descheduler/descheduler:v0.20.0
+            image: k8s.gcr.io/descheduler/descheduler:v0.23.1
             volumeMounts:
             - mountPath: /policy-dir
               name: policy-volume
@@ -31,6 +31,22 @@ spec:
               requests:
                 cpu: "500m"
                 memory: "256Mi"
+            livenessProbe:
+              failureThreshold: 3
+              httpGet:
+                path: /healthz
+                port: 10258
+                scheme: HTTPS
+              initialDelaySeconds: 3
+              periodSeconds: 10
+            securityContext:
+              allowPrivilegeEscalation: false
+              capabilities:
+                drop:
+                  - ALL
+              privileged: false
+              readOnlyRootFilesystem: true
+              runAsNonRoot: true
           restartPolicy: "Never"
           serviceAccountName: descheduler-sa
           volumes:

kubernetes/deployment/deployment.yaml

@@ -0,0 +1,62 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: descheduler
+  namespace: kube-system
+  labels:
+    app: descheduler
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: descheduler
+  template:
+    metadata:
+      labels:
+        app: descheduler
+    spec:
+      priorityClassName: system-cluster-critical
+      serviceAccountName: descheduler-sa
+      containers:
+        - name: descheduler
+          image: k8s.gcr.io/descheduler/descheduler:v0.23.1
+          imagePullPolicy: IfNotPresent
+          command:
+            - "/bin/descheduler"
+          args:
+            - "--policy-config-file"
+            - "/policy-dir/policy.yaml"
+            - "--descheduling-interval"
+            - "5m"
+            - "--v"
+            - "3"
+          ports:
+          - containerPort: 10258
+            protocol: TCP
+          livenessProbe:
+            failureThreshold: 3
+            httpGet:
+              path: /healthz
+              port: 10258
+              scheme: HTTPS
+            initialDelaySeconds: 3
+            periodSeconds: 10
+          resources:
+            requests:
+              cpu: 500m
+              memory: 256Mi
+          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: descheduler-policy-configmap

kubernetes/deployment/kustomization.yaml

@@ -0,0 +1,6 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+  - ../base
+  - deployment.yaml

kubernetes/job/job.yaml

@@ -14,7 +14,7 @@ spec:
       priorityClassName: system-cluster-critical
       containers:
         - name: descheduler
-          image: k8s.gcr.io/descheduler/descheduler:v0.20.0
+          image: k8s.gcr.io/descheduler/descheduler:v0.23.1
           volumeMounts:
           - mountPath: /policy-dir
             name: policy-volume
@@ -29,6 +29,22 @@ spec:
             requests:
               cpu: "500m"
               memory: "256Mi"
+          livenessProbe:
+            failureThreshold: 3
+            httpGet:
+              path: /healthz
+              port: 10258
+              scheme: HTTPS
+            initialDelaySeconds: 3
+            periodSeconds: 10
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop:
+                - ALL
+            privileged: false
+            readOnlyRootFilesystem: true
+            runAsNonRoot: true
       restartPolicy: "Never"
       serviceAccountName: descheduler-sa
       volumes:

metrics/metrics.go

@@ -0,0 +1,72 @@
+/*
+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 metrics
+
+import (
+	"sync"
+
+	"k8s.io/component-base/metrics"
+	"k8s.io/component-base/metrics/legacyregistry"
+	"sigs.k8s.io/descheduler/pkg/version"
+)
+
+const (
+	// DeschedulerSubsystem - subsystem name used by descheduler
+	DeschedulerSubsystem = "descheduler"
+)
+
+var (
+	PodsEvicted = metrics.NewCounterVec(
+		&metrics.CounterOpts{
+			Subsystem:      DeschedulerSubsystem,
+			Name:           "pods_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", "node"})
+
+	buildInfo = metrics.NewGauge(
+		&metrics.GaugeOpts{
+			Subsystem:      DeschedulerSubsystem,
+			Name:           "build_info",
+			Help:           "Build info about descheduler, including Go version, Descheduler version, Git SHA, Git branch",
+			ConstLabels:    map[string]string{"GoVersion": version.Get().GoVersion, "DeschedulerVersion": version.Get().GitVersion, "GitBranch": version.Get().GitBranch, "GitSha1": version.Get().GitSha1},
+			StabilityLevel: metrics.ALPHA,
+		},
+	)
+
+	metricsList = []metrics.Registerable{
+		PodsEvicted,
+		buildInfo,
+	}
+)
+
+var registerMetrics sync.Once
+
+// Register all metrics.
+func Register() {
+	// Register the metrics.
+	registerMetrics.Do(func() {
+		RegisterMetrics(metricsList...)
+	})
+}
+
+// RegisterMetrics registers a list of metrics.
+func RegisterMetrics(extraMetrics ...metrics.Registerable) {
+	for _, metric := range extraMetrics {
+		legacyregistry.MustRegister(metric)
+	}
+}

pkg/api/types.go

@@ -32,11 +32,23 @@ 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
 
+	// EvictSystemCriticalPods allows eviction of pods of any priority (including Kubernetes system pods)
+	EvictSystemCriticalPods *bool
+
+	// IgnorePVCPods prevents pods with PVCs from being evicted.
+	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
@@ -69,9 +81,13 @@ type StrategyParameters struct {
 	PodsHavingTooManyRestarts         *PodsHavingTooManyRestarts
 	PodLifeTime                       *PodLifeTime
 	RemoveDuplicates                  *RemoveDuplicates
+	FailedPods                        *FailedPods
+	IncludeSoftConstraints            bool
 	Namespaces                        *Namespaces
 	ThresholdPriority                 *int32
 	ThresholdPriorityClassName        string
+	LabelSelector                     *metav1.LabelSelector
+	NodeFit                           bool
 }
 
 type Percentage float64
@@ -96,3 +112,10 @@ type PodLifeTime struct {
 	MaxPodLifeTimeSeconds *uint
 	PodStatusPhases       []string
 }
+
+type FailedPods struct {
+	ExcludeOwnerKinds       []string
+	MinPodLifetimeSeconds   *uint
+	Reasons                 []string
+	IncludingInitContainers bool
+}

pkg/api/v1alpha1/types.go

@@ -32,11 +32,23 @@ 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"`
 
+	// EvictSystemCriticalPods allows eviction of pods of any priority (including Kubernetes system pods)
+	EvictSystemCriticalPods *bool `json:"evictSystemCriticalPods,omitempty"`
+
+	// IgnorePVCPods prevents pods with PVCs from being evicted.
+	IgnorePVCPods *bool `json:"ignorePvcPods,omitempty"`
+
 	// 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
@@ -67,9 +79,13 @@ 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"`
 }
 
 type Percentage float64
@@ -94,3 +110,10 @@ 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 2020 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.
@@ -23,6 +24,7 @@ package v1alpha1
 import (
 	unsafe "unsafe"
 
+	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	conversion "k8s.io/apimachinery/pkg/conversion"
 	runtime "k8s.io/apimachinery/pkg/runtime"
 	api "sigs.k8s.io/descheduler/pkg/api"
@@ -55,6 +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 {
@@ -121,8 +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.MaxNoOfPodsToEvictPerNode = (*int)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNode))
+	out.EvictSystemCriticalPods = (*bool)(unsafe.Pointer(in.EvictSystemCriticalPods))
+	out.IgnorePVCPods = (*bool)(unsafe.Pointer(in.IgnorePVCPods))
+	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
 }
 
@@ -134,8 +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.MaxNoOfPodsToEvictPerNode = (*int)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNode))
+	out.EvictSystemCriticalPods = (*bool)(unsafe.Pointer(in.EvictSystemCriticalPods))
+	out.IgnorePVCPods = (*bool)(unsafe.Pointer(in.IgnorePVCPods))
+	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
 }
 
@@ -168,6 +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))
@@ -284,9 +354,13 @@ 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
 	return nil
 }
 
@@ -301,9 +375,13 @@ 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
 	return nil
 }
 

pkg/api/v1alpha1/zz_generated.deepcopy.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2020 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.
@@ -21,6 +22,7 @@ limitations under the License.
 package v1alpha1
 
 import (
+	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	runtime "k8s.io/apimachinery/pkg/runtime"
 )
 
@@ -40,16 +42,36 @@ 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)
 		**out = **in
 	}
+	if in.EvictSystemCriticalPods != nil {
+		in, out := &in.EvictSystemCriticalPods, &out.EvictSystemCriticalPods
+		*out = new(bool)
+		**out = **in
+	}
+	if in.IgnorePVCPods != nil {
+		in, out := &in.IgnorePVCPods, &out.IgnorePVCPods
+		*out = new(bool)
+		**out = **in
+	}
 	if in.MaxNoOfPodsToEvictPerNode != nil {
 		in, out := &in.MaxNoOfPodsToEvictPerNode, &out.MaxNoOfPodsToEvictPerNode
 		*out = new(int)
 		**out = **in
 	}
+	if in.MaxNoOfPodsToEvictPerNamespace != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
+		*out = new(int)
+		**out = **in
+	}
 	return
 }
 
@@ -92,6 +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
@@ -283,6 +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)
@@ -293,6 +351,11 @@ func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
 		*out = new(int32)
 		**out = **in
 	}
+	if in.LabelSelector != nil {
+		in, out := &in.LabelSelector, &out.LabelSelector
+		*out = new(v1.LabelSelector)
+		(*in).DeepCopyInto(*out)
+	}
 	return
 }
 

pkg/api/v1alpha1/zz_generated.defaults.go

@@ -1,7 +1,8 @@
+//go:build !ignore_autogenerated
 // +build !ignore_autogenerated
 
 /*
-Copyright 2020 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 2020 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.
@@ -21,6 +22,7 @@ limitations under the License.
 package api
 
 import (
+	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	runtime "k8s.io/apimachinery/pkg/runtime"
 )
 
@@ -40,14 +42,34 @@ 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)
 		**out = **in
 	}
+	if in.EvictSystemCriticalPods != nil {
+		in, out := &in.EvictSystemCriticalPods, &out.EvictSystemCriticalPods
+		*out = new(bool)
+		**out = **in
+	}
+	if in.IgnorePVCPods != nil {
+		in, out := &in.IgnorePVCPods, &out.IgnorePVCPods
+		*out = new(bool)
+		**out = **in
+	}
 	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
@@ -92,6 +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
@@ -283,6 +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)
@@ -293,6 +351,11 @@ func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
 		*out = new(int32)
 		**out = **in
 	}
+	if in.LabelSelector != nil {
+		in, out := &in.LabelSelector, &out.LabelSelector
+		*out = new(v1.LabelSelector)
+		(*in).DeepCopyInto(*out)
+	}
 	return
 }
 

pkg/apis/componentconfig/types.go

@@ -50,6 +50,9 @@ type DeschedulerConfiguration struct {
 	// EvictLocalStoragePods allows pods using local storage to be evicted.
 	EvictLocalStoragePods bool
 
+	// IgnorePVCPods sets whether PVC pods should be allowed to be evicted
+	IgnorePVCPods bool
+
 	// 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

@@ -19,6 +19,8 @@ package v1alpha1
 import (
 	"time"
 
+	componentbaseconfig "k8s.io/component-base/config"
+
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 )
 
@@ -48,4 +50,11 @@ type DeschedulerConfiguration struct {
 
 	// EvictLocalStoragePods allows pods using local storage to be evicted.
 	EvictLocalStoragePods bool `json:"evictLocalStoragePods,omitempty"`
+
+	// IgnorePVCPods sets whether PVC pods should be allowed to be evicted
+	IgnorePVCPods bool `json:"ignorePvcPods,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 2020 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 +57,8 @@ func autoConvert_v1alpha1_DeschedulerConfiguration_To_componentconfig_Deschedule
 	out.NodeSelector = in.NodeSelector
 	out.MaxNoOfPodsToEvictPerNode = in.MaxNoOfPodsToEvictPerNode
 	out.EvictLocalStoragePods = in.EvictLocalStoragePods
+	out.IgnorePVCPods = in.IgnorePVCPods
+	out.Logging = in.Logging
 	return nil
 }
 
@@ -72,6 +75,8 @@ func autoConvert_componentconfig_DeschedulerConfiguration_To_v1alpha1_Deschedule
 	out.NodeSelector = in.NodeSelector
 	out.MaxNoOfPodsToEvictPerNode = in.MaxNoOfPodsToEvictPerNode
 	out.EvictLocalStoragePods = in.EvictLocalStoragePods
+	out.IgnorePVCPods = in.IgnorePVCPods
+	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 2020 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.
@@ -28,6 +29,7 @@ import (
 func (in *DeschedulerConfiguration) DeepCopyInto(out *DeschedulerConfiguration) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
+	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 2020 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 2020 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.
@@ -28,6 +29,7 @@ import (
 func (in *DeschedulerConfiguration) DeepCopyInto(out *DeschedulerConfiguration) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
+	in.Logging.DeepCopyInto(&out.Logging)
 	return
 }
 

pkg/descheduler/descheduler.go

@@ -21,22 +21,35 @@ import (
 	"fmt"
 
 	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"
 	"sigs.k8s.io/descheduler/pkg/descheduler/client"
 	"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 {
-	ctx := context.Background()
+func Run(ctx context.Context, rs *options.DeschedulerServer) error {
+	metrics.Register()
+
 	rsclient, err := client.CreateClient(rs.KubeconfigFile)
 	if err != nil {
 		return err
@@ -56,79 +69,229 @@ func Run(rs *options.DeschedulerServer) error {
 		return err
 	}
 
-	stopChannel := make(chan struct{})
-	return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion, stopChannel)
+	return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion)
 }
 
-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()
 
-	strategyFuncs := map[string]strategyFunction{
+	// 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,
-		"LowNodeUtilization":                          strategies.LowNodeUtilization,
+		"LowNodeUtilization":                          nodeutilization.LowNodeUtilization,
+		"HighNodeUtilization":                         nodeutilization.HighNodeUtilization,
 		"RemovePodsViolatingInterPodAntiAffinity":     strategies.RemovePodsViolatingInterPodAntiAffinity,
 		"RemovePodsViolatingNodeAffinity":             strategies.RemovePodsViolatingNodeAffinity,
 		"RemovePodsViolatingNodeTaints":               strategies.RemovePodsViolatingNodeTaints,
 		"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
 	}
 
-	maxNoOfPodsToEvictPerNode := rs.MaxNoOfPodsToEvictPerNode
-	if deschedulerPolicy.MaxNoOfPodsToEvictPerNode != nil {
-		maxNoOfPodsToEvictPerNode = *deschedulerPolicy.MaxNoOfPodsToEvictPerNode
+	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.")
+		}
 	}
 
-	wait.Until(func() {
-		nodes, err := nodeutil.ReadyNodes(ctx, rs.Client, nodeInformer, nodeSelector, stopChannel)
+	evictSystemCriticalPods := false
+	if deschedulerPolicy.EvictSystemCriticalPods != nil {
+		evictSystemCriticalPods = *deschedulerPolicy.EvictSystemCriticalPods
+		if evictSystemCriticalPods {
+			klog.V(1).InfoS("Warning: EvictSystemCriticalPods is set to True. This could cause eviction of Kubernetes system pods.")
+		}
+	}
+
+	ignorePvcPods := false
+	if deschedulerPolicy.IgnorePVCPods != nil {
+		ignorePvcPods = *deschedulerPolicy.IgnorePVCPods
+	}
+
+	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,
 			evictLocalStoragePods,
+			evictSystemCriticalPods,
+			ignorePvcPods,
+			evictBarePods,
+			!rs.DisableMetrics,
 		)
 
-		for name, f := range strategyFuncs {
-			if strategy := deschedulerPolicy.Strategies[api.StrategyName(name)]; strategy.Enabled {
-				f(ctx, rs.Client, strategy, nodes, podEvictor)
+		for name, strategy := range deschedulerPolicy.Strategies {
+			if f, ok := strategyFuncs[name]; ok {
+				if strategy.Enabled {
+					f(ctx, rs.Client, strategy, nodes, podEvictor, getPodsAssignedToNode)
+				}
+			} else {
+				klog.ErrorS(fmt.Errorf("unknown strategy name"), "skipping strategy", "strategy", name)
 			}
 		}
 
+		klog.V(1).InfoS("Number of evicted pods", "totalEvicted", podEvictor.TotalEvicted())
+
 		// 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

@@ -7,7 +7,7 @@ import (
 	"testing"
 	"time"
 
-	"k8s.io/api/core/v1"
+	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/wait"
 	fakeclientset "k8s.io/client-go/kubernetes/fake"
@@ -35,24 +35,27 @@ 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)
 	}
 	rs.Client = client
 	rs.DeschedulingInterval = 100 * time.Millisecond
+	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 {
+	case err := <-errChan:
 		if err != nil {
 			t.Fatalf("Unable to run descheduler strategies: %v", err)
 		}
-	}()
+	case <-time.After(1 * time.Second):
+		// Wait for few cycles and then verify the only pod still exists
+	}
 
-	// Wait for few cycles and then verify the only pod still exists
-	time.Sleep(300 * time.Millisecond)
 	pods, err := client.CoreV1().Pods(p1.Namespace).List(ctx, metav1.ListOptions{})
 	if err != nil {
 		t.Errorf("Unable to list pods: %v", err)
@@ -95,3 +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

@@ -25,12 +25,15 @@ import (
 	policy "k8s.io/api/policy/v1beta1"
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/util/errors"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/client-go/kubernetes/scheme"
 	clientcorev1 "k8s.io/client-go/kubernetes/typed/core/v1"
 	"k8s.io/client-go/tools/record"
 	"k8s.io/klog/v2"
+	"sigs.k8s.io/descheduler/metrics"
+	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 
@@ -42,49 +45,70 @@ 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
-	policyGroupVersion    string
-	dryRun                bool
-	maxPodsToEvictPerNode int
-	nodepodCount          nodePodEvictedCount
-	evictLocalStoragePods bool
+	client                     clientset.Interface
+	nodes                      []*v1.Node
+	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,
 	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,
-		policyGroupVersion:    policyGroupVersion,
-		dryRun:                dryRun,
-		maxPodsToEvictPerNode: maxPodsToEvictPerNode,
-		nodepodCount:          nodePodCount,
-		evictLocalStoragePods: evictLocalStoragePods,
+		client:                     client,
+		nodes:                      nodes,
+		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
 	}
@@ -94,23 +118,42 @@ func (pe *PodEvictor) TotalEvicted() int {
 // EvictPod returns non-nil error only when evicting a pod on a node is not
 // possible (due to maxPodsToEvictPerNode constraint). Success is true when the pod
 // is evicted on the server side.
-func (pe *PodEvictor) EvictPod(ctx context.Context, pod *v1.Pod, node *v1.Node, reasons ...string) (bool, error) {
-	var reason string
+func (pe *PodEvictor) EvictPod(ctx context.Context, pod *v1.Pod, node *v1.Node, strategy string, reasons ...string) (bool, error) {
+	reason := strategy
 	if len(reasons) > 0 {
-		reason = " (" + strings.Join(reasons, ", ") + ")"
+		reason += " (" + strings.Join(reasons, ", ") + ")"
 	}
-	if pe.maxPodsToEvictPerNode > 0 && pe.nodepodCount[node]+1 > pe.maxPodsToEvictPerNode {
-		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)
+		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)
 	} else {
@@ -124,10 +167,7 @@ func (pe *PodEvictor) EvictPod(ctx context.Context, pod *v1.Pod, node *v1.Node,
 	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{
@@ -153,7 +193,9 @@ func evictPod(ctx context.Context, client clientset.Interface, pod *v1.Pod, poli
 }
 
 type Options struct {
-	priority *int32
+	priority      *int32
+	nodeFit       bool
+	labelSelector labels.Selector
 }
 
 // WithPriorityThreshold sets a threshold for pod's priority class.
@@ -165,6 +207,24 @@ func WithPriorityThreshold(priority int32) func(opts *Options) {
 	}
 }
 
+// WithNodeFit sets whether or not to consider taints, node selectors,
+// and pod affinity when evicting. A pod whose tolerations, node selectors,
+// and affinity match a node other than the one it is currently running on
+// is evictable.
+func WithNodeFit(nodeFit bool) func(opts *Options) {
+	return func(opts *Options) {
+		opts.nodeFit = nodeFit
+	}
+}
+
+// WithLabelSelector sets whether or not to apply label filtering when evicting.
+// Any pod matching the label selector is considered evictable.
+func WithLabelSelector(labelSelector labels.Selector) func(opts *Options) {
+	return func(opts *Options) {
+		opts.labelSelector = labelSelector
+	}
+}
+
 type constraint func(pod *v1.Pod) error
 
 type evictable struct {
@@ -181,43 +241,98 @@ func (pe *PodEvictor) Evictable(opts ...func(opts *Options)) *evictable {
 	}
 
 	ev := &evictable{}
-	if !pe.evictLocalStoragePods {
+	if pe.evictFailedBarePods {
 		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
-			if IsPodWithLocalStorage(pod) {
-				return fmt.Errorf("pod has local storage and descheduler is not configured with --evict-local-storage-pods")
+			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 options.priority != nil {
+	if !pe.evictSystemCriticalPods {
 		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
-			if IsPodEvictableBasedOnPriority(pod, *options.priority) {
-				return nil
+			// Moved from IsEvictable function to allow for disabling
+			if utils.IsCriticalPriorityPod(pod) {
+				return fmt.Errorf("pod has system critical priority")
 			}
-			return fmt.Errorf("pod has higher priority than specified priority class threshold")
+			return nil
+		})
+
+		if options.priority != nil {
+			ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+				if IsPodEvictableBasedOnPriority(pod, *options.priority) {
+					return nil
+				}
+				return fmt.Errorf("pod has higher priority than specified priority class threshold")
+			})
+		}
+	}
+	if !pe.evictLocalStoragePods {
+		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+			if utils.IsPodWithLocalStorage(pod) {
+				return fmt.Errorf("pod has local storage and descheduler is not configured with evictLocalStoragePods")
+			}
+			return nil
 		})
 	}
+	if pe.ignorePvcPods {
+		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+			if utils.IsPodWithPVC(pod) {
+				return fmt.Errorf("pod has a PVC and descheduler is configured to ignore PVC pods")
+			}
+			return nil
+		})
+	}
+	if options.nodeFit {
+		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+			if !nodeutil.PodFitsAnyOtherNode(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
+		})
+	}
+	if options.labelSelector != nil && !options.labelSelector.Empty() {
+		ev.constraints = append(ev.constraints, func(pod *v1.Pod) error {
+			if !options.labelSelector.Matches(labels.Set(pod.Labels)) {
+				return fmt.Errorf("pod labels do not match the labelSelector filter in the policy parameter")
+			}
+			return nil
+		})
+	}
+
 	return ev
 }
 
