Merge pull request #1662 from ingvagabund/backport-1661-1.32

TestLowNodeUtilizationKubernetesMetrics: set the right -cpu value for the cpu stressor

.gitattributes

@@ -0,0 +1,6 @@
+# Always check-out / check-in files with LF line endings.
+* text=auto eol=lf
+
+go.sum merge=union
+**/zz_generated.*.go linguist-generated=true
+vendor/** linguist-vendored

.github/workflows/helm.yaml

@@ -20,34 +20,42 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           fetch-depth: 0
 
       - name: Set up Helm
-        uses: azure/setup-helm@v2.1
+        uses: azure/setup-helm@v4.2.0
         with:
-          version: v3.9.2
+          version: v3.15.1
 
-      - uses: actions/setup-python@v3.1.2
+      - uses: actions/setup-python@v5.1.1
         with:
-          python-version: 3.7
+          python-version: 3.12
 
-      - uses: actions/setup-go@v3
+      - uses: actions/setup-go@v5
         with:
-          go-version: '1.19.3'
+          go-version-file: 'go.mod'
 
       - name: Set up chart-testing
-        uses: helm/chart-testing-action@v2.2.1
+        uses: helm/chart-testing-action@v2.6.1
         with:
-          version: v3.7.0
+          version: v3.11.0
+
+      - name: Install Helm Unit Test Plugin
+        run: |
+          helm plugin install https://github.com/helm-unittest/helm-unittest
+
+      - name: Run Helm Unit