 // IsEvictable decides when a pod is evictable
 func (ev *evictable) IsEvictable(pod *v1.Pod) bool {
 	checkErrs := []error{}
-	if IsCriticalPod(pod) {
-		checkErrs = append(checkErrs, fmt.Errorf("pod is critical"))
-	}
 
 	ownerRefList := podutil.OwnerRef(pod)
-	if IsDaemonsetPod(ownerRefList) {
+	if utils.IsDaemonsetPod(ownerRefList) {
 		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"))
 	}
 
-	if IsMirrorPod(pod) {
-		checkErrs = append(checkErrs, fmt.Errorf("pod is a mirror pod"))
+	if utils.IsStaticPod(pod) {
+		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 {
@@ -230,43 +345,16 @@ func (ev *evictable) IsEvictable(pod *v1.Pod) bool {
 		klog.V(4).InfoS("Pod lacks an eviction annotation and fails the following checks", "pod", klog.KObj(pod), "checks", errors.NewAggregate(checkErrs).Error())
 		return false
 	}
+
 	return true
 }
 
-func IsCriticalPod(pod *v1.Pod) bool {
-	return utils.IsCriticalPod(pod)
-}
-
-func IsDaemonsetPod(ownerRefList []metav1.OwnerReference) bool {
-	for _, ownerRef := range ownerRefList {
-		if ownerRef.Kind == "DaemonSet" {
-			return true
-		}
-	}
-	return false
-}
-
-// IsMirrorPod checks whether the pod is a mirror pod.
-func IsMirrorPod(pod *v1.Pod) bool {
-	return utils.IsMirrorPod(pod)
-}
-
 // HaveEvictAnnotation checks if the pod have evict annotation
 func HaveEvictAnnotation(pod *v1.Pod) bool {
 	_, found := pod.ObjectMeta.Annotations[evictPodAnnotationKey]
 	return found
 }
 
-func IsPodWithLocalStorage(pod *v1.Pod) bool {
-	for _, volume := range pod.Spec.Volumes {
-		if volume.HostPath != nil || volume.EmptyDir != nil {
-			return true
-		}
-	}
-
-	return false
-}
-
 // IsPodEvictableBasedOnPriority checks if the given pod is evictable based on priority resolved from pod Spec.
 func IsPodEvictableBasedOnPriority(pod *v1.Pod, priority int32) bool {
 	return pod.Spec.Priority == nil || *pod.Spec.Priority < priority

pkg/descheduler/evictions/evictions_test.go

@@ -62,7 +62,7 @@ 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)
 		}
@@ -73,48 +73,99 @@ func TestIsEvictable(t *testing.T) {
 	n1 := test.BuildTestNode("node1", 1000, 2000, 13, nil)
 	lowPriority := int32(800)
 	highPriority := int32(900)
+
+	nodeTaintKey := "hardware"
+	nodeTaintValue := "gpu"
+
+	nodeLabelKey := "datacenter"
+	nodeLabelValue := "east"
 	type testCase struct {
-		pod                   *v1.Pod
-		runBefore             func(*v1.Pod)
-		evictLocalStoragePods bool
-		priorityThreshold     *int32
-		result                bool
+		pod                     *v1.Pod
+		nodes                   []*v1.Node
+		runBefore               func(*v1.Pod, []*v1.Node)
+		evictFailedBarePods     bool
+		evictLocalStoragePods   bool
+		evictSystemCriticalPods bool
+		priorityThreshold       *int32
+		nodeFit                 bool
+		result                  bool
 	}
 
 	testCases := []testCase{
-		{
+		{ // Failed pod eviction with no ownerRefs.
+			pod: test.BuildTestPod("bare_pod_failed", 400, 0, n1.Name, nil),
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
+				pod.Status.Phase = v1.PodFailed
+			},
+			evictFailedBarePods: false,
+			result:              false,
+		}, { // Normal pod eviction with no ownerRefs and evictFailedBarePods enabled
+			pod: test.BuildTestPod("bare_pod", 400, 0, n1.Name, nil),
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
+			},
+			evictFailedBarePods: true,
+			result:              false,
+		}, { // Failed pod eviction with no ownerRefs
+			pod: test.BuildTestPod("bare_pod_failed_but_can_be_evicted", 400, 0, n1.Name, nil),
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
+				pod.Status.Phase = v1.PodFailed
+			},
+			evictFailedBarePods: true,
+			result:              true,
+		}, { // Normal pod eviction with normal ownerRefs
 			pod: test.BuildTestPod("p1", 400, 0, n1.Name, nil),
-			runBefore: func(pod *v1.Pod) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 			},
-			evictLocalStoragePods: false,
-			result:                true,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 			},
-			evictLocalStoragePods: false,
-			result:                true,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
 			},
-			evictLocalStoragePods: false,
-			result:                true,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
 				pod.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
 			},
-			evictLocalStoragePods: false,
-			result:                true,
-		}, {
+			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()
+			},
+			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()
+			},
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 				pod.Spec.Volumes = []v1.Volume{
 					{
@@ -127,11 +178,12 @@ func TestIsEvictable(t *testing.T) {
 					},
 				}
 			},
-			evictLocalStoragePods: false,
-			result:                false,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 				pod.Spec.Volumes = []v1.Volume{
 					{
@@ -144,11 +196,12 @@ func TestIsEvictable(t *testing.T) {
 					},
 				}
 			},
-			evictLocalStoragePods: true,
-			result:                true,
-		}, {
+			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) {
+			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{
@@ -162,52 +215,58 @@ func TestIsEvictable(t *testing.T) {
 					},
 				}
 			},
-			evictLocalStoragePods: false,
-			result:                true,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
 			},
-			evictLocalStoragePods: false,
-			result:                false,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.Annotations = map[string]string{"descheduler.alpha.kubernetes.io/evict": "true"}
 				pod.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
 			},
-			evictLocalStoragePods: false,
-			result:                true,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 				pod.Annotations = test.GetMirrorPodAnnotation()
 			},
-			evictLocalStoragePods: false,
-			result:                false,
-		}, {
+			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) {
+			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"
 			},
-			evictLocalStoragePods: false,
-			result:                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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 				priority := utils.SystemCriticalPriority
 				pod.Spec.Priority = &priority
 			},
-			evictLocalStoragePods: false,
-			result:                false,
-		}, {
+			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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 				priority := utils.SystemCriticalPriority
 				pod.Spec.Priority = &priority
@@ -215,41 +274,183 @@ func TestIsEvictable(t *testing.T) {
 					"descheduler.alpha.kubernetes.io/evict": "true",
 				}
 			},
-			evictLocalStoragePods: false,
-			result:                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) {
+			runBefore: func(pod *v1.Pod, nodes []*v1.Node) {
 				pod.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 				pod.Spec.Priority = &highPriority
 			},
-			evictLocalStoragePods: false,
-			priorityThreshold:     &lowPriority,
-			result:                false,
-		}, {
+			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) {
+			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
 			},
-			evictLocalStoragePods: false,
-			priorityThreshold:     &lowPriority,
-			result:                true,
+			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
+			},
+			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
+			},
+			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
+			},
+			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
+			},
+			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 {
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							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 {
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							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,
+					}
+				}
+			},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			nodeFit:                 true,
+			result:                  true,
 		},
 	}
 
 	for _, test := range testCases {
-		test.runBefore(test.pod)
+		test.runBefore(test.pod, test.nodes)
+		nodes := append(test.nodes, n1)
 
 		podEvictor := &PodEvictor{
-			evictLocalStoragePods: test.evictLocalStoragePods,
+			evictLocalStoragePods:   test.evictLocalStoragePods,
+			evictSystemCriticalPods: test.evictSystemCriticalPods,
+			evictFailedBarePods:     test.evictFailedBarePods,
+			nodes:                   nodes,
 		}
 
 		evictable := podEvictor.Evictable()
+		var opts []func(opts *Options)
 		if test.priorityThreshold != nil {
-			evictable = podEvictor.Evictable(WithPriorityThreshold(*test.priorityThreshold))
+			opts = append(opts, WithPriorityThreshold(*test.priorityThreshold))
 		}
+		if test.nodeFit {
+			opts = append(opts, WithNodeFit(true))
+		}
+		evictable = podEvictor.Evictable(opts...)
 
 		result := evictable.IsEvictable(test.pod)
 		if result != test.result {
@@ -286,18 +487,18 @@ func TestPodTypes(t *testing.T) {
 	}
 	// A Mirror Pod.
 	p4.Annotations = test.GetMirrorPodAnnotation()
-	if !IsMirrorPod(p4) {
+	if !utils.IsMirrorPod(p4) {
 		t.Errorf("Expected p4 to be a mirror pod.")
 	}
-	if !IsPodWithLocalStorage(p3) {
+	if !utils.IsPodWithLocalStorage(p3) {
 		t.Errorf("Expected p3 to be a pod with local storage.")
 	}
 	ownerRefList := podutil.OwnerRef(p2)
-	if !IsDaemonsetPod(ownerRefList) {
+	if !utils.IsDaemonsetPod(ownerRefList) {
 		t.Errorf("Expected p2 to be a daemonset pod.")
 	}
 	ownerRefList = podutil.OwnerRef(p1)
-	if IsDaemonsetPod(ownerRefList) || IsPodWithLocalStorage(p1) || IsCriticalPod(p1) || IsMirrorPod(p1) {
+	if utils.IsDaemonsetPod(ownerRefList) || utils.IsPodWithLocalStorage(p1) || utils.IsCriticalPriorityPod(p1) || utils.IsMirrorPod(p1) || utils.IsStaticPod(p1) {
 		t.Errorf("Expected p1 to be a normal pod.")
 	}
 

pkg/descheduler/node/node.go

@@ -30,7 +30,7 @@ import (
 
 // ReadyNodes returns ready nodes irrespective of whether they are
 // schedulable or not.
-func ReadyNodes(ctx context.Context, client clientset.Interface, nodeInformer coreinformers.NodeInformer, nodeSelector string, stopChannel <-chan struct{}) ([]*v1.Node, error) {
+func ReadyNodes(ctx context.Context, client clientset.Interface, nodeInformer coreinformers.NodeInformer, nodeSelector string) ([]*v1.Node, error) {
 	ns, err := labels.Parse(nodeSelector)
 	if err != nil {
 		return []*v1.Node{}, err
@@ -96,7 +96,38 @@ func IsReady(node *v1.Node) bool {
 	return true
 }
 
-// IsNodeUnschedulable checks if the node is unschedulable. This is helper function to check only in case of
+// 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 {
+
+	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))
+			return true
+		}
+	}
+	return false
+}
+
+// IsNodeUnschedulable checks if the node is unschedulable. This is a helper function to check only in case of
 // underutilized node so that they won't be accounted for.
 func IsNodeUnschedulable(node *v1.Node) bool {
 	return node.Spec.Unschedulable

pkg/descheduler/node/node_test.go

@@ -74,7 +74,7 @@ func TestReadyNodesWithNodeSelector(t *testing.T) {
 	sharedInformerFactory.WaitForCacheSync(stopChannel)
 	defer close(stopChannel)
 
-	nodes, _ := ReadyNodes(ctx, fakeClient, nodeInformer, nodeSelector, nil)
+	nodes, _ := ReadyNodes(ctx, fakeClient, nodeInformer, nodeSelector)
 
 	if nodes[0].Name != "node1" {
 		t.Errorf("Expected node1, got %s", nodes[0].Name)
@@ -200,10 +200,15 @@ func TestPodFitsCurrentNode(t *testing.T) {
 	}
 }
 
-func TestPodFitsAnyNode(t *testing.T) {
+func TestPodFitsAnyOtherNode(t *testing.T) {
 
 	nodeLabelKey := "kubernetes.io/desiredNode"
 	nodeLabelValue := "yes"
+	nodeTaintKey := "hardware"
+	nodeTaintValue := "gpu"
+
+	// Staging node has no scheduling restrictions, but the pod always starts here and PodFitsAnyOtherNode() doesn't take into account the node the pod is running on.
+	nodeNames := []string{"node1", "node2", "stagingNode"}
 
 	tests := []struct {
 		description string
@@ -212,33 +217,12 @@ func TestPodFitsAnyNode(t *testing.T) {
 		success     bool
 	}{
 		{
-			description: "Pod expected to fit one of the nodes",
-			pod: &v1.Pod{
-				Spec: v1.PodSpec{
-					Affinity: &v1.Affinity{
-						NodeAffinity: &v1.NodeAffinity{
-							RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
-								NodeSelectorTerms: []v1.NodeSelectorTerm{
-									{
-										MatchExpressions: []v1.NodeSelectorRequirement{
-											{
-												Key:      nodeLabelKey,
-												Operator: "In",
-												Values: []string{
-													nodeLabelValue,
-												},
-											},
-										},
-									},
-								},
-							},
-						},
-					},
-				},
-			},
+			description: "Pod fits another node matching node affinity",
+			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
 			nodes: []*v1.Node{
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[0],
 						Labels: map[string]string{
 							nodeLabelKey: nodeLabelValue,
 						},
@@ -246,42 +230,27 @@ func TestPodFitsAnyNode(t *testing.T) {
 				},
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[1],
 						Labels: map[string]string{
 							nodeLabelKey: "no",
 						},
 					},
 				},
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[2],
+					},
+				},
 			},
 			success: true,
 		},
 		{
-			description: "Pod expected to fit one of the nodes (error node first)",
-			pod: &v1.Pod{
-				Spec: v1.PodSpec{
-					Affinity: &v1.Affinity{
-						NodeAffinity: &v1.NodeAffinity{
-							RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
-								NodeSelectorTerms: []v1.NodeSelectorTerm{
-									{
-										MatchExpressions: []v1.NodeSelectorRequirement{
-											{
-												Key:      nodeLabelKey,
-												Operator: "In",
-												Values: []string{
-													nodeLabelValue,
-												},
-											},
-										},
-									},
-								},
-							},
-						},
-					},
-				},
-			},
+			description: "Pod expected to fit one of the nodes",
+			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
 			nodes: []*v1.Node{
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[0],
 						Labels: map[string]string{
 							nodeLabelKey: "no",
 						},
@@ -289,42 +258,27 @@ func TestPodFitsAnyNode(t *testing.T) {
 				},
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[1],
 						Labels: map[string]string{
 							nodeLabelKey: nodeLabelValue,
 						},
 					},
 				},
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[2],
+					},
+				},
 			},
 			success: true,
 		},
 		{
 			description: "Pod expected to fit none of the nodes",
-			pod: &v1.Pod{
-				Spec: v1.PodSpec{
-					Affinity: &v1.Affinity{
-						NodeAffinity: &v1.NodeAffinity{
-							RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
-								NodeSelectorTerms: []v1.NodeSelectorTerm{
-									{
-										MatchExpressions: []v1.NodeSelectorRequirement{
-											{
-												Key:      nodeLabelKey,
-												Operator: "In",
-												Values: []string{
-													nodeLabelValue,
-												},
-											},
-										},
-									},
-								},
-							},
-						},
-					},
-				},
-			},
+			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
 			nodes: []*v1.Node{
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[0],
 						Labels: map[string]string{
 							nodeLabelKey: "unfit1",
 						},
@@ -332,42 +286,27 @@ func TestPodFitsAnyNode(t *testing.T) {
 				},
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[1],
 						Labels: map[string]string{
 							nodeLabelKey: "unfit2",
 						},
 					},
 				},
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[2],
+					},
+				},
 			},
 			success: false,
 		},
 		{
 			description: "Nodes are unschedulable but labels match, should fail",
-			pod: &v1.Pod{
-				Spec: v1.PodSpec{
-					Affinity: &v1.Affinity{
-						NodeAffinity: &v1.NodeAffinity{
-							RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
-								NodeSelectorTerms: []v1.NodeSelectorTerm{
-									{
-										MatchExpressions: []v1.NodeSelectorRequirement{
-											{
-												Key:      nodeLabelKey,
-												Operator: "In",
-												Values: []string{
-													nodeLabelValue,
-												},
-											},
-										},
-									},
-								},
-							},
-						},
-					},
-				},
-			},
+			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
 			nodes: []*v1.Node{
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[0],
 						Labels: map[string]string{
 							nodeLabelKey: nodeLabelValue,
 						},
@@ -378,20 +317,136 @@ func TestPodFitsAnyNode(t *testing.T) {
 				},
 				{
 					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[1],
 						Labels: map[string]string{
 							nodeLabelKey: "no",
 						},
 					},
 				},
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[2],
+					},
+				},
+			},
+			success: false,
+		},
+		{
+			description: "Two nodes matches node selector, one of them is tained, should pass",
+			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			nodes: []*v1.Node{
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[0],
+						Labels: map[string]string{
+							nodeLabelKey: nodeLabelValue,
+						},
+					},
+					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],
+					},
+				},
+			},
+			success: true,
+		},
+		{
+			description: "Both nodes are tained, should fail",
+			pod:         createPodManifest(nodeNames[2], nodeLabelKey, nodeLabelValue),
+			nodes: []*v1.Node{
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[0],
+						Labels: map[string]string{
+							nodeLabelKey: nodeLabelValue,
+						},
+					},
+					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,
+						},
+					},
+					Spec: v1.NodeSpec{
+						Taints: []v1.Taint{
+							{
+								Key:    nodeTaintKey,
+								Value:  nodeTaintValue,
+								Effect: v1.TaintEffectNoExecute,
+							},
+						},
+					},
+				},
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeNames[2],
+					},
+				},
 			},
 			success: false,
 		},
 	}
 
 	for _, tc := range tests {
-		actual := PodFitsAnyNode(tc.pod, tc.nodes)
+		actual := PodFitsAnyOtherNode(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,
+										},
+									},
+								},
+							},
+						},
+					},
+				},
+			},
+		},
+	})
+}

pkg/descheduler/pod/pods.go

@@ -17,116 +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
 }
 
-// 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)
+// WithLabelSelector sets a pod label selector
+func (o *Options) WithLabelSelector(labelSelector *metav1.LabelSelector) *Options {
+	o.labelSelector = labelSelector
+	return o
+}
+
+// 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 nil, 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
+}
+
+// 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
+			}
+			if len(pod.Spec.NodeName) == 0 {
+				return []string{}, nil
+			}
+			return []string{pod.Spec.NodeName}, nil
+		},
+	})
+	if err != nil {
+		return nil, err
 	}
 
-	pods := make([]*v1.Pod, 0)
-
-	fieldSelectorString := "spec.nodeName=" + node.Name + ",status.phase!=" + string(v1.PodSucceeded) + ",status.phase!=" + string(v1.PodFailed)
-
-	if len(options.includedNamespaces) > 0 {
-		fieldSelector, err := fields.ParseSelector(fieldSelectorString)
+	// 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 []*v1.Pod{}, err
+			return nil, err
 		}
-
-		for _, namespace := range options.includedNamespaces {
-			podList, err := client.CoreV1().Pods(namespace).List(ctx,
-				metav1.ListOptions{FieldSelector: fieldSelector.String()})
-			if err != nil {
-				return []*v1.Pod{}, err
+		pods := make([]*v1.Pod, 0, len(objs))
+		for _, obj := range objs {
+			pod, ok := obj.(*v1.Pod)
+			if !ok {
+				continue
 			}
-			for i := range podList.Items {
-				if options.filter != nil && !options.filter(&podList.Items[i]) {
-					continue
-				}
-				pods = append(pods, &podList.Items[i])
+			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()})
-	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

@@ -18,15 +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"
 )
 
@@ -38,39 +38,80 @@ 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,
+			expectedPodCount: 2,
+		},
+		{
+			name: "test listing pods with label selector",
+			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{
+				MatchExpressions: []metav1.LabelSelectorRequirement{
+					{
+						Key:      "foo",
+						Operator: metav1.LabelSelectorOpIn,
+						Values:   []string{"bar", "bar1"},
+					},
+				},
+			},
 			expectedPodCount: 2,
 		},
 	}
 	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)
-		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")
@@ -77,28 +78,37 @@ 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...)
 	}
 
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority), evictions.WithNodeFit(nodeFit))
 
 	duplicatePods := make(map[podOwner]map[string][]*v1.Pod)
 	ownerKeyOccurence := make(map[podOwner]int32)
 	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
@@ -183,9 +193,17 @@ func RemoveDuplicatePods(
 	}
 
 	// 1. how many pods can be evicted to respect uniform placement of pods among viable nodes?
-	for ownerKey, nodes := range duplicatePods {
-		upperAvg := int(math.Ceil(float64(ownerKeyOccurence[ownerKey]) / float64(nodeCount)))
-		for nodeName, pods := range nodes {
+	for ownerKey, podNodes := range duplicatePods {
+		targetNodes := getTargetNodes(podNodes, nodes)
+
+		klog.V(2).InfoS("Adjusting feasible nodes", "owner", ownerKey, "from", nodeCount, "to", len(targetNodes))
+		if len(targetNodes) < 2 {
+			klog.V(1).InfoS("Less than two feasible nodes for duplicates to land, skipping eviction", "owner", ownerKey)
+			continue
+		}
+
+		upperAvg := int(math.Ceil(float64(ownerKeyOccurence[ownerKey]) / float64(len(targetNodes))))
+		for nodeName, pods := range podNodes {
 			klog.V(2).InfoS("Average occurrence per node", "node", klog.KObj(nodeMap[nodeName]), "ownerKey", ownerKey, "avg", upperAvg)
 			// list of duplicated pods does not contain the original referential pod
 			if len(pods)+1 > upperAvg {
@@ -202,6 +220,73 @@ func RemoveDuplicatePods(
 	}
 }
 
+func getNodeAffinityNodeSelector(pod *v1.Pod) *v1.NodeSelector {
+	if pod.Spec.Affinity == nil {
+		return nil
+	}
+	if pod.Spec.Affinity.NodeAffinity == nil {
+		return nil
+	}
+	return pod.Spec.Affinity.NodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution
+}
+
+func getTargetNodes(podNodes map[string][]*v1.Pod, nodes []*v1.Node) []*v1.Node {
+	// In order to reduce the number of pods processed, identify pods which have
+	// equal (tolerations, nodeselectors, node affinity) terms and considered them
+	// as identical. Identical pods wrt. (tolerations, nodeselectors, node affinity) terms
+	// will produce the same result when checking if a pod is feasible for a node.
+	// Thus, improving efficiency of processing pods marked for eviction.
+
+	// Collect all distinct pods which differ in at least taints, node affinity or node selector terms
+	distinctPods := map[*v1.Pod]struct{}{}
+	for _, pods := range podNodes {
+		for _, pod := range pods {
+			duplicated := false
+			for dp := range distinctPods {
+				if utils.TolerationsEqual(pod.Spec.Tolerations, dp.Spec.Tolerations) &&
+					utils.NodeSelectorsEqual(getNodeAffinityNodeSelector(pod), getNodeAffinityNodeSelector(dp)) &&
+					reflect.DeepEqual(pod.Spec.NodeSelector, dp.Spec.NodeSelector) {
+					duplicated = true
+					continue
+				}
+			}
+			if duplicated {
+				continue
+			}
+			distinctPods[pod] = struct{}{}
+		}
+	}
+
+	// For each distinct pod get a list of nodes where it can land
+	targetNodesMap := map[string]*v1.Node{}
+	for pod := range distinctPods {
+		matchingNodes := map[string]*v1.Node{}
+		for _, node := range nodes {
+			if !utils.TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, node.Spec.Taints, func(taint *v1.Taint) bool {
+				return taint.Effect == v1.TaintEffectNoSchedule || taint.Effect == v1.TaintEffectNoExecute
+			}) {
+				continue
+			}
+			if match, err := utils.PodMatchNodeSelector(pod, node); err == nil && !match {
+				continue
+			}
+			matchingNodes[node.Name] = node
+		}
+		if len(matchingNodes) > 1 {
+			for nodeName := range matchingNodes {
+				targetNodesMap[nodeName] = matchingNodes[nodeName]
+			}
+		}
+	}
+
+	targetNodes := []*v1.Node{}
+	for _, node := range targetNodesMap {
+		targetNodes = append(targetNodes, node)
+	}
+
+	return targetNodes
+}
+
 func hasExcludedOwnerRefKind(ownerRefs []metav1.OwnerReference, strategy api.DeschedulerStrategy) bool {
 	if strategy.Params == nil || strategy.Params.RemoveDuplicates == nil {
 		return false

pkg/descheduler/strategies/duplicates_test.go

@@ -21,13 +21,16 @@ import (
 	"testing"
 
 	v1 "k8s.io/api/core/v1"
+	policyv1 "k8s.io/api/policy/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"
-	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"
 )
@@ -42,10 +45,28 @@ 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)
+	node3 := test.BuildTestNode("n3", 2000, 3000, 10, func(node *v1.Node) {
+		node.Spec.Taints = []v1.Taint{
+			{
+				Key:    "hardware",
+				Value:  "gpu",
+				Effect: v1.TaintEffectNoSchedule,
+			},
+		}
+	})
+	node4 := test.BuildTestNode("n4", 2000, 3000, 10, func(node *v1.Node) {
+		node.ObjectMeta.Labels = map[string]string{
+			"datacenter": "east",
+		}
+	})
+	node5 := test.BuildTestNode("n5", 2000, 3000, 10, func(node *v1.Node) {
+		node.Spec = v1.NodeSpec{
+			Unschedulable: true,
+		}
+	})
 
 	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
 	p1.Namespace = "dev"
@@ -72,6 +93,14 @@ func TestFindDuplicatePods(t *testing.T) {
 	p13.Namespace = "different-images"
 	p14 := test.BuildTestPod("p14", 100, 0, node1.Name, nil)
 	p14.Namespace = "different-images"
+	p15 := test.BuildTestPod("p15", 100, 0, node1.Name, nil)
+	p15.Namespace = "node-fit"
+	p16 := test.BuildTestPod("NOT1", 100, 0, node1.Name, nil)
+	p16.Namespace = "node-fit"
+	p17 := test.BuildTestPod("NOT2", 100, 0, node1.Name, nil)
+	p17.Namespace = "node-fit"
+	p18 := test.BuildTestPod("TARGET", 100, 0, node1.Name, nil)
+	p18.Namespace = "node-fit"
 
 	// ### Evictable Pods ###
 
@@ -125,94 +154,157 @@ func TestFindDuplicatePods(t *testing.T) {
 		Image: "foo",
 	})
 
+	// ### Pods Evictable Based On Node Fit ###
+
+	ownerRef3 := test.GetReplicaSetOwnerRefList()
+	p15.ObjectMeta.OwnerReferences = ownerRef3
+	p16.ObjectMeta.OwnerReferences = ownerRef3
+	p17.ObjectMeta.OwnerReferences = ownerRef3
+	p18.ObjectMeta.OwnerReferences = ownerRef3
+
+	p15.Spec.NodeSelector = map[string]string{
+		"datacenter": "west",
+	}
+	p16.Spec.NodeSelector = map[string]string{
+		"datacenter": "west",
+	}
+	p17.Spec.NodeSelector = map[string]string{
+		"datacenter": "west",
+	}
+
 	testCases := []struct {
 		description             string
-		maxPodsToEvictPerNode   int
-		pods                    []v1.Pod
-		expectedEvictedPodCount int
+		pods                    []*v1.Pod
+		nodes                   []*v1.Node
+		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.",
+			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.",
+			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.",
+			pods:                    []*v1.Pod{p1, p2, p3},
+			nodes:                   []*v1.Node{node1, node5},
+			expectedEvictedPodCount: 0,
+			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,
-				[]*v1.Node{node1, node2},
+				nil,
+				nil,
+				testCase.nodes,
+				false,
+				false,
+				false,
+				false,
 				false,
 			)
 
-			RemoveDuplicatePods(ctx, fakeClient, testCase.strategy, []*v1.Node{node1, node2}, 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)
@@ -223,8 +315,6 @@ 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"},
@@ -237,27 +327,137 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 		}
 	}
 
+	setTolerationsK1 := func(pod *v1.Pod) {
+		test.SetRSOwnerRef(pod)
+		pod.Spec.Tolerations = []v1.Toleration{
+			{
+				Key:      "k1",
+				Value:    "v1",
+				Operator: v1.TolerationOpEqual,
+				Effect:   v1.TaintEffectNoSchedule,
+			},
+		}
+	}
+	setTolerationsK2 := func(pod *v1.Pod) {
+		test.SetRSOwnerRef(pod)
+		pod.Spec.Tolerations = []v1.Toleration{
+			{
+				Key:      "k2",
+				Value:    "v2",
+				Operator: v1.TolerationOpEqual,
+				Effect:   v1.TaintEffectNoSchedule,
+			},
+		}
+	}
+
+	setMasterNoScheduleTaint := func(node *v1.Node) {
+		node.Spec.Taints = []v1.Taint{
+			{
+				Effect: v1.TaintEffectNoSchedule,
+				Key:    "node-role.kubernetes.io/master",
+			},
+		}
+	}
+
+	setMasterNoScheduleLabel := func(node *v1.Node) {
+		if node.ObjectMeta.Labels == nil {
+			node.ObjectMeta.Labels = map[string]string{}
+		}
+		node.ObjectMeta.Labels["node-role.kubernetes.io/master"] = ""
+	}
+
+	setWorkerLabel := func(node *v1.Node) {
+		if node.ObjectMeta.Labels == nil {
+			node.ObjectMeta.Labels = map[string]string{}
+		}
+		node.ObjectMeta.Labels["node-role.kubernetes.io/worker"] = "k1"
+		node.ObjectMeta.Labels["node-role.kubernetes.io/worker"] = "k2"
+	}
+
+	setNotMasterNodeSelectorK1 := func(pod *v1.Pod) {
+		test.SetRSOwnerRef(pod)
+		pod.Spec.Affinity = &v1.Affinity{
+			NodeAffinity: &v1.NodeAffinity{
+				RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
+					NodeSelectorTerms: []v1.NodeSelectorTerm{
+						{
+							MatchExpressions: []v1.NodeSelectorRequirement{
+								{
+									Key:      "node-role.kubernetes.io/master",
+									Operator: v1.NodeSelectorOpDoesNotExist,
+								},
+								{
+									Key:      "k1",
+									Operator: v1.NodeSelectorOpDoesNotExist,
+								},
+							},
+						},
+					},
+				},
+			},
+		}
+	}
+
+	setNotMasterNodeSelectorK2 := func(pod *v1.Pod) {
+		test.SetRSOwnerRef(pod)
+		pod.Spec.Affinity = &v1.Affinity{
+			NodeAffinity: &v1.NodeAffinity{
+				RequiredDuringSchedulingIgnoredDuringExecution: &v1.NodeSelector{
+					NodeSelectorTerms: []v1.NodeSelectorTerm{
+						{
+							MatchExpressions: []v1.NodeSelectorRequirement{
+								{
+									Key:      "node-role.kubernetes.io/master",
+									Operator: v1.NodeSelectorOpDoesNotExist,
+								},
+								{
+									Key:      "k2",
+									Operator: v1.NodeSelectorOpDoesNotExist,
+								},
+							},
+						},
+					},
+				},
+			},
+		}
+	}
+
+	setWorkerLabelSelectorK1 := func(pod *v1.Pod) {
+		test.SetRSOwnerRef(pod)
+		if pod.Spec.NodeSelector == nil {
+			pod.Spec.NodeSelector = map[string]string{}
+		}
+		pod.Spec.NodeSelector["node-role.kubernetes.io/worker"] = "k1"
+	}
+
+	setWorkerLabelSelectorK2 := func(pod *v1.Pod) {
+		test.SetRSOwnerRef(pod)
+		if pod.Spec.NodeSelector == nil {
+			pod.Spec.NodeSelector = map[string]string{}
+		}
+		pod.Spec.NodeSelector["node-role.kubernetes.io/worker"] = "k2"
+	}
+
 	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{
@@ -269,17 +469,17 @@ 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{
@@ -290,17 +490,17 @@ 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{
@@ -312,27 +512,27 @@ 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{
@@ -344,10 +544,10 @@ 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{
@@ -359,14 +559,14 @@ 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{
@@ -378,9 +578,9 @@ 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{
@@ -390,24 +590,148 @@ func TestRemoveDuplicatesUniformly(t *testing.T) {
 			},
 			strategy: api.DeschedulerStrategy{},
 		},
+		{
+			description: "Evict pods uniformly respecting taints",
+			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),
+			},
+			expectedEvictedPodCount: 2,
+			nodes: []*v1.Node{
+				test.BuildTestNode("worker1", 2000, 3000, 10, nil),
+				test.BuildTestNode("worker2", 2000, 3000, 10, nil),
+				test.BuildTestNode("worker3", 2000, 3000, 10, nil),
+				test.BuildTestNode("master1", 2000, 3000, 10, setMasterNoScheduleTaint),
+				test.BuildTestNode("master2", 2000, 3000, 10, setMasterNoScheduleTaint),
+				test.BuildTestNode("master3", 2000, 3000, 10, setMasterNoScheduleTaint),
+			},
+			strategy: api.DeschedulerStrategy{},
+		},
+		{
+			description: "Evict pods uniformly respecting RequiredDuringSchedulingIgnoredDuringExecution node affinity",
+			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),
+			},
+			expectedEvictedPodCount: 2,
+			nodes: []*v1.Node{
+				test.BuildTestNode("worker1", 2000, 3000, 10, nil),
+				test.BuildTestNode("worker2", 2000, 3000, 10, nil),
+				test.BuildTestNode("worker3", 2000, 3000, 10, nil),
+				test.BuildTestNode("master1", 2000, 3000, 10, setMasterNoScheduleLabel),
+				test.BuildTestNode("master2", 2000, 3000, 10, setMasterNoScheduleLabel),
+				test.BuildTestNode("master3", 2000, 3000, 10, setMasterNoScheduleLabel),
+			},
+			strategy: api.DeschedulerStrategy{},
+		},
+		{
+			description: "Evict pods uniformly respecting node selector",
+			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),
+			},
+			expectedEvictedPodCount: 2,
+			nodes: []*v1.Node{
+				test.BuildTestNode("worker1", 2000, 3000, 10, setWorkerLabel),
+				test.BuildTestNode("worker2", 2000, 3000, 10, setWorkerLabel),
+				test.BuildTestNode("worker3", 2000, 3000, 10, setWorkerLabel),
+				test.BuildTestNode("master1", 2000, 3000, 10, nil),
+				test.BuildTestNode("master2", 2000, 3000, 10, nil),
+				test.BuildTestNode("master3", 2000, 3000, 10, nil),
+			},
+			strategy: api.DeschedulerStrategy{},
+		},
+		{
+			description: "Evict pods uniformly respecting node selector with zero target nodes",
+			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),
+			},
+			expectedEvictedPodCount: 0,
+			nodes: []*v1.Node{
+				test.BuildTestNode("worker1", 2000, 3000, 10, nil),
+				test.BuildTestNode("worker2", 2000, 3000, 10, nil),
+				test.BuildTestNode("worker3", 2000, 3000, 10, nil),
+				test.BuildTestNode("master1", 2000, 3000, 10, nil),
+				test.BuildTestNode("master2", 2000, 3000, 10, nil),
+				test.BuildTestNode("master3", 2000, 3000, 10, nil),
+			},
+			strategy: api.DeschedulerStrategy{},
+		},
 	}
 
 	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",
+				policyv1.SchemeGroupVersion.String(),
 				false,
-				testCase.maxPodsToEvictPerNode,
+				nil,
+				nil,
 				testCase.nodes,
 				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 +1,343 @@
+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, 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:             "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,
+				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/lownodeutilization_test.go

@@ -1,776 +0,0 @@
-/*
-Copyright 2017 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 strategies
-
-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/kubernetes/fake"
-	core "k8s.io/client-go/testing"
-	"sigs.k8s.io/descheduler/pkg/api"
-	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
-	"sigs.k8s.io/descheduler/pkg/utils"
-	"sigs.k8s.io/descheduler/test"
-)
-
-var (
-	lowPriority  = int32(0)
-	highPriority = int32(10000)
-)
-
-func TestLowNodeUtilization(t *testing.T) {
-	ctx := context.Background()
-	n1NodeName := "n1"
-	n2NodeName := "n2"
-	n3NodeName := "n3"
-
-	testCases := []struct {
-		name                         string
-		thresholds, targetThresholds api.ResourceThresholds
-		nodes                        map[string]*v1.Node
-		pods                         map[string]*v1.PodList
-		maxPodsToEvictPerNode        int
-		expectedPodsEvicted          int
-		evictedPods                  []string
-	}{
-		{
-			name: "no evictable pods",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  30,
-				v1.ResourcePods: 30,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  50,
-				v1.ResourcePods: 50,
-			},
-			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, test.SetNodeUnschedulable),
-			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						// These won't be evicted.
-						*test.BuildTestPod("p1", 400, 0, n1NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p3", 400, 0, n1NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p4", 400, 0, n1NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p5", 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("p6", 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{
-						*test.BuildTestPod("p9", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {},
-			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   0,
-		},
-		{
-			name: "without priorities",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  30,
-				v1.ResourcePods: 30,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  50,
-				v1.ResourcePods: 50,
-			},
-			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, test.SetNodeUnschedulable),
-			},
-			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.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 0, n1NodeName, test.SetRSOwnerRef),
-						// These won't be evicted.
-						*test.BuildTestPod("p6", 400, 0, n1NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p7", 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("p8", 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{
-						*test.BuildTestPod("p9", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {},
-			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   4,
-		},
-		{
-			name: "without priorities stop when cpu capacity is depleted",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  30,
-				v1.ResourcePods: 30,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  50,
-				v1.ResourcePods: 50,
-			},
-			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, test.SetNodeUnschedulable),
-			},
-			pods: map[string]*v1.PodList{
-				n1NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p1", 400, 300, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p2", 400, 300, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p3", 400, 300, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p4", 400, 300, n1NodeName, test.SetRSOwnerRef),
-						*test.BuildTestPod("p5", 400, 300, n1NodeName, test.SetRSOwnerRef),
-						// These won't be evicted.
-						*test.BuildTestPod("p6", 400, 300, n1NodeName, test.SetDSOwnerRef),
-						*test.BuildTestPod("p7", 400, 300, 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("p8", 400, 300, n1NodeName, func(pod *v1.Pod) {
-							// A Critical Pod.
-							pod.Namespace = "kube-system"
-							priority := utils.SystemCriticalPriority
-							pod.Spec.Priority = &priority
-						}),
-					},
-				},
-				n2NodeName: {
-					Items: []v1.Pod{
-						*test.BuildTestPod("p9", 400, 2100, n2NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {},
-			},
-			maxPodsToEvictPerNode: 0,
-			// 4 pods available for eviction based on v1.ResourcePods, only 3 pods can be evicted before cpu is depleted
-			expectedPodsEvicted: 3,
-		},
-		{
-			name: "with priorities",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  30,
-				v1.ResourcePods: 30,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  50,
-				v1.ResourcePods: 50,
-			},
-			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, 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, highPriority)
-						}),
-						*test.BuildTestPod("p2", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodPriority(pod, highPriority)
-						}),
-						*test.BuildTestPod("p3", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodPriority(pod, highPriority)
-						}),
-						*test.BuildTestPod("p4", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodPriority(pod, highPriority)
-						}),
-						*test.BuildTestPod("p5", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.SetPodPriority(pod, lowPriority)
-						}),
-						// These won't be evicted.
-						*test.BuildTestPod("p6", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetDSOwnerRef(pod)
-							test.SetPodPriority(pod, highPriority)
-						}),
-						*test.BuildTestPod("p7", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							// A pod with local storage.
-							test.SetNormalOwnerRef(pod)
-							test.SetPodPriority(pod, lowPriority)
-							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("p8", 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{
-						*test.BuildTestPod("p9", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {},
-			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   4,
-		},
-		{
-			name: "without priorities evicting best-effort pods only",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  30,
-				v1.ResourcePods: 30,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  50,
-				v1.ResourcePods: 50,
-			},
-			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, 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)
-							test.MakeBestEffortPod(pod)
-						}),
-						*test.BuildTestPod("p3", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-						}),
-						*test.BuildTestPod("p4", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.MakeBestEffortPod(pod)
-						}),
-						*test.BuildTestPod("p5", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetRSOwnerRef(pod)
-							test.MakeBestEffortPod(pod)
-						}),
-						// These won't be evicted.
-						*test.BuildTestPod("p6", 400, 0, n1NodeName, func(pod *v1.Pod) {
-							test.SetDSOwnerRef(pod)
-						}),
-						*test.BuildTestPod("p7", 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("p8", 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{
-						*test.BuildTestPod("p9", 400, 0, n1NodeName, test.SetRSOwnerRef),
-					},
-				},
-				n3NodeName: {},
-			},
-			maxPodsToEvictPerNode: 0,
-			expectedPodsEvicted:   4,
-			evictedPods:           []string{"p1", "p2", "p4", "p5"},
-		},
-	}
-
-	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())
-			})
-			podsForEviction := make(map[string]struct{})
-			for _, pod := range test.evictedPods {
-				podsForEviction[pod] = struct{}{}
-			}
-
-			evictionFailed := false
-			if len(test.evictedPods) > 0 {
-				fakeClient.Fake.AddReactor("create", "pods", func(action core.Action) (bool, runtime.Object, error) {
-					getAction := action.(core.CreateAction)
-					obj := getAction.GetObject()
-					if eviction, ok := obj.(*v1beta1.Eviction); ok {
-						if _, exists := podsForEviction[eviction.Name]; exists {
-							return true, obj, nil
-						}
-						evictionFailed = true
-						return true, nil, fmt.Errorf("pod %q was unexpectedly evicted", eviction.Name)
-					}
-					return true, obj, nil
-				})
-			}
-
-			var nodes []*v1.Node
-			for _, node := range test.nodes {
-				nodes = append(nodes, node)
-			}
-
-			podEvictor := evictions.NewPodEvictor(
-				fakeClient,
-				"v1",
-				false,
-				test.maxPodsToEvictPerNode,
-				nodes,
-				false,
-			)
-
-			strategy := api.DeschedulerStrategy{
-				Enabled: true,
-				Params: &api.StrategyParameters{
-					NodeResourceUtilizationThresholds: &api.NodeResourceUtilizationThresholds{
-						Thresholds:       test.thresholds,
-						TargetThresholds: test.targetThresholds,
-					},
-				},
-			}
-			LowNodeUtilization(ctx, fakeClient, strategy, nodes, podEvictor)
-
-			podsEvicted := podEvictor.TotalEvicted()
-			if test.expectedPodsEvicted != podsEvicted {
-				t.Errorf("Expected %#v pods to be evicted but %#v got evicted", test.expectedPodsEvicted, podsEvicted)
-			}
-			if evictionFailed {
-				t.Errorf("Pod evictions failed unexpectedly")
-			}
-		})
-	}
-}
-
-func TestValidateStrategyConfig(t *testing.T) {
-	tests := []struct {
-		name             string
-		thresholds       api.ResourceThresholds
-		targetThresholds api.ResourceThresholds
-		errInfo          error
-	}{
-		{
-			name: "passing invalid thresholds",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    20,
-				v1.ResourceMemory: 120,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    80,
-				v1.ResourceMemory: 80,
-			},
-			errInfo: fmt.Errorf("thresholds config is not valid: %v", fmt.Errorf(
-				"%v threshold not in [%v, %v] range", v1.ResourceMemory, MinResourcePercentage, MaxResourcePercentage)),
-		},
-		{
-			name: "passing invalid targetThresholds",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    20,
-				v1.ResourceMemory: 20,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    80,
-				"resourceInvalid": 80,
-			},
-			errInfo: fmt.Errorf("targetThresholds config is not valid: %v",
-				fmt.Errorf("only cpu, memory, or pods thresholds can be specified")),
-		},
-		{
-			name: "thresholds and targetThresholds configured different num of resources",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    20,
-				v1.ResourceMemory: 20,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    80,
-				v1.ResourceMemory: 80,
-				v1.ResourcePods:   80,
-			},
-			errInfo: fmt.Errorf("thresholds and targetThresholds configured different resources"),
-		},
-		{
-			name: "thresholds and targetThresholds configured different resources",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    20,
-				v1.ResourceMemory: 20,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:  80,
-				v1.ResourcePods: 80,
-			},
-			errInfo: fmt.Errorf("thresholds and targetThresholds configured different resources"),
-		},
-		{
-			name: "thresholds' CPU config value is greater than targetThresholds'",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    90,
-				v1.ResourceMemory: 20,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    80,
-				v1.ResourceMemory: 80,
-			},
-			errInfo: fmt.Errorf("thresholds' %v percentage is greater than targetThresholds'", v1.ResourceCPU),
-		},
-		{
-			name: "passing valid strategy config",
-			thresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    20,
-				v1.ResourceMemory: 20,
-			},
-			targetThresholds: api.ResourceThresholds{
-				v1.ResourceCPU:    80,
-				v1.ResourceMemory: 80,
-			},
-			errInfo: nil,
-		},
-	}
-
-	for _, testCase := range tests {
-		validateErr := validateStrategyConfig(testCase.thresholds, testCase.targetThresholds)
-
-		if validateErr == nil || testCase.errInfo == nil {
-			if validateErr != testCase.errInfo {
-				t.Errorf("expected validity of strategy config: thresholds %#v targetThresholds %#v to be %v but got %v instead",
-					testCase.thresholds, testCase.targetThresholds, testCase.errInfo, validateErr)
-			}
-		} else if validateErr.Error() != testCase.errInfo.Error() {
-			t.Errorf("expected validity of strategy config: thresholds %#v targetThresholds %#v to be %v but got %v instead",
-				testCase.thresholds, testCase.targetThresholds, testCase.errInfo, validateErr)
-		}
-	}
-}
-
-func TestValidateThresholds(t *testing.T) {
-	tests := []struct {
-		name    string
-		input   api.ResourceThresholds
-		errInfo error
-	}{
-		{
-			name:    "passing nil map for threshold",
-			input:   nil,
-			errInfo: fmt.Errorf("no resource threshold is configured"),
-		},
-		{
-			name:    "passing no threshold",
-			input:   api.ResourceThresholds{},
-			errInfo: fmt.Errorf("no resource threshold is configured"),
-		},
-		{
-			name: "passing unsupported resource name",
-			input: api.ResourceThresholds{
-				v1.ResourceCPU:     40,
-				v1.ResourceStorage: 25.5,
-			},
-			errInfo: fmt.Errorf("only cpu, memory, or pods thresholds can be specified"),
-		},
-		{
-			name: "passing invalid resource name",
-			input: api.ResourceThresholds{
-				v1.ResourceCPU: 40,
-				"coolResource": 42.0,
-			},
-			errInfo: fmt.Errorf("only cpu, memory, or pods thresholds can be specified"),
-		},
-		{
-			name: "passing invalid resource value",
-			input: api.ResourceThresholds{
-				v1.ResourceCPU:    110,
-				v1.ResourceMemory: 80,
-			},
-			errInfo: fmt.Errorf("%v threshold not in [%v, %v] range", v1.ResourceCPU, MinResourcePercentage, MaxResourcePercentage),
-		},
-		{
-			name: "passing a valid threshold with max and min resource value",
-			input: api.ResourceThresholds{
-				v1.ResourceCPU:    100,
-				v1.ResourceMemory: 0,
-			},
-			errInfo: nil,
-		},
-		{
-			name: "passing a valid threshold with only cpu",
-			input: api.ResourceThresholds{
-				v1.ResourceCPU: 80,
-			},
-			errInfo: nil,
-		},
-		{
-			name: "passing a valid threshold with cpu, memory and pods",
-			input: api.ResourceThresholds{
-				v1.ResourceCPU:    20,
-				v1.ResourceMemory: 30,
-				v1.ResourcePods:   40,
-			},
-			errInfo: nil,
-		},
-	}
-
-	for _, test := range tests {
-		validateErr := validateThresholds(test.input)
-
-		if validateErr == nil || test.errInfo == nil {
-			if validateErr != test.errInfo {
-				t.Errorf("expected validity of threshold: %#v to be %v but got %v instead", test.input, test.errInfo, validateErr)
-			}
-		} else if validateErr.Error() != test.errInfo.Error() {
-			t.Errorf("expected validity of threshold: %#v to be %v but got %v instead", test.input, test.errInfo, validateErr)
-		}
-	}
-}
-
-func TestWithTaints(t *testing.T) {
-	ctx := context.Background()
-	strategy := api.DeschedulerStrategy{
-		Enabled: true,
-		Params: &api.StrategyParameters{
-			NodeResourceUtilizationThresholds: &api.NodeResourceUtilizationThresholds{
-				Thresholds: api.ResourceThresholds{
-					v1.ResourcePods: 20,
-				},
-				TargetThresholds: api.ResourceThresholds{
-					v1.ResourcePods: 70,
-				},
-			},
-		},
-	}
-
-	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
-	n2 := test.BuildTestNode("n2", 1000, 3000, 10, nil)
-	n3 := test.BuildTestNode("n3", 1000, 3000, 10, nil)
-	n3withTaints := n3.DeepCopy()
-	n3withTaints.Spec.Taints = []v1.Taint{
-		{
-			Key:    "key",
-			Value:  "value",
-			Effect: v1.TaintEffectNoSchedule,
-		},
-	}
-
-	podThatToleratesTaint := test.BuildTestPod("tolerate_pod", 200, 0, n1.Name, test.SetRSOwnerRef)
-	podThatToleratesTaint.Spec.Tolerations = []v1.Toleration{
-		{
-			Key:   "key",
-			Value: "value",
-		},
-	}
-
-	tests := []struct {
-		name              string
-		nodes             []*v1.Node
-		pods              []*v1.Pod
-		evictionsExpected int
-	}{
-		{
-			name:  "No taints",
-			nodes: []*v1.Node{n1, n2, n3},
-			pods: []*v1.Pod{
-				//Node 1 pods
-				test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_2_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_3_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_4_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_5_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_6_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_7_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_8_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				// Node 2 pods
-				test.BuildTestPod(fmt.Sprintf("pod_9_%s", n2.Name), 200, 0, n2.Name, test.SetRSOwnerRef),
-			},
-			evictionsExpected: 1,
-		},
-		{
-			name:  "No pod tolerates node taint",
-			nodes: []*v1.Node{n1, n3withTaints},
-			pods: []*v1.Pod{
-				//Node 1 pods
-				test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_2_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_3_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_4_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_5_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_6_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_7_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_8_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				// Node 3 pods
-				test.BuildTestPod(fmt.Sprintf("pod_9_%s", n3withTaints.Name), 200, 0, n3withTaints.Name, test.SetRSOwnerRef),
-			},
-			evictionsExpected: 0,
-		},
-		{
-			name:  "Pod which tolerates node taint",
-			nodes: []*v1.Node{n1, n3withTaints},
-			pods: []*v1.Pod{
-				//Node 1 pods
-				test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_2_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_3_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_4_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_5_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_6_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				test.BuildTestPod(fmt.Sprintf("pod_7_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
-				podThatToleratesTaint,
-				// Node 3 pods
-				test.BuildTestPod(fmt.Sprintf("pod_9_%s", n3withTaints.Name), 200, 0, n3withTaints.Name, test.SetRSOwnerRef),
-			},
-			evictionsExpected: 1,
-		},
-	}
-
-	for _, item := range tests {
-		t.Run(item.name, func(t *testing.T) {
-			var objs []runtime.Object
-			for _, node := range item.nodes {
-				objs = append(objs, node)
-			}
-
-			for _, pod := range item.pods {
-				objs = append(objs, pod)
-			}
-
-			fakeClient := fake.NewSimpleClientset(objs...)
-
-			podEvictor := evictions.NewPodEvictor(
-				fakeClient,
-				"policy/v1",
-				false,
-				item.evictionsExpected,
-				item.nodes,
-				false,
-			)
-
-			LowNodeUtilization(ctx, fakeClient, strategy, item.nodes, podEvictor)
-
-			if item.evictionsExpected != podEvictor.TotalEvicted() {
-				t.Errorf("Expected %v evictions, got %v", item.evictionsExpected, podEvictor.TotalEvicted())
-			}
-		})
-	}
-}

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
@@ -58,13 +59,28 @@ 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...)
 	}
 
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	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)
@@ -75,16 +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)
 					}),
-					podutil.WithNamespaces(includedNamespaces),
-					podutil.WithoutNamespaces(excludedNamespaces),
 				)
 				if err != nil {
 					klog.ErrorS(err, "Failed to get pods", "node", klog.KObj(node))
@@ -104,5 +117,4 @@ func RemovePodsViolatingNodeAffinity(ctx context.Context, client clientset.Inter
 			klog.ErrorS(nil, "Invalid nodeAffinityType", "nodeAffinity", nodeAffinity)
 		}
 	}
-	klog.V(1).InfoS("Number of evicted pods", "totalEvicted", podEvictor.TotalEvicted())
 }

pkg/descheduler/strategies/node_affinity_test.go

@@ -20,17 +20,20 @@ import (
 	"context"
 	"testing"
 
-	"k8s.io/api/core/v1"
+	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{
@@ -40,6 +43,16 @@ func TestRemovePodsViolatingNodeAffinity(t *testing.T) {
 		},
 	}
 
+	requiredDuringSchedulingIgnoredDuringExecutionWithNodeFitStrategy := api.DeschedulerStrategy{
+		Enabled: true,
+		Params: &api.StrategyParameters{
+			NodeAffinityType: []string{
+				"requiredDuringSchedulingIgnoredDuringExecution",
+			},
+			NodeFit: true,
+		},
+	}
+
 	nodeLabelKey := "kubernetes.io/desiredNode"
 	nodeLabelValue := "yes"
 	nodeWithLabels := test.BuildTestNode("nodeWithLabels", 2000, 3000, 10, nil)
@@ -48,10 +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{
@@ -80,20 +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",
@@ -106,64 +124,116 @@ 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:                requiredDuringSchedulingIgnoredDuringExecutionStrategy,
-			pods:                    addPodsToNode(nodeWithoutLabels),
+			strategy:                requiredDuringSchedulingIgnoredDuringExecutionWithNodeFitStrategy,
+			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, nil),
+			nodes:                   []*v1.Node{nodeWithLabels, unschedulableNodeWithLabels},
+			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,
+				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,
-			"v1",
-			false,
-			tc.maxPodsToEvictPerNode,
-			tc.nodes,
-			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

@@ -20,12 +20,14 @@ 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"
 	"sigs.k8s.io/descheduler/pkg/utils"
 
 	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
 )
@@ -47,16 +49,20 @@ 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
-	if strategy.Params != nil && strategy.Params.Namespaces != nil {
-		includedNamespaces = strategy.Params.Namespaces.Include
-		excludedNamespaces = strategy.Params.Namespaces.Exclude
+	var includedNamespaces, excludedNamespaces sets.String
+	var labelSelector *metav1.LabelSelector
+	if strategy.Params != nil {
+		if strategy.Params.Namespaces != nil {
+			includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+			excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
+		}
+		labelSelector = strategy.Params.LabelSelector
 	}
 
 	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, strategy.Params)
@@ -65,18 +71,27 @@ func RemovePodsViolatingNodeTaints(ctx context.Context, client clientset.Interfa
 		return
 	}
 
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	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
+	}
 
 	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.ListAllPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
 		if err != nil {
 			//no pods evicted as error encountered retrieving evictable Pods
 			return

pkg/descheduler/strategies/node_taint_test.go

@@ -5,13 +5,16 @@ import (
 	"fmt"
 	"testing"
 
-	"k8s.io/api/core/v1"
+	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/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
@@ -44,11 +47,21 @@ func addTolerationToPod(pod *v1.Pod, key, value string, index int) *v1.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{
+			"datacenter": "east",
+		}
+	})
+	node4 := test.BuildTestNode("n4", 2000, 3000, 10, func(node *v1.Node) {
+		node.Spec = v1.NodeSpec{
+			Unschedulable: true,
+		}
+	})
 
 	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
 	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, nil)
@@ -69,12 +82,15 @@ func TestDeletePodsViolatingNodeTaints(t *testing.T) {
 	p10 := test.BuildTestPod("p10", 100, 0, node2.Name, nil)
 	p11 := test.BuildTestPod("p11", 100, 0, node2.Name, nil)
 	p11.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
+	p12 := test.BuildTestPod("p11", 100, 0, node2.Name, nil)
+	p12.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 
 	// The following 4 pods won't get evicted.
 	// A Critical Pod.
 	p7.Namespace = "kube-system"
 	priority := utils.SystemCriticalPriority
 	p7.Spec.Priority = &priority
+	p7.ObjectMeta.OwnerReferences = test.GetNormalPodOwnerRefList()
 
 	// A daemonset.
 	p8.ObjectMeta.OwnerReferences = test.GetDaemonSetOwnerRefList()
@@ -97,89 +113,172 @@ func TestDeletePodsViolatingNodeTaints(t *testing.T) {
 	p3 = addTolerationToPod(p3, "testTaint", "test", 1)
 	p4 = addTolerationToPod(p4, "testTaintX", "testX", 1)
 
+	p12.Spec.NodeSelector = map[string]string{
+		"datacenter": "west",
+	}
+
+	var uint1 uint = 1
+
 	tests := []struct {
-		description             string
-		nodes                   []*v1.Node
-		pods                    []v1.Pod
-		evictLocalStoragePods   bool
-		maxPodsToEvictPerNode   int
-		expectedEvictedPodCount int
+		description                    string
+		nodes                          []*v1.Node
+		pods                           []*v1.Pod
+		evictLocalStoragePods          bool
+		evictSystemCriticalPods        bool
+		maxPodsToEvictPerNode          *uint
+		maxNoOfPodsToEvictPerNamespace *uint
+		expectedEvictedPodCount        uint
+		nodeFit                        bool
 	}{
 
 		{
 			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,
-			maxPodsToEvictPerNode:   0,
+			evictSystemCriticalPods: false,
 			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,
-			maxPodsToEvictPerNode:   0,
+			evictSystemCriticalPods: false,
 			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,
-			maxPodsToEvictPerNode:   1,
+			evictSystemCriticalPods: false,
+			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,
-			maxPodsToEvictPerNode:   0,
-			expectedEvictedPodCount: 0,
+			evictSystemCriticalPods: false,
+			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,
-			maxPodsToEvictPerNode:   0,
-			expectedEvictedPodCount: 1,
+			evictSystemCriticalPods: false,
+			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,
-			maxPodsToEvictPerNode:   0,
-			expectedEvictedPodCount: 1,
+			evictSystemCriticalPods: false,
+			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{node1, node2},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: true,
+			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},
+			nodes:                   []*v1.Node{node1, node2},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			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},
+			nodes:                   []*v1.Node{node1, node3},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			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},
+			nodes:                   []*v1.Node{node1, node4},
+			evictLocalStoragePods:   false,
+			evictSystemCriticalPods: false,
+			expectedEvictedPodCount: 0, //p2 gets 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,
+				tc.evictLocalStoragePods,
+				tc.evictSystemCriticalPods,
+				false,
+				false,
+				false,
+			)
+
+			strategy := api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: tc.nodeFit,
+				},
+			}
+
+			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,
-			"v1",
-			false,
-			tc.maxPodsToEvictPerNode,
-			tc.nodes,
-			tc.evictLocalStoragePods,
-		)
-
-		RemovePodsViolatingNodeTaints(ctx, fakeClient, api.DeschedulerStrategy{}, 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

@@ -0,0 +1,166 @@
+/*
+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 nodeutilization
+
+import (
+	"context"
+	"fmt"
+
+	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"
+	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, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc) {
+	if err := validateNodeUtilizationParams(strategy.Params); err != nil {
+		klog.ErrorS(err, "Invalid HighNodeUtilization parameters")
+		return
+	}
+
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, strategy.Params)
+	if err != nil {
+		klog.ErrorS(err, "Failed to get threshold priority from strategy's params")
+		return
+	}
+
+	thresholds := strategy.Params.NodeResourceUtilizationThresholds.Thresholds
+	targetThresholds := strategy.Params.NodeResourceUtilizationThresholds.TargetThresholds
+	if err := validateHighUtilizationStrategyConfig(thresholds, targetThresholds); err != nil {
+		klog.ErrorS(err, "HighNodeUtilization config is not valid")
+		return
+	}
+	targetThresholds = make(api.ResourceThresholds)
+
+	setDefaultForThresholds(thresholds, targetThresholds)
+	resourceNames := getResourceNames(targetThresholds)
+
+	sourceNodes, highNodes := classifyNodes(
+		getNodeUsage(nodes, thresholds, targetThresholds, resourceNames, getPodsAssignedToNode),
+		func(node *v1.Node, usage NodeUsage) bool {
+			return isNodeWithLowUtilization(usage)
+		},
+		func(node *v1.Node, usage NodeUsage) bool {
+			if nodeutil.IsNodeUnschedulable(node) {
+				klog.V(2).InfoS("Node is unschedulable", "node", klog.KObj(node))
+				return false
+			}
+			return !isNodeWithLowUtilization(usage)
+		})
+
+	// log message in one line
+	keysAndValues := []interface{}{
+		"CPU", thresholds[v1.ResourceCPU],
+		"Mem", thresholds[v1.ResourceMemory],
+		"Pods", thresholds[v1.ResourcePods],
+	}
+	for name := range thresholds {
+		if !isBasicResource(name) {
+			keysAndValues = append(keysAndValues, string(name), int64(thresholds[name]))
+		}
+	}
+
+	klog.V(1).InfoS("Criteria for a node below target utilization", keysAndValues...)
+	klog.V(1).InfoS("Number of underutilized nodes", "totalNumber", len(sourceNodes))
+
+	if len(sourceNodes) == 0 {
+		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 or equal than NumberOfNodes, nothing to do here", "underutilizedNodes", len(sourceNodes), "numberOfNodes", strategy.Params.NodeResourceUtilizationThresholds.NumberOfNodes)
+		return
+	}
+	if len(sourceNodes) == len(nodes) {
+		klog.V(1).InfoS("All nodes are underutilized, nothing to do here")
+		return
+	}
+	if len(highNodes) == 0 {
+		klog.V(1).InfoS("No node is available to schedule the pods, nothing to do here")
+		return
+	}
+
+	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 {
+		for name := range totalAvailableUsage {
+			if totalAvailableUsage[name].CmpInt64(0) < 1 {
+				return false
+			}
+		}
+
+		return true
+	}
+	evictPodsFromSourceNodes(
+		ctx,
+		sourceNodes,
+		highNodes,
+		podEvictor,
+		evictable.IsEvictable,
+		resourceNames,
+		"HighNodeUtilization",
+		continueEvictionCond)
+
+}
+
+func validateHighUtilizationStrategyConfig(thresholds, targetThresholds api.ResourceThresholds) error {
+	if targetThresholds != nil {
+		return fmt.Errorf("targetThresholds is not applicable for HighNodeUtilization")
+	}
+	if err := validateThresholds(thresholds); err != nil {
+		return fmt.Errorf("thresholds config is not valid: %v", err)
+	}
+	return nil
+}
+
+func setDefaultForThresholds(thresholds, targetThresholds api.ResourceThresholds) {
+	// check if Pods/CPU/Mem are set, if not, set them to 100
+	if _, ok := thresholds[v1.ResourcePods]; !ok {
+		thresholds[v1.ResourcePods] = MaxResourcePercentage
+	}
+	if _, ok := thresholds[v1.ResourceCPU]; !ok {
+		thresholds[v1.ResourceCPU] = MaxResourcePercentage
+	}
+	if _, ok := thresholds[v1.ResourceMemory]; !ok {
+		thresholds[v1.ResourceMemory] = MaxResourcePercentage
+	}
+
+	// Default targetThreshold resource values to 100
+	targetThresholds[v1.ResourcePods] = MaxResourcePercentage
+	targetThresholds[v1.ResourceCPU] = MaxResourcePercentage
+	targetThresholds[v1.ResourceMemory] = MaxResourcePercentage
+
+	for name := range thresholds {
+		if !isBasicResource(name) {
+			targetThresholds[name] = MaxResourcePercentage
+		}
+	}
+}

pkg/descheduler/strategies/nodeutilization/highnodeutilization_test.go

@@ -0,0 +1,685 @@
+/*
+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 nodeutilization
+
+import (
+	"context"
+	"fmt"
+	"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) {
+	n1NodeName := "n1"
+	n2NodeName := "n2"
+	n3NodeName := "n3"
+
+	nodeSelectorKey := "datacenter"
+	nodeSelectorValue := "west"
+
+	testCases := []struct {
+		name                string
+		thresholds          api.ResourceThresholds
+		nodes               []*v1.Node
+		pods                []*v1.Pod
+		expectedPodsEvicted uint
+		evictedPods         []string
+	}{
+		{
+			name: "no node below threshold usage",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  20,
+				v1.ResourcePods: 20,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 0,
+		},
+		{
+			name: "no evictable pods",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  40,
+				v1.ResourcePods: 40,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 0,
+		},
+		{
+			name: "no node to schedule evicted pods",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  20,
+				v1.ResourcePods: 20,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 0,
+		},
+		{
+			name: "without priorities",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  30,
+				v1.ResourcePods: 30,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 2,
+			evictedPods:         []string{"p1", "p7"},
+		},
+		{
+			name: "without priorities stop when resource capacity is depleted",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  30,
+				v1.ResourcePods: 30,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 1,
+		},
+		{
+			name: "with priorities",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU: 30,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 1,
+			evictedPods:         []string{"p1"},
+		},
+		{
+			name: "without priorities evicting best-effort pods only",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU: 30,
+			},
+			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: []*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),
+			},
+			expectedPodsEvicted: 1,
+			evictedPods:         []string{"p1"},
+		},
+		{
+			name: "with extended resource",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:   20,
+				extendedResource: 40,
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, func(node *v1.Node) {
+					test.SetNodeExtendedResource(node, extendedResource, 8)
+				}),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, func(node *v1.Node) {
+					test.SetNodeExtendedResource(node, extendedResource, 8)
+				}),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
+			},
+			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)
+				}),
+			},
+			expectedPodsEvicted: 2,
+			evictedPods:         []string{"p1", "p2"},
+		},
+		{
+			name: "with extended resource in some of nodes",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:   40,
+				extendedResource: 40,
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 10, func(node *v1.Node) {
+					test.SetNodeExtendedResource(node, extendedResource, 8)
+				}),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+				test.BuildTestNode(n3NodeName, 4000, 3000, 10, test.SetNodeUnschedulable),
+			},
+			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),
+			},
+			expectedPodsEvicted: 0,
+		},
+		{
+			name: "Other node match pod node selector",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  30,
+				v1.ResourcePods: 30,
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 9, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeSelectorKey: nodeSelectorValue,
+					}
+				}),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+			},
+			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,
+					}
+				}),
+			},
+			expectedPodsEvicted: 1,
+		},
+		{
+			name: "Other node does not match pod node selector",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:  30,
+				v1.ResourcePods: 30,
+			},
+			nodes: []*v1.Node{
+				test.BuildTestNode(n1NodeName, 4000, 3000, 9, nil),
+				test.BuildTestNode(n2NodeName, 4000, 3000, 10, nil),
+			},
+			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,
+					}
+				}),
+			},
+			expectedPodsEvicted: 0,
+		},
+	}
+
+	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 testCase.evictedPods {
+				podsForEviction[pod] = struct{}{}
+			}
+
+			evictionFailed := false
+			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()
+					if eviction, ok := obj.(*v1beta1.Eviction); ok {
+						if _, exists := podsForEviction[eviction.Name]; exists {
+							return true, obj, nil
+						}
+						evictionFailed = true
+						return true, nil, fmt.Errorf("pod %q was unexpectedly evicted", eviction.Name)
+					}
+					return true, obj, nil
+				})
+			}
+
+			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,
+				nil,
+				nil,
+				testCase.nodes,
+				false,
+				false,
+				false,
+				false,
+				false,
+			)
+
+			strategy := api.DeschedulerStrategy{
+				Enabled: true,
+				Params: &api.StrategyParameters{
+					NodeResourceUtilizationThresholds: &api.NodeResourceUtilizationThresholds{
+						Thresholds: testCase.thresholds,
+					},
+					NodeFit: true,
+				},
+			}
+			HighNodeUtilization(ctx, fakeClient, strategy, testCase.nodes, podEvictor, getPodsAssignedToNode)
+
+			podsEvicted := podEvictor.TotalEvicted()
+			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")
+			}
+		})
+	}
+}
+
+func TestValidateHighNodeUtilizationStrategyConfig(t *testing.T) {
+	tests := []struct {
+		name             string
+		thresholds       api.ResourceThresholds
+		targetThresholds api.ResourceThresholds
+		errInfo          error
+	}{
+		{
+			name: "passing target thresholds",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:    20,
+				v1.ResourceMemory: 20,
+			},
+			targetThresholds: api.ResourceThresholds{
+				v1.ResourceCPU:    80,
+				v1.ResourceMemory: 80,
+			},
+			errInfo: fmt.Errorf("targetThresholds is not applicable for HighNodeUtilization"),
+		},
+		{
+			name:       "passing empty thresholds",
+			thresholds: api.ResourceThresholds{},
+			errInfo:    fmt.Errorf("thresholds config is not valid: no resource threshold is configured"),
+		},
+		{
+			name: "passing invalid thresholds",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:    80,
+				v1.ResourceMemory: 120,
+			},
+			errInfo: fmt.Errorf("thresholds config is not valid: %v", fmt.Errorf(
+				"%v threshold not in [%v, %v] range", v1.ResourceMemory, MinResourcePercentage, MaxResourcePercentage)),
+		},
+		{
+			name: "passing valid strategy config",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:    80,
+				v1.ResourceMemory: 80,
+			},
+			errInfo: nil,
+		},
+		{
+			name: "passing valid strategy config with extended resource",
+			thresholds: api.ResourceThresholds{
+				v1.ResourceCPU:    80,
+				v1.ResourceMemory: 80,
+				extendedResource:  80,
+			},
+			errInfo: nil,
+		},
+	}
+
+	for _, testCase := range tests {
+		validateErr := validateHighUtilizationStrategyConfig(testCase.thresholds, testCase.targetThresholds)
+
+		if validateErr == nil || testCase.errInfo == nil {
+			if validateErr != testCase.errInfo {
+				t.Errorf("expected validity of strategy config: thresholds %#v targetThresholds %#v to be %v but got %v instead",
+					testCase.thresholds, testCase.targetThresholds, testCase.errInfo, validateErr)
+			}
+		} else if validateErr.Error() != testCase.errInfo.Error() {
+			t.Errorf("expected validity of strategy config: thresholds %#v targetThresholds %#v to be %v but got %v instead",
+				testCase.thresholds, testCase.targetThresholds, testCase.errInfo, validateErr)
+		}
+	}
+}
+
+func TestHighNodeUtilizationWithTaints(t *testing.T) {
+	strategy := api.DeschedulerStrategy{
+		Enabled: true,
+		Params: &api.StrategyParameters{
+			NodeResourceUtilizationThresholds: &api.NodeResourceUtilizationThresholds{
+				Thresholds: api.ResourceThresholds{
+					v1.ResourceCPU: 40,
+				},
+			},
+		},
+	}
+
+	n1 := test.BuildTestNode("n1", 1000, 3000, 10, nil)
+	n2 := test.BuildTestNode("n2", 1000, 3000, 10, nil)
+	n3 := test.BuildTestNode("n3", 1000, 3000, 10, nil)
+	n3withTaints := n3.DeepCopy()
+	n3withTaints.Spec.Taints = []v1.Taint{
+		{
+			Key:    "key",
+			Value:  "value",
+			Effect: v1.TaintEffectNoSchedule,
+		},
+	}
+
+	podThatToleratesTaint := test.BuildTestPod("tolerate_pod", 200, 0, n1.Name, test.SetRSOwnerRef)
+	podThatToleratesTaint.Spec.Tolerations = []v1.Toleration{
+		{
+			Key:   "key",
+			Value: "value",
+		},
+	}
+
+	tests := []struct {
+		name              string
+		nodes             []*v1.Node
+		pods              []*v1.Pod
+		evictionsExpected uint
+	}{
+		{
+			name:  "No taints",
+			nodes: []*v1.Node{n1, n2, n3},
+			pods: []*v1.Pod{
+				//Node 1 pods
+				test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
+				test.BuildTestPod(fmt.Sprintf("pod_2_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
+				test.BuildTestPod(fmt.Sprintf("pod_3_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
+				// Node 2 pods
+				test.BuildTestPod(fmt.Sprintf("pod_4_%s", n2.Name), 200, 0, n2.Name, test.SetRSOwnerRef),
+			},
+			evictionsExpected: 1,
+		},
+		{
+			name:  "No pod tolerates node taint",
+			nodes: []*v1.Node{n1, n3withTaints},
+			pods: []*v1.Pod{
+				//Node 1 pods
+				test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, test.SetRSOwnerRef),
+				// Node 3 pods
+				test.BuildTestPod(fmt.Sprintf("pod_2_%s", n3withTaints.Name), 200, 0, n3withTaints.Name, test.SetRSOwnerRef),
+			},
+			evictionsExpected: 0,
+		},
+		{
+			name:  "Pod which tolerates node taint",
+			nodes: []*v1.Node{n1, n3withTaints},
+			pods: []*v1.Pod{
+				//Node 1 pods
+				test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 100, 0, n1.Name, test.SetRSOwnerRef),
+				podThatToleratesTaint,
+				// Node 3 pods
+				test.BuildTestPod(fmt.Sprintf("pod_9_%s", n3withTaints.Name), 500, 0, n3withTaints.Name, test.SetRSOwnerRef),
+			},
+			evictionsExpected: 1,
+		},
+	}
+
+	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)
+			}
+
+			for _, pod := range item.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,
+				"policy/v1",
+				false,
+				&item.evictionsExpected,
+				nil,
+				item.nodes,
+				false,
+				false,
+				false,
+				false,
+				false,
+			)
+
+			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

@@ -0,0 +1,189 @@
+/*
+Copyright 2017 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 nodeutilization
+
+import (
+	"context"
+	"fmt"
+
+	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"
+	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, 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")
+		return
+	}
+	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, strategy.Params)
+	if err != nil {
+		klog.ErrorS(err, "Failed to get threshold priority from strategy's params")
+		return
+	}
+
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	thresholds := strategy.Params.NodeResourceUtilizationThresholds.Thresholds
+	targetThresholds := strategy.Params.NodeResourceUtilizationThresholds.TargetThresholds
+	if err := validateLowUtilizationStrategyConfig(thresholds, targetThresholds); 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 _, ok := thresholds[v1.ResourceCPU]; !ok {
+		thresholds[v1.ResourceCPU] = MaxResourcePercentage
+		targetThresholds[v1.ResourceCPU] = MaxResourcePercentage
+	}
+	if _, ok := thresholds[v1.ResourceMemory]; !ok {
+		thresholds[v1.ResourceMemory] = MaxResourcePercentage
+		targetThresholds[v1.ResourceMemory] = MaxResourcePercentage
+	}
+	resourceNames := getResourceNames(thresholds)
+
+	lowNodes, sourceNodes := classifyNodes(
+		getNodeUsage(nodes, thresholds, targetThresholds, resourceNames, getPodsAssignedToNode),
+		// The node has to be schedulable (to be able to move workload there)
+		func(node *v1.Node, usage NodeUsage) 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)
+		},
+		func(node *v1.Node, usage NodeUsage) bool {
+			return isNodeAboveTargetUtilization(usage)
+		},
+	)
+
+	// log message in one line
+	keysAndValues := []interface{}{
+		"CPU", thresholds[v1.ResourceCPU],
+		"Mem", thresholds[v1.ResourceMemory],
+		"Pods", thresholds[v1.ResourcePods],
+	}
+	for name := range thresholds {
+		if !isBasicResource(name) {
+			keysAndValues = append(keysAndValues, string(name), int64(thresholds[name]))
+		}
+	}
+	klog.V(1).InfoS("Criteria for a node under utilization", keysAndValues...)
+	klog.V(1).InfoS("Number of underutilized nodes", "totalNumber", len(lowNodes))
+
+	// log message in one line
+	keysAndValues = []interface{}{
+		"CPU", targetThresholds[v1.ResourceCPU],
+		"Mem", targetThresholds[v1.ResourceMemory],
+		"Pods", targetThresholds[v1.ResourcePods],
+	}
+	for name := range targetThresholds {
+		if !isBasicResource(name) {
+			keysAndValues = append(keysAndValues, string(name), int64(targetThresholds[name]))
+		}
+	}
+	klog.V(1).InfoS("Criteria for a node above target utilization", keysAndValues...)
+	klog.V(1).InfoS("Number of overutilized nodes", "totalNumber", len(sourceNodes))
+
+	if len(lowNodes) == 0 {
+		klog.V(1).InfoS("No node is underutilized, nothing to do here, you might tune your thresholds further")
+		return
+	}
+
+	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
+	}
+
+	if len(lowNodes) == len(nodes) {
+		klog.V(1).InfoS("All nodes are underutilized, nothing to do here")
+		return
+	}
+
+	if len(sourceNodes) == 0 {
+		klog.V(1).InfoS("All nodes are under target utilization, nothing to do here")
+		return
+	}
+
+	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) {
+			return false
+		}
+		for name := range totalAvailableUsage {
+			if totalAvailableUsage[name].CmpInt64(0) < 1 {
+				return false
+			}
+		}
+
+		return true
+	}
+
+	evictPodsFromSourceNodes(
+		ctx,
+		sourceNodes,
+		lowNodes,
+		podEvictor,
+		evictable.IsEvictable,
+		resourceNames,
+		"LowNodeUtilization",
+		continueEvictionCond)
+
+	klog.V(1).InfoS("Total number of pods evicted", "evictedPods", podEvictor.TotalEvicted())
+}
+
+// validateLowUtilizationStrategyConfig checks if the strategy's config is valid
+func validateLowUtilizationStrategyConfig(thresholds, targetThresholds api.ResourceThresholds) error {
+	// validate thresholds and targetThresholds config
+	if err := validateThresholds(thresholds); err != nil {
+		return fmt.Errorf("thresholds config is not valid: %v", err)
+	}
+	if err := validateThresholds(targetThresholds); err != nil {
+		return fmt.Errorf("targetThresholds config is not valid: %v", err)
+	}
+
+	// validate if thresholds and targetThresholds have same resources configured
+	if len(thresholds) != len(targetThresholds) {
+		return fmt.Errorf("thresholds and targetThresholds configured different resources")
+	}
+	for resourceName, value := range thresholds {
+		if targetValue, ok := targetThresholds[resourceName]; !ok {
+			return fmt.Errorf("thresholds and targetThresholds configured different resources")
+		} else if value > targetValue {
+			return fmt.Errorf("thresholds' %v percentage is greater than targetThresholds'", resourceName)
+		}
+	}
+	return nil
+}

pkg/descheduler/strategies/nodeutilization/nodeutilization.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2017 The Kubernetes Authors.
+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.
@@ -14,7 +14,7 @@ See the License for the specific language governing permissions and
 limitations under the License.
 */
 
-package strategies
+package nodeutilization
 
 import (
 	"context"
@@ -23,12 +23,10 @@ import (
 
 	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"
-	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
@@ -43,6 +41,8 @@ type NodeUsage struct {
 	highResourceThreshold map[v1.ResourceName]*resource.Quantity
 }
 
+type continueEvictionCond func(nodeUsage NodeUsage, totalAvailableUsage map[v1.ResourceName]*resource.Quantity) bool
+
 // NodePodsMap is a set of (node, pods) pairs
 type NodePodsMap map[*v1.Node][]*v1.Pod
 
@@ -53,7 +53,7 @@ const (
 	MaxResourcePercentage = 100
 )
 
-func validateLowNodeUtilizationParams(params *api.StrategyParameters) error {
+func validateNodeUtilizationParams(params *api.StrategyParameters) error {
 	if params == nil || params.NodeResourceUtilizationThresholds == nil {
 		return fmt.Errorf("NodeResourceUtilizationThresholds not set")
 	}
@@ -64,174 +64,70 @@ func validateLowNodeUtilizationParams(params *api.StrategyParameters) error {
 	return nil
 }
 
-// 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) {
-	// 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 := validateLowNodeUtilizationParams(strategy.Params); err != nil {
-		klog.ErrorS(err, "Invalid LowNodeUtilization parameters")
-		return
-	}
-	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, strategy.Params)
-	if err != nil {
-		klog.ErrorS(err, "Failed to get threshold priority from strategy's params")
-		return
-	}
-
-	thresholds := strategy.Params.NodeResourceUtilizationThresholds.Thresholds
-	targetThresholds := strategy.Params.NodeResourceUtilizationThresholds.TargetThresholds
-	if err := validateStrategyConfig(thresholds, targetThresholds); 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 _, ok := thresholds[v1.ResourceCPU]; !ok {
-		thresholds[v1.ResourceCPU] = MaxResourcePercentage
-		targetThresholds[v1.ResourceCPU] = MaxResourcePercentage
-	}
-	if _, ok := thresholds[v1.ResourceMemory]; !ok {
-		thresholds[v1.ResourceMemory] = MaxResourcePercentage
-		targetThresholds[v1.ResourceMemory] = MaxResourcePercentage
-	}
-
-	lowNodes, targetNodes := classifyNodes(
-		getNodeUsage(ctx, client, nodes, thresholds, targetThresholds),
-		// The node has to be schedulable (to be able to move workload there)
-		func(node *v1.Node, usage NodeUsage) 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)
-		},
-		func(node *v1.Node, usage NodeUsage) bool {
-			return isNodeAboveTargetUtilization(usage)
-		},
-	)
-
-	klog.V(1).InfoS("Criteria for a node under utilization",
-		"CPU", thresholds[v1.ResourceCPU], "Mem", thresholds[v1.ResourceMemory], "Pods", thresholds[v1.ResourcePods])
-
-	if len(lowNodes) == 0 {
-		klog.V(1).InfoS("No node is underutilized, nothing to do here, you might tune your thresholds further")
-		return
-	}
-	klog.V(1).InfoS("Total number of underutilized nodes", "totalNumber", len(lowNodes))
-
-	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)
-		return
-	}
-
-	if len(lowNodes) == len(nodes) {
-		klog.V(1).InfoS("All nodes are underutilized, nothing to do here")
-		return
-	}
-
-	if len(targetNodes) == 0 {
-		klog.V(1).InfoS("All nodes are under target utilization, nothing to do here")
-		return
-	}
-
-	klog.V(1).InfoS("Criteria for a node above target utilization",
-		"CPU", targetThresholds[v1.ResourceCPU], "Mem", targetThresholds[v1.ResourceMemory], "Pods", targetThresholds[v1.ResourcePods])
-
-	klog.V(1).InfoS("Number of nodes above target utilization", "totalNumber", len(targetNodes))
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
-
-	evictPodsFromTargetNodes(
-		ctx,
-		targetNodes,
-		lowNodes,
-		podEvictor,
-		evictable.IsEvictable)
-
-	klog.V(1).InfoS("Total number of pods evicted", "evictedPods", podEvictor.TotalEvicted())
-}
-
-// validateStrategyConfig checks if the strategy's config is valid
-func validateStrategyConfig(thresholds, targetThresholds api.ResourceThresholds) error {
-	// validate thresholds and targetThresholds config
-	if err := validateThresholds(thresholds); err != nil {
-		return fmt.Errorf("thresholds config is not valid: %v", err)
-	}
-	if err := validateThresholds(targetThresholds); err != nil {
-		return fmt.Errorf("targetThresholds config is not valid: %v", err)
-	}
-
-	// validate if thresholds and targetThresholds have same resources configured
-	if len(thresholds) != len(targetThresholds) {
-		return fmt.Errorf("thresholds and targetThresholds configured different resources")
-	}
-	for resourceName, value := range thresholds {
-		if targetValue, ok := targetThresholds[resourceName]; !ok {
-			return fmt.Errorf("thresholds and targetThresholds configured different resources")
-		} else if value > targetValue {
-			return fmt.Errorf("thresholds' %v percentage is greater than targetThresholds'", resourceName)
-		}
-	}
-	return nil
-}
-
 // validateThresholds checks if thresholds have valid resource name and resource percentage configured
 func validateThresholds(thresholds api.ResourceThresholds) error {
 	if thresholds == nil || len(thresholds) == 0 {
 		return fmt.Errorf("no resource threshold is configured")
 	}
 	for name, percent := range thresholds {
-		switch name {
-		case v1.ResourceCPU, v1.ResourceMemory, v1.ResourcePods:
-			if percent < MinResourcePercentage || percent > MaxResourcePercentage {
-				return fmt.Errorf("%v threshold not in [%v, %v] range", name, MinResourcePercentage, MaxResourcePercentage)
-			}
-		default:
-			return fmt.Errorf("only cpu, memory, or pods thresholds can be specified")
+		if percent < MinResourcePercentage || percent > MaxResourcePercentage {
+			return fmt.Errorf("%v threshold not in [%v, %v] range", name, MinResourcePercentage, MaxResourcePercentage)
 		}
 	}
 	return nil
 }
 
 func getNodeUsage(
-	ctx context.Context,
-	client clientset.Interface,
 	nodes []*v1.Node,
 	lowThreshold, highThreshold api.ResourceThresholds,
+	resourceNames []v1.ResourceName,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 ) []NodeUsage {
-	nodeUsageList := []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),
-			allPods: pods,
-			// 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.
-			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),
-			},
-			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),
-			},
+			node:                  node,
+			usage:                 nodeUtilization(node, pods, resourceNames),
+			allPods:               pods,
+			lowResourceThreshold:  lowResourceThreshold,
+			highResourceThreshold: highResourceThreshold,
 		})
 	}
 
@@ -248,7 +144,7 @@ func resourceUsagePercentages(nodeUsage NodeUsage) map[v1.ResourceName]float64 {
 	for resourceName, resourceUsage := range nodeUsage.usage {
 		cap := nodeCapacity[resourceName]
 		if !cap.IsZero() {
-			resourceUsagePercentage[resourceName] = 100 * float64(resourceUsage.Value()) / float64(cap.Value())
+			resourceUsagePercentage[resourceName] = 100 * float64(resourceUsage.MilliValue()) / float64(cap.MilliValue())
 		}
 	}
 
@@ -278,43 +174,55 @@ func classifyNodes(
 	return lowNodes, highNodes
 }
 
-// evictPodsFromTargetNodes evicts pods based on priority, if all the pods on the node have priority, if not
+// evictPodsFromSourceNodes evicts pods based on priority, if all the pods on the node have priority, if not
 // evicts them based on QoS as fallback option.
 // TODO: @ravig Break this function into smaller functions.
-func evictPodsFromTargetNodes(
+func evictPodsFromSourceNodes(
 	ctx context.Context,
-	targetNodes, lowNodes []NodeUsage,
+	sourceNodes, destinationNodes []NodeUsage,
 	podEvictor *evictions.PodEvictor,
 	podFilter func(pod *v1.Pod) bool,
+	resourceNames []v1.ResourceName,
+	strategy string,
+	continueEviction continueEvictionCond,
 ) {
 
-	sortNodesByUsage(targetNodes)
+	sortNodesByUsage(sourceNodes)
 
-	// upper bound on total number of pods/cpu/memory to be moved
+	// upper bound on total number of pods/cpu/memory and optional extended resources to be moved
 	totalAvailableUsage := map[v1.ResourceName]*resource.Quantity{
 		v1.ResourcePods:   {},
 		v1.ResourceCPU:    {},
 		v1.ResourceMemory: {},
 	}
 
-	var taintsOfLowNodes = make(map[string][]v1.Taint, len(lowNodes))
-	for _, node := range lowNodes {
-		taintsOfLowNodes[node.node.Name] = node.node.Spec.Taints
+	var taintsOfDestinationNodes = make(map[string][]v1.Taint, len(destinationNodes))
+	for _, node := range destinationNodes {
+		taintsOfDestinationNodes[node.node.Name] = node.node.Spec.Taints
 
-		for name := range totalAvailableUsage {
+		for _, name := range resourceNames {
+			if _, ok := totalAvailableUsage[name]; !ok {
+				totalAvailableUsage[name] = resource.NewQuantity(0, resource.DecimalSI)
+			}
 			totalAvailableUsage[name].Add(*node.highResourceThreshold[name])
 			totalAvailableUsage[name].Sub(*node.usage[name])
 		}
 	}
 
-	klog.V(1).InfoS(
-		"Total capacity to be moved",
+	// log message in one line
+	keysAndValues := []interface{}{
 		"CPU", totalAvailableUsage[v1.ResourceCPU].MilliValue(),
 		"Mem", totalAvailableUsage[v1.ResourceMemory].Value(),
 		"Pods", totalAvailableUsage[v1.ResourcePods].Value(),
-	)
+	}
+	for name := range totalAvailableUsage {
+		if !isBasicResource(name) {
+			keysAndValues = append(keysAndValues, string(name), totalAvailableUsage[name].Value())
+		}
+	}
+	klog.V(1).InfoS("Total capacity to be moved", keysAndValues...)
 
-	for _, node := range targetNodes {
+	for _, node := range sourceNodes {
 		klog.V(3).InfoS("Evicting pods from node", "node", klog.KObj(node.node), "usage", node.usage)
 
 		nonRemovablePods, removablePods := classifyPods(node.allPods, podFilter)
@@ -328,7 +236,7 @@ func evictPodsFromTargetNodes(
 		klog.V(1).InfoS("Evicting pods based on priority, if they have same priority, they'll be evicted based on QoS tiers")
 		// sort the evictable Pods based on priority. This also sorts them based on QoS. If there are multiple pods with same priority, they are sorted based on QoS tiers.
 		podutil.SortPodsBasedOnPriorityLowToHigh(removablePods)
-		evictPods(ctx, removablePods, node, totalAvailableUsage, taintsOfLowNodes, podEvictor)
+		evictPods(ctx, removablePods, node, totalAvailableUsage, taintsOfDestinationNodes, podEvictor, strategy, continueEviction)
 		klog.V(1).InfoS("Evicted pods from node", "node", klog.KObj(node.node), "evictedPods", podEvictor.NodeEvicted(node.node), "usage", node.usage)
 	}
 }
@@ -340,33 +248,18 @@ func evictPods(
 	totalAvailableUsage map[v1.ResourceName]*resource.Quantity,
 	taintsOfLowNodes map[string][]v1.Taint,
 	podEvictor *evictions.PodEvictor,
+	strategy string,
+	continueEviction continueEvictionCond,
 ) {
-	// stop if node utilization drops below target threshold or any of required capacity (cpu, memory, pods) is moved
-	continueCond := func() bool {
-		if !isNodeAboveTargetUtilization(nodeUsage) {
-			return false
-		}
-		if totalAvailableUsage[v1.ResourcePods].CmpInt64(0) < 1 {
-			return false
-		}
-		if totalAvailableUsage[v1.ResourceCPU].CmpInt64(0) < 1 {
-			return false
-		}
-		if totalAvailableUsage[v1.ResourceMemory].CmpInt64(0) < 1 {
-			return false
-		}
-		return true
-	}
 
-	if continueCond() {
+	if continueEviction(nodeUsage, 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, "LowNodeUtilization")
+			success, err := podEvictor.EvictPod(ctx, pod, nodeUsage.node, strategy)
 			if err != nil {
 				klog.ErrorS(err, "Error evicting pod", "pod", klog.KObj(pod))
 				break
@@ -375,20 +268,32 @@ func evictPods(
 			if success {
 				klog.V(3).InfoS("Evicted pods", "pod", klog.KObj(pod), "err", err)
 
-				cpuQuantity := utils.GetResourceRequestQuantity(pod, v1.ResourceCPU)
-				nodeUsage.usage[v1.ResourceCPU].Sub(cpuQuantity)
-				totalAvailableUsage[v1.ResourceCPU].Sub(cpuQuantity)
+				for name := range totalAvailableUsage {
+					if name == v1.ResourcePods {
+						nodeUsage.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)
+						totalAvailableUsage[name].Sub(quantity)
+					}
+				}
 
-				memoryQuantity := utils.GetResourceRequestQuantity(pod, v1.ResourceMemory)
-				nodeUsage.usage[v1.ResourceMemory].Sub(memoryQuantity)
-				totalAvailableUsage[v1.ResourceMemory].Sub(memoryQuantity)
+				keysAndValues := []interface{}{
+					"node", nodeUsage.node.Name,
+					"CPU", nodeUsage.usage[v1.ResourceCPU].MilliValue(),
+					"Mem", nodeUsage.usage[v1.ResourceMemory].Value(),
+					"Pods", nodeUsage.usage[v1.ResourcePods].Value(),
+				}
+				for name := range totalAvailableUsage {
+					if !isBasicResource(name) {
+						keysAndValues = append(keysAndValues, string(name), totalAvailableUsage[name].Value())
+					}
+				}
 
-				nodeUsage.usage[v1.ResourcePods].Sub(*resource.NewQuantity(1, resource.DecimalSI))
-				totalAvailableUsage[v1.ResourcePods].Sub(*resource.NewQuantity(1, resource.DecimalSI))
-
-				klog.V(3).InfoS("Updated node usage", "updatedUsage", nodeUsage)
-				// check if node utilization drops below target threshold or any required capacity (cpu, memory, pods) is moved
-				if !continueCond() {
+				klog.V(3).InfoS("Updated node usage", keysAndValues...)
+				// check if pods can be still evicted
+				if !continueEviction(nodeUsage, totalAvailableUsage) {
 					break
 				}
 			}
@@ -401,6 +306,15 @@ func sortNodesByUsage(nodes []NodeUsage) {
 	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) {
+				ti = ti + nodes[i].usage[name].Value()
+				tj = tj + nodes[j].usage[name].Value()
+			}
+		}
+
 		// To return sorted in descending order
 		return ti > tj
 	})
@@ -431,16 +345,42 @@ func isNodeWithLowUtilization(usage NodeUsage) bool {
 	return true
 }
 
-func nodeUtilization(node *v1.Node, pods []*v1.Pod) map[v1.ResourceName]*resource.Quantity {
+// getResourceNames returns list of resource names in resource thresholds
+func getResourceNames(thresholds api.ResourceThresholds) []v1.ResourceName {
+	resourceNames := make([]v1.ResourceName, 0, len(thresholds))
+	for name := range thresholds {
+		resourceNames = append(resourceNames, name)
+	}
+	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, quantity := range req {
-			if name == v1.ResourceCPU || name == v1.ResourceMemory {
+		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)

pkg/descheduler/strategies/nodeutilization/nodeutilization_test.go

@@ -0,0 +1,158 @@
+/*
+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 nodeutilization
+
+import (
+	"fmt"
+	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	"math"
+	"sigs.k8s.io/descheduler/pkg/api"
+	"testing"
+)
+
+var (
+	lowPriority      = int32(0)
+	highPriority     = int32(10000)
+	extendedResource = v1.ResourceName("example.com/foo")
+)
+
+func TestValidateThresholds(t *testing.T) {
+	tests := []struct {
+		name    string
+		input   api.ResourceThresholds
+		errInfo error
+	}{
+		{
+			name:    "passing nil map for threshold",
+			input:   nil,
+			errInfo: fmt.Errorf("no resource threshold is configured"),
+		},
+		{
+			name:    "passing no threshold",
+			input:   api.ResourceThresholds{},
+			errInfo: fmt.Errorf("no resource threshold is configured"),
+		},
+		{
+			name: "passing extended resource name other than cpu/memory/pods",
+			input: api.ResourceThresholds{
+				v1.ResourceCPU:   40,
+				extendedResource: 50,
+			},
+			errInfo: nil,
+		},
+		{
+			name: "passing invalid resource value",
+			input: api.ResourceThresholds{
+				v1.ResourceCPU:    110,
+				v1.ResourceMemory: 80,
+			},
+			errInfo: fmt.Errorf("%v threshold not in [%v, %v] range", v1.ResourceCPU, MinResourcePercentage, MaxResourcePercentage),
+		},
+		{
+			name: "passing a valid threshold with max and min resource value",
+			input: api.ResourceThresholds{
+				v1.ResourceCPU:    100,
+				v1.ResourceMemory: 0,
+			},
+			errInfo: nil,
+		},
+		{
+			name: "passing a valid threshold with only cpu",
+			input: api.ResourceThresholds{
+				v1.ResourceCPU: 80,
+			},
+			errInfo: nil,
+		},
+		{
+			name: "passing a valid threshold with cpu, memory and pods",
+			input: api.ResourceThresholds{
+				v1.ResourceCPU:    20,
+				v1.ResourceMemory: 30,
+				v1.ResourcePods:   40,
+			},
+			errInfo: nil,
+		},
+		{
+			name: "passing a valid threshold with only extended resource",
+			input: api.ResourceThresholds{
+				extendedResource: 80,
+			},
+			errInfo: nil,
+		},
+		{
+			name: "passing a valid threshold with cpu, memory, pods and extended resource",
+			input: api.ResourceThresholds{
+				v1.ResourceCPU:    20,
+				v1.ResourceMemory: 30,
+				v1.ResourcePods:   40,
+				extendedResource:  50,
+			},
+			errInfo: nil,
+		},
+	}
+
+	for _, test := range tests {
+		validateErr := validateThresholds(test.input)
+
+		if validateErr == nil || test.errInfo == nil {
+			if validateErr != test.errInfo {
+				t.Errorf("expected validity of threshold: %#v to be %v but got %v instead", test.input, test.errInfo, validateErr)
+			}
+		} else if validateErr.Error() != test.errInfo.Error() {
+			t.Errorf("expected validity of threshold: %#v to be %v but got %v instead", test.input, test.errInfo, validateErr)
+		}
+	}
+}
+
+func TestResourceUsagePercentages(t *testing.T) {
+	resourceUsagePercentage := resourceUsagePercentages(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),
+				},
+			},
+		},
+		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),
+		},
+	})
+
+	expectedUsageInIntPercentage := map[v1.ResourceName]float64{
+		v1.ResourceCPU:    63,
+		v1.ResourceMemory: 90,
+		v1.ResourcePods:   37,
+	}
+
+	for resourceName, percentage := range expectedUsageInIntPercentage {
+		if math.Floor(resourceUsagePercentage[resourceName]) != percentage {
+			t.Errorf("Incorrect percentange computation, expected %v, got math.Floor(%v) instead", percentage, resourceUsagePercentage[resourceName])
+		}
+	}
+
+	t.Logf("resourceUsagePercentage: %#v\n", resourceUsagePercentage)
+}

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"
@@ -48,16 +49,20 @@ 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
-	if strategy.Params != nil && strategy.Params.Namespaces != nil {
-		includedNamespaces = strategy.Params.Namespaces.Include
-		excludedNamespaces = strategy.Params.Namespaces.Exclude
+	var includedNamespaces, excludedNamespaces sets.String
+	var labelSelector *metav1.LabelSelector
+	if strategy.Params != nil {
+		if strategy.Params.Namespaces != nil {
+			includedNamespaces = sets.NewString(strategy.Params.Namespaces.Include...)
+			excludedNamespaces = sets.NewString(strategy.Params.Namespaces.Exclude...)
+		}
+		labelSelector = strategy.Params.LabelSelector
 	}
 
 	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, strategy.Params)
@@ -66,17 +71,26 @@ func RemovePodsViolatingInterPodAntiAffinity(ctx context.Context, client clients
 		return
 	}
 
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	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),
-		)
+		pods, err := podutil.ListPodsOnANode(node.Name, getPodsAssignedToNode, podFilter)
 		if err != nil {
 			return
 		}

pkg/descheduler/strategies/pod_antiaffinity_test.go

@@ -20,29 +20,49 @@ import (
 	"context"
 	"testing"
 
-	"k8s.io/api/core/v1"
+	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/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
 
 func TestPodAntiAffinity(t *testing.T) {
-	ctx := context.Background()
-	node := test.BuildTestNode("n1", 2000, 3000, 10, nil)
-	p1 := test.BuildTestPod("p1", 100, 0, node.Name, nil)
-	p2 := test.BuildTestPod("p2", 100, 0, node.Name, nil)
-	p3 := test.BuildTestPod("p3", 100, 0, node.Name, nil)
-	p4 := test.BuildTestPod("p4", 100, 0, node.Name, nil)
-	p5 := test.BuildTestPod("p5", 100, 0, node.Name, nil)
-	p6 := test.BuildTestPod("p6", 100, 0, node.Name, nil)
-	p7 := test.BuildTestPod("p7", 100, 0, node.Name, nil)
+	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{
+			"datacenter": "east",
+		}
+	})
+
+	node3 := test.BuildTestNode("n3", 2000, 3000, 10, func(node *v1.Node) {
+		node.Spec = v1.NodeSpec{
+			Unschedulable: true,
+		}
+	})
+
+	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)
+	p4 := test.BuildTestPod("p4", 100, 0, node1.Name, nil)
+	p5 := test.BuildTestPod("p5", 100, 0, node1.Name, nil)
+	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, node.Name, func(pod *v1.Pod) {
+	nonEvictablePod := test.BuildTestPod("non-evict", 100, 0, node1.Name, func(pod *v1.Pod) {
 		pod.Spec.Priority = &criticalPriority
 	})
 	p2.Labels = map[string]string{"foo": "bar"}
@@ -57,6 +77,8 @@ 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")
@@ -65,68 +87,146 @@ 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)
 	test.SetPodPriority(p6, 50)
 	test.SetPodPriority(p7, 0)
 
+	// Set pod node selectors
+	p8.Spec.NodeSelector = map[string]string{
+		"datacenter": "west",
+	}
+
+	var uint1 uint = 1
+	var uint3 uint = 3
+
 	tests := []struct {
-		description             string
-		maxPodsToEvictPerNode   int
-		pods                    []v1.Pod
-		expectedEvictedPodCount int
+		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:   &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:   &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:   &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,
+		},
 	}
 
 	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, node, nil
-		})
+		t.Run(test.description, func(t *testing.T) {
 
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			"v1",
-			false,
-			test.maxPodsToEvictPerNode,
-			[]*v1.Node{node},
-			false,
-		)
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
 
-		RemovePodsViolatingInterPodAntiAffinity(ctx, fakeClient, api.DeschedulerStrategy{}, []*v1.Node{node}, 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,
+				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

@@ -21,7 +21,8 @@ import (
 	"fmt"
 
 	v1 "k8s.io/api/core/v1"
-	v1meta "k8s.io/apimachinery/pkg/apis/meta/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.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 {
@@ -107,22 +119,20 @@ 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, 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),
-	)
+func listOldPodsOnNode(
+	nodeName string,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
+	filter podutil.FilterFunc,
+	maxPodLifeTimeSeconds uint,
+) []*v1.Pod {
+	pods, err := podutil.ListPodsOnANode(nodeName, getPodsAssignedToNode, filter)
 	if err != nil {
 		return nil
 	}
 
 	var oldPods []*v1.Pod
 	for _, pod := range pods {
-		podAgeSeconds := uint(v1meta.Now().Sub(pod.GetCreationTimestamp().Local()).Seconds())
+		podAgeSeconds := uint(metav1.Now().Sub(pod.GetCreationTimestamp().Local()).Seconds())
 		if podAgeSeconds > maxPodLifeTimeSeconds {
 			oldPods = append(oldPods, pod)
 		}

pkg/descheduler/strategies/pod_lifetime_test.go

@@ -22,26 +22,28 @@ import (
 	"time"
 
 	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 TestPodLifeTime(t *testing.T) {
-	ctx := context.Background()
-	node := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+	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())
 
 	// Setup pods, one should be evicted
-	p1 := test.BuildTestPod("p1", 100, 0, node.Name, nil)
+	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, nil)
 	p1.Namespace = "dev"
 	p1.ObjectMeta.CreationTimestamp = newerPodCreationTime
-	p2 := test.BuildTestPod("p2", 100, 0, node.Name, nil)
+	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, nil)
 	p2.Namespace = "dev"
 	p2.ObjectMeta.CreationTimestamp = olderPodCreationTime
 
@@ -50,10 +52,10 @@ func TestPodLifeTime(t *testing.T) {
 	p2.ObjectMeta.OwnerReferences = ownerRef1
 
 	// Setup pods, zero should be evicted
-	p3 := test.BuildTestPod("p3", 100, 0, node.Name, nil)
+	p3 := test.BuildTestPod("p3", 100, 0, node1.Name, nil)
 	p3.Namespace = "dev"
 	p3.ObjectMeta.CreationTimestamp = newerPodCreationTime
-	p4 := test.BuildTestPod("p4", 100, 0, node.Name, nil)
+	p4 := test.BuildTestPod("p4", 100, 0, node1.Name, nil)
 	p4.Namespace = "dev"
 	p4.ObjectMeta.CreationTimestamp = newerPodCreationTime
 
@@ -62,10 +64,10 @@ func TestPodLifeTime(t *testing.T) {
 	p4.ObjectMeta.OwnerReferences = ownerRef2
 
 	// Setup pods, one should be evicted
-	p5 := test.BuildTestPod("p5", 100, 0, node.Name, nil)
+	p5 := test.BuildTestPod("p5", 100, 0, node1.Name, nil)
 	p5.Namespace = "dev"
 	p5.ObjectMeta.CreationTimestamp = newerPodCreationTime
-	p6 := test.BuildTestPod("p6", 100, 0, node.Name, nil)
+	p6 := test.BuildTestPod("p6", 100, 0, node1.Name, nil)
 	p6.Namespace = "dev"
 	p6.ObjectMeta.CreationTimestamp = metav1.NewTime(time.Now().Add(time.Second * 605))
 
@@ -74,10 +76,10 @@ func TestPodLifeTime(t *testing.T) {
 	p6.ObjectMeta.OwnerReferences = ownerRef3
 
 	// Setup pods, zero should be evicted
-	p7 := test.BuildTestPod("p7", 100, 0, node.Name, nil)
+	p7 := test.BuildTestPod("p7", 100, 0, node1.Name, nil)
 	p7.Namespace = "dev"
 	p7.ObjectMeta.CreationTimestamp = newerPodCreationTime
-	p8 := test.BuildTestPod("p8", 100, 0, node.Name, nil)
+	p8 := test.BuildTestPod("p8", 100, 0, node1.Name, nil)
 	p8.Namespace = "dev"
 	p8.ObjectMeta.CreationTimestamp = metav1.NewTime(time.Now().Add(time.Second * 595))
 
@@ -86,10 +88,10 @@ func TestPodLifeTime(t *testing.T) {
 	p6.ObjectMeta.OwnerReferences = ownerRef4
 
 	// Setup two old pods with different status phases
-	p9 := test.BuildTestPod("p9", 100, 0, node.Name, nil)
+	p9 := test.BuildTestPod("p9", 100, 0, node1.Name, nil)
 	p9.Namespace = "dev"
 	p9.ObjectMeta.CreationTimestamp = olderPodCreationTime
-	p10 := test.BuildTestPod("p10", 100, 0, node.Name, nil)
+	p10 := test.BuildTestPod("p10", 100, 0, node1.Name, nil)
 	p10.Namespace = "dev"
 	p10.ObjectMeta.CreationTimestamp = olderPodCreationTime
 
@@ -98,13 +100,51 @@ func TestPodLifeTime(t *testing.T) {
 	p9.ObjectMeta.OwnerReferences = ownerRef1
 	p10.ObjectMeta.OwnerReferences = ownerRef1
 
+	p11 := test.BuildTestPod("p11", 100, 0, node1.Name, func(pod *v1.Pod) {
+		pod.Spec.Volumes = []v1.Volume{
+			{
+				Name: "pvc", VolumeSource: v1.VolumeSource{
+					PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{ClaimName: "foo"},
+				},
+			},
+		}
+		pod.Namespace = "dev"
+		pod.ObjectMeta.CreationTimestamp = olderPodCreationTime
+		pod.ObjectMeta.OwnerReferences = ownerRef1
+	})
+
+	// Setup two old pods with different labels
+	p12 := test.BuildTestPod("p12", 100, 0, node1.Name, nil)
+	p12.Namespace = "dev"
+	p12.ObjectMeta.CreationTimestamp = olderPodCreationTime
+	p13 := test.BuildTestPod("p13", 100, 0, node1.Name, nil)
+	p13.Namespace = "dev"
+	p13.ObjectMeta.CreationTimestamp = olderPodCreationTime
+
+	p12.ObjectMeta.Labels = map[string]string{"foo": "bar"}
+	p13.ObjectMeta.Labels = map[string]string{"foo": "bar1"}
+	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
-		expectedEvictedPodCount int
+		pods                    []*v1.Pod
+		nodes                   []*v1.Node
+		expectedEvictedPodCount uint
+		ignorePvcPods           bool
 	}{
 		{
 			description: "Two pods in the `dev` Namespace, 1 is new and 1 very is old. 1 should be evicted.",
@@ -114,8 +154,8 @@ 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,
 		},
 		{
@@ -126,8 +166,8 @@ 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,
 		},
 		{
@@ -138,8 +178,8 @@ 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,
 		},
 		{
@@ -150,8 +190,8 @@ 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,
 		},
 		{
@@ -165,34 +205,111 @@ func TestPodLifeTime(t *testing.T) {
 					},
 				},
 			},
-			maxPodsToEvictPerNode:   5,
-			pods:                    []v1.Pod{*p9, *p10},
+			pods:                    []*v1.Pod{p9, p10},
+			nodes:                   []*v1.Node{node1},
 			expectedEvictedPodCount: 1,
 		},
+		{
+			description: "does not evict pvc pods with ignorePvcPods set to true",
+			strategy: api.DeschedulerStrategy{
+				Enabled: true,
+				Params: &api.StrategyParameters{
+					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
+				},
+			},
+			pods:                    []*v1.Pod{p11},
+			nodes:                   []*v1.Node{node1},
+			expectedEvictedPodCount: 0,
+			ignorePvcPods:           true,
+		},
+		{
+			description: "evicts pvc pods with ignorePvcPods set to false (or unset)",
+			strategy: api.DeschedulerStrategy{
+				Enabled: true,
+				Params: &api.StrategyParameters{
+					PodLifeTime: &api.PodLifeTime{MaxPodLifeTimeSeconds: &maxLifeTime},
+				},
+			},
+			pods:                    []*v1.Pod{p11},
+			nodes:                   []*v1.Node{node1},
+			expectedEvictedPodCount: 1,
+		},
+		{
+			description: "No pod to evicted since all 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{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{}
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
-			return true, &v1.PodList{Items: tc.pods}, nil
-		})
-		fakeClient.Fake.AddReactor("get", "nodes", func(action core.Action) (bool, runtime.Object, error) {
-			return true, node, nil
-		})
-		podEvictor := evictions.NewPodEvictor(
-			fakeClient,
-			"v1",
-			false,
-			tc.maxPodsToEvictPerNode,
-			[]*v1.Node{node},
-			false,
-		)
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
 
-		PodLifeTime(ctx, fakeClient, tc.strategy, []*v1.Node{node}, 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)
-		}
+			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,
+				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)
+			}
+		})
 	}
-
 }

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
@@ -61,24 +62,33 @@ 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...)
 	}
 
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
+	nodeFit := false
+	if strategy.Params != nil {
+		nodeFit = strategy.Params.NodeFit
+	}
+
+	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),
-		)
+		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

@@ -18,22 +18,24 @@ package strategies
 
 import (
 	"context"
+	"fmt"
 	"testing"
 
-	"fmt"
-
-	"k8s.io/api/core/v1"
+	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)
@@ -55,7 +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.
@@ -80,8 +82,25 @@ func initPods(node *v1.Node) []v1.Pod {
 }
 
 func TestRemovePodsHavingTooManyRestarts(t *testing.T) {
-	ctx := context.Background()
-	createStrategy := func(enabled, includingInitContainers bool, restartThresholds int32) api.DeschedulerStrategy {
+	node1 := test.BuildTestNode("node1", 2000, 3000, 10, nil)
+	node2 := test.BuildTestNode("node2", 2000, 3000, 10, func(node *v1.Node) {
+		node.Spec.Taints = []v1.Taint{
+			{
+				Key:    "hardware",
+				Value:  "gpu",
+				Effect: v1.TaintEffectNoSchedule,
+			},
+		}
+	})
+	node3 := test.BuildTestNode("node3", 2000, 3000, 10, func(node *v1.Node) {
+		node.Spec = v1.NodeSpec{
+			Unschedulable: true,
+		}
+	})
+
+	pods := initPods(node1)
+
+	createStrategy := func(enabled, includingInitContainers bool, restartThresholds int32, nodeFit bool) api.DeschedulerStrategy {
 		return api.DeschedulerStrategy{
 			Enabled: enabled,
 			Params: &api.StrategyParameters{
@@ -89,96 +108,144 @@ func TestRemovePodsHavingTooManyRestarts(t *testing.T) {
 					PodRestartThreshold:     restartThresholds,
 					IncludingInitContainers: includingInitContainers,
 				},
+				NodeFit: nodeFit,
 			},
 		}
 	}
 
+	var uint3 uint = 3
+
 	tests := []struct {
-		description             string
-		pods                    []v1.Pod
-		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),
+			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),
+			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 pods evictions",
-			strategy:                createStrategy(true, true, 1*25),
+			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 pods evictions",
-			strategy:                createStrategy(true, false, 1*25),
+			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 pods evictions",
-			strategy:                createStrategy(true, true, 1*20),
+			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 pods evictions",
-			strategy:                createStrategy(true, false, 1*20),
+			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),
+			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),
+			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 pods evictions",
-			strategy:                createStrategy(true, true, 1),
+			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:   &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:   &uint3,
 		},
 	}
 
 	for _, tc := range tests {
-		node := test.BuildTestNode("node1", 2000, 3000, 10, nil)
-		pods := initPods(node)
+		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,
+				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,
-			"v1",
-			false,
-			tc.maxPodsToEvictPerNode,
-			[]*v1.Node{node},
-			false,
-		)
-
-		RemovePodsHavingTooManyRestarts(ctx, fakeClient, tc.strategy, []*v1.Node{node}, 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

@@ -25,12 +25,15 @@ import (
 	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"
+	utilerrors "k8s.io/apimachinery/pkg/util/errors"
 	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/descheduler/strategies/validation"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
@@ -45,48 +48,30 @@ type topology struct {
 	pods []*v1.Pod
 }
 
-func validateAndParseTopologySpreadParams(ctx context.Context, client clientset.Interface, params *api.StrategyParameters) (int32, sets.String, sets.String, error) {
-	var includedNamespaces, excludedNamespaces sets.String
-	if params == nil {
-		return 0, includedNamespaces, 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 0, includedNamespaces, excludedNamespaces, fmt.Errorf("only one of Include/Exclude namespaces can be set")
-	}
-	if params.ThresholdPriority != nil && params.ThresholdPriorityClassName != "" {
-		return 0, includedNamespaces, excludedNamespaces, fmt.Errorf("only one of thresholdPriority and thresholdPriorityClassName can be set")
-	}
-	thresholdPriority, err := utils.GetPriorityFromStrategyParams(ctx, client, params)
-	if err != nil {
-		return 0, includedNamespaces, excludedNamespaces, 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...)
-	}
-
-	return thresholdPriority, includedNamespaces, excludedNamespaces, nil
-}
-
 func RemovePodsViolatingTopologySpreadConstraint(
 	ctx context.Context,
 	client clientset.Interface,
 	strategy api.DeschedulerStrategy,
 	nodes []*v1.Node,
 	podEvictor *evictions.PodEvictor,
+	getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc,
 ) {
-	thresholdPriority, includedNamespaces, excludedNamespaces, err := validateAndParseTopologySpreadParams(ctx, client, strategy.Params)
+	strategyParams, err := validation.ValidateAndParseStrategyParams(ctx, client, strategy.Params)
 	if err != nil {
-		klog.ErrorS(err, "Invalid PodLifeTime parameters")
+		klog.ErrorS(err, "Invalid RemovePodsViolatingTopologySpreadConstraint parameters")
 		return
 	}
 
+	evictable := podEvictor.Evictable(
+		evictions.WithPriorityThreshold(strategyParams.ThresholdPriority),
+		evictions.WithNodeFit(strategyParams.NodeFit),
+		evictions.WithLabelSelector(strategyParams.LabelSelector),
+	)
+
 	nodeMap := make(map[string]*v1.Node, len(nodes))
 	for _, node := range nodes {
 		nodeMap[node.Name] = node
 	}
-	evictable := podEvictor.Evictable(evictions.WithPriorityThreshold(thresholdPriority))
 
 	// 1. for each namespace for which there is Topology Constraint
 	// 2. for each TopologySpreadConstraint in that namespace
@@ -110,8 +95,8 @@ func RemovePodsViolatingTopologySpreadConstraint(
 	podsForEviction := make(map[*v1.Pod]struct{})
 	// 1. for each namespace...
 	for _, namespace := range namespaces.Items {
-		if (len(includedNamespaces) > 0 && !includedNamespaces.Has(namespace.Name)) ||
-			(len(excludedNamespaces) > 0 && 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{})
@@ -121,13 +106,11 @@ func RemovePodsViolatingTopologySpreadConstraint(
 		}
 
 		// ...where there is a topology constraint
-		//namespaceTopologySpreadConstrainPods := make([]v1.Pod, 0, len(namespacePods.Items))
 		namespaceTopologySpreadConstraints := make(map[v1.TopologySpreadConstraint]struct{})
 		for _, pod := range namespacePods.Items {
 			for _, constraint := range pod.Spec.TopologySpreadConstraints {
-				// Only deal with hard topology constraints
-				// TODO(@damemi): add support for soft constraints
-				if constraint.WhenUnsatisfiable != v1.DoNotSchedule {
+				// Ignore soft topology constraints if they are not included
+				if constraint.WhenUnsatisfiable == v1.ScheduleAnyway && (strategy.Params == nil || !strategy.Params.IncludeSoftConstraints) {
 					continue
 				}
 				namespaceTopologySpreadConstraints[constraint] = struct{}{}
@@ -158,6 +141,10 @@ 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
@@ -188,7 +175,10 @@ func RemovePodsViolatingTopologySpreadConstraint(
 	}
 
 	for pod := range podsForEviction {
-		if _, err := podEvictor.EvictPod(ctx, pod, nodeMap[pod.Spec.NodeName], "PodTopologySpread"); err != nil && !evictable.IsEvictable(pod) {
+		if !evictable.IsEvictable(pod) {
+			continue
+		}
+		if _, err := podEvictor.EvictPod(ctx, pod, nodeMap[pod.Spec.NodeName], "PodTopologySpread"); err != nil {
 			klog.ErrorS(err, "Error evicting pod", "pod", klog.KObj(pod))
 			break
 		}
@@ -241,6 +231,7 @@ func balanceDomains(
 	sumPods float64,
 	isEvictable func(*v1.Pod) bool,
 	nodeMap map[string]*v1.Node) {
+
 	idealAvg := sumPods / float64(len(constraintTopologies))
 	sortedDomains := sortDomains(constraintTopologies, isEvictable)
 	// i is the index for belowOrEqualAvg
@@ -249,7 +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--
 		}
 
@@ -282,19 +273,11 @@ 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 the pod has a hard nodeAffinity or nodeSelector 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)
-			if aboveToEvict[k].Spec.NodeSelector != nil ||
-				(aboveToEvict[k].Spec.Affinity != nil &&
-					aboveToEvict[k].Spec.Affinity.NodeAffinity != nil &&
-					aboveToEvict[k].Spec.Affinity.NodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution != nil &&
-					nodesPodFitsOnBesidesCurrent(aboveToEvict[k], nodeMap) == 0) {
-				klog.V(2).InfoS("Ignoring pod for eviction due to node selector/affinity", "pod", klog.KObj(aboveToEvict[k]))
+			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]))
 				continue
 			}
+
 			podsForEviction[aboveToEvict[k]] = struct{}{}
 		}
 		sortedDomains[j].pods = sortedDomains[j].pods[:len(sortedDomains[j].pods)-movePods]
@@ -302,18 +285,54 @@ func balanceDomains(
 	}
 }
 
-// nodesPodFitsOnBesidesCurrent counts the number of nodes this pod could fit on based on its affinity
+// 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 nodesPodFitsOnBesidesCurrent(pod *v1.Pod, nodeMap map[string]*v1.Node) int {
-	count := 0
-	for _, node := range nodeMap {
-		if nodeutil.PodFitsCurrentNode(pod, node) && node != nodeMap[pod.Spec.NodeName] {
-			count++
-		}
+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
 	}
-	return count
+
+	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

pkg/descheduler/strategies/topologyspreadconstraint_test.go

@@ -3,24 +3,25 @@ package strategies
 import (
 	"context"
 	"fmt"
-	"sigs.k8s.io/descheduler/pkg/api"
 	"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/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
@@ -57,8 +58,12 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 0,
-			strategy:             api.DeschedulerStrategy{},
-			namespaces:           []string{"ns1"},
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
 		},
 		{
 			name: "2 domains, sizes [3,1], maxSkew=1, move 1 pod to achieve [2,2]",
@@ -68,17 +73,10 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			pods: createTestPods([]testPodList{
 				{
-					count:  1,
-					node:   "n1",
-					labels: map[string]string{"foo": "bar"},
-					constraints: []v1.TopologySpreadConstraint{
-						{
-							MaxSkew:           1,
-							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
-							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
-						},
-					},
+					count:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
 				},
 				{
 					count:  2,
@@ -92,11 +90,15 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 1,
-			strategy:             api.DeschedulerStrategy{},
-			namespaces:           []string{"ns1"},
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
 		},
 		{
-			name: "2 domains, sizes [3,1], maxSkew=1, move 1 pod to achieve [2,2], exclude kube-system namespace",
+			name: "2 domains, sizes [3,1], maxSkew=1, move 1 pod to achieve [2,2] (soft constraints)",
 			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" }),
@@ -110,7 +112,7 @@ func TestTopologySpreadConstraint(t *testing.T) {
 						{
 							MaxSkew:           1,
 							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
+							WhenUnsatisfiable: v1.ScheduleAnyway,
 							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
 						},
 					},
@@ -127,7 +129,70 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 1,
-			strategy:             api.DeschedulerStrategy{Enabled: true, Params: &api.StrategyParameters{Namespaces: &api.Namespaces{Exclude: []string{"kube-system"}}}},
+			strategy:             api.DeschedulerStrategy{Params: &api.StrategyParameters{IncludeSoftConstraints: true}},
+			namespaces:           []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [3,1], maxSkew=1, no pods eligible, move 0 pods",
+			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:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+					noOwners:    true,
+				},
+				{
+					count:    2,
+					node:     "n1",
+					labels:   map[string]string{"foo": "bar"},
+					noOwners: true,
+				},
+				{
+					count:    1,
+					node:     "n2",
+					labels:   map[string]string{"foo": "bar"},
+					noOwners: true,
+				},
+			}),
+			expectedEvictedCount: 0,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [3,1], maxSkew=1, move 1 pod to achieve [2,2], exclude kube-system namespace",
+			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:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:  2,
+					node:   "n1",
+					labels: map[string]string{"foo": "bar"},
+				},
+				{
+					count:  1,
+					node:   "n2",
+					labels: map[string]string{"foo": "bar"},
+				},
+			}),
+			expectedEvictedCount: 1,
+			strategy:             api.DeschedulerStrategy{Enabled: true, Params: &api.StrategyParameters{NodeFit: true, Namespaces: &api.Namespaces{Exclude: []string{"kube-system"}}}},
 			namespaces:           []string{"ns1"},
 		},
 		{
@@ -138,17 +203,10 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			pods: createTestPods([]testPodList{
 				{
-					count:  1,
-					node:   "n1",
-					labels: map[string]string{"foo": "bar"},
-					constraints: []v1.TopologySpreadConstraint{
-						{
-							MaxSkew:           1,
-							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
-							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
-						},
-					},
+					count:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
 				},
 				{
 					count:  4,
@@ -162,8 +220,12 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 1,
-			strategy:             api.DeschedulerStrategy{},
-			namespaces:           []string{"ns1"},
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
 		},
 		{
 			name: "2 domains, sizes [4,0], maxSkew=1, move 2 pods to achieve [2,2]",
@@ -173,17 +235,10 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			pods: createTestPods([]testPodList{
 				{
-					count:  1,
-					node:   "n1",
-					labels: map[string]string{"foo": "bar"},
-					constraints: []v1.TopologySpreadConstraint{
-						{
-							MaxSkew:           1,
-							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
-							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
-						},
-					},
+					count:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
 				},
 				{
 					count:  3,
@@ -192,8 +247,12 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 2,
-			strategy:             api.DeschedulerStrategy{},
-			namespaces:           []string{"ns1"},
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
 		},
 		{
 			name: "2 domains, sizes [4,0], maxSkew=1, only move 1 pod since pods with nodeSelector and nodeAffinity aren't evicted",
@@ -203,17 +262,10 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			pods: createTestPods([]testPodList{
 				{
-					count:  1,
-					node:   "n1",
-					labels: map[string]string{"foo": "bar"},
-					constraints: []v1.TopologySpreadConstraint{
-						{
-							MaxSkew:           1,
-							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
-							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
-						},
-					},
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					constraints:  getDefaultTopologyConstraints(1),
 					nodeSelector: map[string]string{"zone": "zoneA"},
 				},
 				{
@@ -239,8 +291,59 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 1,
-			strategy:             api.DeschedulerStrategy{},
-			namespaces:           []string{"ns1"},
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: true,
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [4,0], maxSkew=1, move 2 pods since selector matches multiple nodes",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels["zone"] = "zoneA"
+					n.Labels["region"] = "boston"
+				}),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels["zone"] = "zoneB"
+					n.Labels["region"] = "boston"
+				}),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					constraints:  getDefaultTopologyConstraints(1),
+					nodeSelector: map[string]string{"region": "boston"},
+				},
+				{
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					nodeSelector: map[string]string{"region": "boston"},
+				},
+				{
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					nodeSelector: map[string]string{"region": "boston"},
+				},
+				{
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					nodeSelector: map[string]string{"region": "boston"},
+				},
+			}),
+			expectedEvictedCount: 2,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
 		},
 		{
 			name: "3 domains, sizes [0, 1, 100], maxSkew=1, move 66 pods to get [34, 33, 34]",
@@ -251,17 +354,10 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			pods: createTestPods([]testPodList{
 				{
-					count:  1,
-					node:   "n2",
-					labels: map[string]string{"foo": "bar"},
-					constraints: []v1.TopologySpreadConstraint{
-						{
-							MaxSkew:           1,
-							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
-							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
-						},
-					},
+					count:       1,
+					node:        "n2",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
 				},
 				{
 					count:  100,
@@ -283,17 +379,10 @@ func TestTopologySpreadConstraint(t *testing.T) {
 			},
 			pods: createTestPods([]testPodList{
 				{
-					count:  1,
-					node:   "n2",
-					labels: map[string]string{"foo": "bar"},
-					constraints: []v1.TopologySpreadConstraint{
-						{
-							MaxSkew:           1,
-							TopologyKey:       "zone",
-							WhenUnsatisfiable: v1.DoNotSchedule,
-							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
-						},
-					},
+					count:       1,
+					node:        "n2",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
 				},
 				{
 					count:  3,
@@ -307,8 +396,12 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				},
 			}),
 			expectedEvictedCount: 3,
-			strategy:             api.DeschedulerStrategy{},
-			namespaces:           []string{"ns1"},
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
 		},
 		{
 			name: "2 domains size [2 6], maxSkew=2, should move 1 to get [3 5]",
@@ -316,6 +409,47 @@ func TestTopologySpreadConstraint(t *testing.T) {
 				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:       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: 1,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					NodeFit: false,
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
+		{
+			name: "2 domains size [2 6], maxSkew=2, can't move any because of node taints",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels["zone"] = "zoneA"
+					n.Spec.Taints = []v1.Taint{
+						{
+							Key:    "hardware",
+							Value:  "gpu",
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) { n.Labels["zone"] = "zoneB" }),
+			},
 			pods: createTestPods([]testPodList{
 				{
 					count:  1,
@@ -341,35 +475,440 @@ func TestTopologySpreadConstraint(t *testing.T) {
 					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)",
+			nodes: []*v1.Node{
+				test.BuildTestNode("n1", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels = map[string]string{"topology.kubernetes.io/zone": "zoneA", "kubernetes.io/hostname": "n1"}
+				}),
+				test.BuildTestNode("n2", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels = map[string]string{"topology.kubernetes.io/zone": "zoneA", "kubernetes.io/hostname": "n2"}
+				}),
+				test.BuildTestNode("n3", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels = map[string]string{"topology.kubernetes.io/zone": "zoneA", "kubernetes.io/hostname": "n3"}
+				}),
+				test.BuildTestNode("n4", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels = map[string]string{"topology.kubernetes.io/zone": "zoneB", "kubernetes.io/hostname": "n4"}
+				}),
+				test.BuildTestNode("n5", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels = map[string]string{"topology.kubernetes.io/zone": "zoneB", "kubernetes.io/hostname": "n5"}
+				}),
+				test.BuildTestNode("n6", 2000, 3000, 10, func(n *v1.Node) {
+					n.Labels = map[string]string{"topology.kubernetes.io/zone": "zoneB", "kubernetes.io/hostname": "n6"}
+				}),
+			},
+			pods: createTestPods([]testPodList{
+				{
+					count:  1,
+					node:   "n1",
+					labels: map[string]string{"app": "whoami"},
+					constraints: []v1.TopologySpreadConstraint{
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "topology.kubernetes.io/zone",
+							WhenUnsatisfiable: v1.ScheduleAnyway,
+						},
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "kubernetes.io/hostname",
+							WhenUnsatisfiable: v1.DoNotSchedule,
+						},
+					},
+				},
+				{
+					count:  1,
+					node:   "n2",
+					labels: map[string]string{"app": "whoami"},
+					constraints: []v1.TopologySpreadConstraint{
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "topology.kubernetes.io/zone",
+							WhenUnsatisfiable: v1.ScheduleAnyway,
+						},
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "kubernetes.io/hostname",
+							WhenUnsatisfiable: v1.DoNotSchedule,
+						},
+					},
+				},
+				{
+					count:  1,
+					node:   "n3",
+					labels: map[string]string{"app": "whoami"},
+					constraints: []v1.TopologySpreadConstraint{
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "topology.kubernetes.io/zone",
+							WhenUnsatisfiable: v1.ScheduleAnyway,
+						},
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "kubernetes.io/hostname",
+							WhenUnsatisfiable: v1.DoNotSchedule,
+						},
+					},
+				},
+				{
+					count:  1,
+					node:   "n4",
+					labels: map[string]string{"app": "whoami"},
+					constraints: []v1.TopologySpreadConstraint{
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "topology.kubernetes.io/zone",
+							WhenUnsatisfiable: v1.ScheduleAnyway,
+						},
+						{
+							LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"app": "whoami"}},
+							MaxSkew:           1,
+							TopologyKey:       "kubernetes.io/hostname",
+							WhenUnsatisfiable: v1.DoNotSchedule,
+						},
+					},
+				},
+			}),
+			expectedEvictedCount: 1,
+			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{
+				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),
+					tolerations: []v1.Toleration{
+						{
+							Key:      "taint-test",
+							Value:    "test",
+							Operator: v1.TolerationOpEqual,
+							Effect:   v1.TaintEffectNoSchedule,
+						},
+					},
+				},
+				{
+					count:        1,
+					node:         "n1",
+					labels:       map[string]string{"foo": "bar"},
+					nodeSelector: map[string]string{"zone": "zoneA"},
+				},
+			}),
 			expectedEvictedCount: 1,
 			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",
+			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{},
+			namespaces:           []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [2,0], maxSkew=1, move 1 pod for node with PreferNoSchedule Taint",
+			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.TaintEffectPreferNoSchedule,
+						},
+					}
+				}),
+			},
+			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: 1,
+			strategy:             api.DeschedulerStrategy{},
+			namespaces:           []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [2,0], maxSkew=1, move 0 pod for node with unmatched label filtering",
+			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:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:  1,
+					node:   "n1",
+					labels: map[string]string{"foo": "bar"},
+				},
+			}),
+			expectedEvictedCount: 0,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					LabelSelector: getLabelSelector("foo", []string{"baz"}, metav1.LabelSelectorOpIn),
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [2,0], maxSkew=1, move 1 pod for node with matched label filtering",
+			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:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:  1,
+					node:   "n1",
+					labels: map[string]string{"foo": "bar"},
+				},
+			}),
+			expectedEvictedCount: 1,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					LabelSelector: getLabelSelector("foo", []string{"bar"}, metav1.LabelSelectorOpIn),
+				},
+			},
+			namespaces: []string{"ns1"},
+		},
+		{
+			name: "2 domains, sizes [2,0], maxSkew=1, move 1 pod for node with matched label filtering (NotIn op)",
+			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:       1,
+					node:        "n1",
+					labels:      map[string]string{"foo": "bar"},
+					constraints: getDefaultTopologyConstraints(1),
+				},
+				{
+					count:  1,
+					node:   "n1",
+					labels: map[string]string{"foo": "bar"},
+				},
+			}),
+			expectedEvictedCount: 1,
+			strategy: api.DeschedulerStrategy{
+				Params: &api.StrategyParameters{
+					LabelSelector: getLabelSelector("foo", []string{"baz"}, metav1.LabelSelectorOpNotIn),
+				},
+			},
+			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,
 				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)
@@ -379,12 +918,15 @@ 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
+	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 {
@@ -398,13 +940,34 @@ func createTestPods(testPods []testPodList) []*v1.Pod {
 					p.Labels = make(map[string]string)
 					p.Labels = tp.labels
 					p.Namespace = "ns1"
-					p.ObjectMeta.OwnerReferences = ownerRef1
+					if !tp.noOwners {
+						p.ObjectMeta.OwnerReferences = ownerRef1
+					}
 					p.Spec.TopologySpreadConstraints = tp.constraints
 					p.Spec.NodeSelector = tp.nodeSelector
 					p.Spec.Affinity = tp.nodeAffinity
+					p.Spec.Tolerations = tp.tolerations
+					p.ObjectMeta.DeletionTimestamp = tp.deletionTimestamp
 				}))
 			podNum++
 		}
 	}
 	return pods
 }
+
+func getLabelSelector(key string, values []string, operator metav1.LabelSelectorOperator) *metav1.LabelSelector {
+	return &metav1.LabelSelector{
+		MatchExpressions: []metav1.LabelSelectorRequirement{{Key: key, Operator: operator, Values: values}},
+	}
+}
+
+func getDefaultTopologyConstraints(maxSkew int32) []v1.TopologySpreadConstraint {
+	return []v1.TopologySpreadConstraint{
+		{
+			MaxSkew:           maxSkew,
+			TopologyKey:       "zone",
+			WhenUnsatisfiable: v1.DoNotSchedule,
+			LabelSelector:     &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}},
+		},
+	}
+}

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

@@ -5,6 +5,7 @@ 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"
@@ -82,18 +83,59 @@ func GetResourceRequestQuantity(pod *v1.Pod, resourceName v1.ResourceName) resou
 	return requestQuantity
 }
 
-// IsMirrorPod returns true if the passed Pod is a Mirror Pod.
+// IsMirrorPod returns true if the pod is a Mirror Pod.
 func IsMirrorPod(pod *v1.Pod) bool {
 	_, ok := pod.Annotations[v1.MirrorPodAnnotationKey]
 	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)
 	return err == nil && source != "api"
 }
 
+// IsCriticalPriorityPod returns true if the pod has critical priority.
+func IsCriticalPriorityPod(pod *v1.Pod) bool {
+	return pod.Spec.Priority != nil && *pod.Spec.Priority >= SystemCriticalPriority
+}
+
+// IsDaemonsetPod returns true if the pod is a IsDaemonsetPod.
+func IsDaemonsetPod(ownerRefList []metav1.OwnerReference) bool {
+	for _, ownerRef := range ownerRefList {
+		if ownerRef.Kind == "DaemonSet" {
+			return true
+		}
+	}
+	return false
+}
+
+// IsPodWithLocalStorage returns true if the pod has local storage.
+func IsPodWithLocalStorage(pod *v1.Pod) bool {
+	for _, volume := range pod.Spec.Volumes {
+		if volume.HostPath != nil || volume.EmptyDir != nil {
+			return true
+		}
+	}
+
+	return false
+}
+
+// IsPodWithLocalStorage returns true if the pod has claimed a persistent volume.
+func IsPodWithPVC(pod *v1.Pod) bool {
+	for _, volume := range pod.Spec.Volumes {
+		if volume.PersistentVolumeClaim != nil {
+			return true
+		}
+	}
+	return false
+}
+
 // GetPodSource returns the source of the pod based on the annotation.
 func GetPodSource(pod *v1.Pod) (string, error) {
 	if pod.Annotations != nil {
@@ -104,23 +146,6 @@ func GetPodSource(pod *v1.Pod) (string, error) {
 	return "", fmt.Errorf("cannot get source of pod %q", pod.UID)
 }
 
-// IsCriticalPod returns true if the pod is a static or mirror pod.
-func IsCriticalPod(pod *v1.Pod) bool {
-	if IsStaticPod(pod) {
-		return true
-	}
-
-	if IsMirrorPod(pod) {
-		return true
-	}
-
-	if pod.Spec.Priority != nil && *pod.Spec.Priority >= SystemCriticalPriority {
-		return true
-	}
-
-	return false
-}
-
 // PodRequestsAndLimits returns a dictionary of all defined resources summed up for all
 // containers of the pod. If PodOverhead feature is enabled, pod overhead is added to the
 // total container resource requests and to the total container limits which have a

pkg/utils/predicates.go

@@ -18,6 +18,8 @@ package utils
 
 import (
 	"fmt"
+	"reflect"
+	"sort"
 
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/labels"
@@ -80,6 +82,116 @@ func podMatchesNodeLabels(pod *v1.Pod, node *v1.Node) bool {
 	return true
 }
 
+func uniqueSortNodeSelectorTerms(srcTerms []v1.NodeSelectorTerm) []v1.NodeSelectorTerm {
+	terms := append([]v1.NodeSelectorTerm{}, srcTerms...)
+	for i := range terms {
+		if terms[i].MatchExpressions != nil {
+			terms[i].MatchExpressions = uniqueSortNodeSelectorRequirements(terms[i].MatchExpressions)
+		}
+		if terms[i].MatchFields != nil {
+			terms[i].MatchFields = uniqueSortNodeSelectorRequirements(terms[i].MatchFields)
+		}
+	}
+
+	if len(terms) < 2 {
+		return terms
+	}
+
+	lastTerm := terms[0]
+	uniqueTerms := append([]v1.NodeSelectorTerm{}, terms[0])
+
+	for _, term := range terms[1:] {
+		if reflect.DeepEqual(term, lastTerm) {
+			continue
+		}
+		lastTerm = term
+		uniqueTerms = append(uniqueTerms, term)
+	}
+
+	return uniqueTerms
+}
+
+// sort NodeSelectorRequirement in (key, operator, values) order
+func uniqueSortNodeSelectorRequirements(srcReqs []v1.NodeSelectorRequirement) []v1.NodeSelectorRequirement {
+	reqs := append([]v1.NodeSelectorRequirement{}, srcReqs...)
+
+	// unique sort Values
+	for i := range reqs {
+		sort.Strings(reqs[i].Values)
+		if len(reqs[i].Values) > 1 {
+			lastString := reqs[i].Values[0]
+			values := []string{lastString}
+			for _, val := range reqs[i].Values {
+				if val == lastString {
+					continue
+				}
+				lastString = val
+				values = append(values, val)
+			}
+			reqs[i].Values = values
+		}
+	}
+
+	if len(reqs) < 2 {
+		return reqs
+	}
+
+	// unique sort reqs
+	sort.Slice(reqs, func(i, j int) bool {
+		if reqs[i].Key < reqs[j].Key {
+			return true
+		}
+		if reqs[i].Key > reqs[j].Key {
+			return false
+		}
+		if reqs[i].Operator < reqs[j].Operator {
+			return true
+		}
+		if reqs[i].Operator > reqs[j].Operator {
+			return false
+		}
+		if len(reqs[i].Values) < len(reqs[j].Values) {
+			return true
+		}
+		if len(reqs[i].Values) > len(reqs[j].Values) {
+			return false
+		}
+		for k := range reqs[i].Values {
+			if reqs[i].Values[k] < reqs[j].Values[k] {
+				return true
+			}
+			if reqs[i].Values[k] > reqs[j].Values[k] {
+				return false
+			}
+		}
+		return true
+	})
+
+	lastReq := reqs[0]
+	uniqueReqs := append([]v1.NodeSelectorRequirement{}, lastReq)
+	for _, req := range reqs[1:] {
+		if reflect.DeepEqual(req, lastReq) {
+			continue
+		}
+		lastReq = req
+		uniqueReqs = append(uniqueReqs, req)
+	}
+	return uniqueReqs
+}
+
+func NodeSelectorsEqual(n1, n2 *v1.NodeSelector) bool {
+	if n1 == nil && n2 == nil {
+		return true
+	}
+	if n1 == nil || n2 == nil {
+		return false
+	}
+	return reflect.DeepEqual(
+		uniqueSortNodeSelectorTerms(n1.NodeSelectorTerms),
+		uniqueSortNodeSelectorTerms(n2.NodeSelectorTerms),
+	)
+}
+
 // TolerationsTolerateTaint checks if taint is tolerated by any of the tolerations.
 func TolerationsTolerateTaint(tolerations []v1.Toleration, taint *v1.Taint) bool {
 	for i := range tolerations {
@@ -95,10 +207,6 @@ type taintsFilterFunc func(*v1.Taint) bool
 // TolerationsTolerateTaintsWithFilter checks if given tolerations tolerates
 // all the taints that apply to the filter in given taint list.
 func TolerationsTolerateTaintsWithFilter(tolerations []v1.Toleration, taints []v1.Taint, applyFilter taintsFilterFunc) bool {
-	if len(taints) == 0 {
-		return true
-	}
-
 	for i := range taints {
 		if applyFilter != nil && !applyFilter(&taints[i]) {
 			continue
@@ -111,3 +219,59 @@ func TolerationsTolerateTaintsWithFilter(tolerations []v1.Toleration, taints []v
 
 	return true
 }
+
+// sort by (key, value, effect, operand)
+func uniqueSortTolerations(srcTolerations []v1.Toleration) []v1.Toleration {
+	tolerations := append([]v1.Toleration{}, srcTolerations...)
+
+	if len(tolerations) < 2 {
+		return tolerations
+	}
+
+	sort.Slice(tolerations, func(i, j int) bool {
+		if tolerations[i].Key < tolerations[j].Key {
+			return true
+		}
+		if tolerations[i].Key > tolerations[j].Key {
+			return false
+		}
+		if tolerations[i].Value < tolerations[j].Value {
+			return true
+		}
+		if tolerations[i].Value > tolerations[j].Value {
+			return false
+		}
+		if tolerations[i].Effect < tolerations[j].Effect {
+			return true
+		}
+		if tolerations[i].Effect > tolerations[j].Effect {
+			return false
+		}
+		return tolerations[i].Operator < tolerations[j].Operator
+	})
+	uniqueTolerations := []v1.Toleration{tolerations[0]}
+	idx := 0
+	for _, t := range tolerations[1:] {
+		if t.MatchToleration(&uniqueTolerations[idx]) {
+			continue
+		}
+		idx++
+		uniqueTolerations = append(uniqueTolerations, t)
+	}
+	return uniqueTolerations
+}
+
+func TolerationsEqual(t1, t2 []v1.Toleration) bool {
+	t1Sorted := uniqueSortTolerations(t1)
+	t2Sorted := uniqueSortTolerations(t2)
+	l1Len := len(t1Sorted)
+	if l1Len != len(t2Sorted) {
+		return false
+	}
+	for i := 0; i < l1Len; i++ {
+		if !t1Sorted[i].MatchToleration(&t2Sorted[i]) {
+			return false
+		}
+	}
+	return true
+}

pkg/utils/predicates_test.go

@@ -0,0 +1,940 @@
+package utils
+
+import (
+	"reflect"
+	"testing"
+
+	v1 "k8s.io/api/core/v1"
+)
+
+func TestUniqueSortTolerations(t *testing.T) {
+	tests := []struct {
+		name                string
+		tolerations         []v1.Toleration
+		expectedTolerations []v1.Toleration
+	}{
+		{
+			name: "sort by key",
+			tolerations: []v1.Toleration{
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key3",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value3",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			expectedTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key3",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value3",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+		},
+		{
+			name: "sort by value",
+			tolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value3",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			expectedTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value3",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+		},
+		{
+			name: "sort by effect",
+			tolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoExecute,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectPreferNoSchedule,
+				},
+			},
+			expectedTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoExecute,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectPreferNoSchedule,
+				},
+			},
+		},
+		{
+			name: "sort unique",
+			tolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			expectedTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			resultTolerations := uniqueSortTolerations(test.tolerations)
+			if !reflect.DeepEqual(resultTolerations, test.expectedTolerations) {
+				t.Errorf("tolerations not sorted as expected, \n\tgot: %#v, \n\texpected: %#v", resultTolerations, test.expectedTolerations)
+			}
+		})
+	}
+}
+
+func TestTolerationsEqual(t *testing.T) {
+	tests := []struct {
+		name                              string
+		leftTolerations, rightTolerations []v1.Toleration
+		equal                             bool
+	}{
+		{
+			name: "identical lists",
+			leftTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			rightTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			equal: true,
+		},
+		{
+			name: "equal lists",
+			leftTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			rightTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			equal: true,
+		},
+		{
+			name: "non-equal lists",
+			leftTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			rightTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoExecute,
+				},
+			},
+			equal: false,
+		},
+		{
+			name: "different sizes lists",
+			leftTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+			},
+			rightTolerations: []v1.Toleration{
+				{
+					Key:      "key1",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value1",
+					Effect:   v1.TaintEffectNoSchedule,
+				},
+				{
+					Key:      "key2",
+					Operator: v1.TolerationOpEqual,
+					Value:    "value2",
+					Effect:   v1.TaintEffectNoExecute,
+				},
+			},
+			equal: false,
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			equal := TolerationsEqual(test.leftTolerations, test.rightTolerations)
+			if equal != test.equal {
+				t.Errorf("TolerationsEqual expected to be %v, got %v", test.equal, equal)
+			}
+		})
+	}
+}
+
+func TestUniqueSortNodeSelectorRequirements(t *testing.T) {
+	tests := []struct {
+		name                 string
+		requirements         []v1.NodeSelectorRequirement
+		expectedRequirements []v1.NodeSelectorRequirement
+	}{
+		{
+			name: "Identical requirements",
+			requirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1"},
+				},
+			},
+			expectedRequirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1"},
+				},
+			},
+		},
+		{
+			name: "Sorted requirements",
+			requirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1"},
+				},
+				{
+					Key:      "k2",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v2"},
+				},
+			},
+			expectedRequirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1"},
+				},
+				{
+					Key:      "k2",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v2"},
+				},
+			},
+		},
+		{
+			name: "Sort values",
+			requirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v2", "v1"},
+				},
+			},
+			expectedRequirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+			},
+		},
+		{
+			name: "Sort by key",
+			requirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k3",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k2",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+			},
+			expectedRequirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k2",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k3",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+			},
+		},
+		{
+			name: "Sort by operator",
+			requirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpExists,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpGt,
+					Values:   []string{"v1", "v2"},
+				},
+			},
+			expectedRequirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpExists,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpGt,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+			},
+		},
+		{
+			name: "Sort by values",
+			requirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v6", "v5"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v2", "v1"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v4", "v1"},
+				},
+			},
+			expectedRequirements: []v1.NodeSelectorRequirement{
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v2"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v1", "v4"},
+				},
+				{
+					Key:      "k1",
+					Operator: v1.NodeSelectorOpIn,
+					Values:   []string{"v5", "v6"},
+				},
+			},
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			resultRequirements := uniqueSortNodeSelectorRequirements(test.requirements)
+			if !reflect.DeepEqual(resultRequirements, test.expectedRequirements) {
+				t.Errorf("Requirements not sorted as expected, \n\tgot: %#v, \n\texpected: %#v", resultRequirements, test.expectedRequirements)
+			}
+		})
+	}
+}
+
+func TestUniqueSortNodeSelectorTerms(t *testing.T) {
+	tests := []struct {
+		name          string
+		terms         []v1.NodeSelectorTerm
+		expectedTerms []v1.NodeSelectorTerm
+	}{
+		{
+			name: "Identical terms",
+			terms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+					MatchFields: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+				},
+			},
+			expectedTerms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+					MatchFields: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "Sorted terms",
+			terms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+					MatchFields: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+				},
+			},
+			expectedTerms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+					MatchFields: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "Sort terms",
+			terms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v2", "v1"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v3", "v1"},
+						},
+					},
+				},
+			},
+			expectedTerms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1", "v2"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1", "v3"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "Unique sort terms",
+			terms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v2", "v1"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v2", "v1"},
+						},
+					},
+				},
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v2", "v1"},
+						},
+					},
+				},
+			},
+			expectedTerms: []v1.NodeSelectorTerm{
+				{
+					MatchExpressions: []v1.NodeSelectorRequirement{
+						{
+							Key:      "k1",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1", "v2"},
+						},
+						{
+							Key:      "k2",
+							Operator: v1.NodeSelectorOpIn,
+							Values:   []string{"v1"},
+						},
+					},
+				},
+			},
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			resultTerms := uniqueSortNodeSelectorTerms(test.terms)
+			if !reflect.DeepEqual(resultTerms, test.expectedTerms) {
+				t.Errorf("Terms not sorted as expected, \n\tgot: %#v, \n\texpected: %#v", resultTerms, test.expectedTerms)
+			}
+		})
+	}
+}
+
+func TestNodeSelectorTermsEqual(t *testing.T) {
+	tests := []struct {
+		name                        string
+		leftSelector, rightSelector v1.NodeSelector
+		equal                       bool
+	}{
+		{
+			name: "identical selectors",
+			leftSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1"},
+							},
+						},
+					},
+				},
+			},
+			rightSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1"},
+							},
+						},
+					},
+				},
+			},
+			equal: true,
+		},
+		{
+			name: "equal selectors",
+			leftSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v1"},
+							},
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+						},
+					},
+				},
+			},
+			rightSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1"},
+							},
+						},
+					},
+				},
+			},
+			equal: true,
+		},
+		{
+			name: "non-equal selectors in values",
+			leftSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1"},
+							},
+						},
+					},
+				},
+			},
+			rightSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+						},
+					},
+				},
+			},
+			equal: false,
+		},
+		{
+			name: "non-equal selectors in keys",
+			leftSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k3",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1"},
+							},
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1"},
+							},
+						},
+					},
+				},
+			},
+			rightSelector: v1.NodeSelector{
+				NodeSelectorTerms: []v1.NodeSelectorTerm{
+					{
+						MatchExpressions: []v1.NodeSelectorRequirement{
+							{
+								Key:      "k1",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+							{
+								Key:      "k2",
+								Operator: v1.NodeSelectorOpIn,
+								Values:   []string{"v1", "v2"},
+							},
+						},
+					},
+				},
+			},
+			equal: false,
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			equal := NodeSelectorsEqual(&test.leftSelector, &test.rightSelector)
+			if equal != test.equal {
+				t.Errorf("NodeSelectorsEqual expected to be %v, got %v", test.equal, equal)
+			}
+		})
+	}
+}

pkg/version/version.go

@@ -0,0 +1,90 @@
+/*
+Copyright 2017 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 version
+
+import (
+	"fmt"
+	"regexp"
+	"runtime"
+	"strings"
+)
+
+var (
+	// version is a constant representing the version tag that
+	// generated this build. It should be set during build via -ldflags.
+	version string
+	// buildDate in ISO8601 format, output of $(date -u +'%Y-%m-%dT%H:%M:%SZ')
+	// It should be set during build via -ldflags.
+	buildDate string
+	// gitbranch is a constant representing git branch for this build.
+	// It should be set during build via -ldflags.
+	gitbranch string
+	// gitbranch is a constant representing git sha1 for this build.
+	// It should be set during build via -ldflags.
+	gitsha1 string
+)
+
+// Info holds the information related to descheduler app version.
+type Info struct {
+	Major      string `json:"major"`
+	Minor      string `json:"minor"`
+	GitVersion string `json:"gitVersion"`
+	GitBranch  string `json:"gitBranch"`
+	GitSha1    string `json:"gitSha1"`
+	BuildDate  string `json:"buildDate"`
+	GoVersion  string `json:"goVersion"`
+	Compiler   string `json:"compiler"`
+	Platform   string `json:"platform"`
+}
+
+// Get returns the overall codebase version. It's for detecting
+// what code a binary was built from.
+func Get() Info {
+	majorVersion, minorVersion := splitVersion(version)
+	return Info{
+		Major:      majorVersion,
+		Minor:      minorVersion,
+		GitVersion: version,
+		GitBranch:  gitbranch,
+		GitSha1:    gitsha1,
+		BuildDate:  buildDate,
+		GoVersion:  runtime.Version(),
+		Compiler:   runtime.Compiler,
+		Platform:   fmt.Sprintf("%s/%s", runtime.GOOS, runtime.GOARCH),
+	}
+}
+
+// splitVersion splits the git version to generate major and minor versions needed.
+func splitVersion(version string) (string, string) {
+	if version == "" {
+		return "", ""
+	}
+
+	// Version from an automated container build environment for a tag. For example v20200521-v0.18.0.
+	m1, _ := regexp.MatchString(`^v\d{8}-v\d+\.\d+\.\d+$`, version)
+
+	// Version from an automated container build environment(not a tag) or a local build. For example v20201009-v0.18.0-46-g939c1c0.
+	m2, _ := regexp.MatchString(`^v\d{8}-v\d+\.\d+\.\d+-\w+-\w+$`, version)
+
+	if m1 || m2 {
+		semVer := strings.Split(version, "-")[1]
+		return strings.Trim(strings.Split(semVer, ".")[0], "v"), strings.Split(semVer, ".")[1] + "+"
+	}
+
+	// Something went wrong
+	return "", ""
+}

test/e2e/e2e_duplicatepods_test.go

@@ -0,0 +1,203 @@
+/*
+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"
+	"time"
+
+	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"
+	"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/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,
+				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)
+			}
+		})
+	}
+}
+
+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)
+	}
+}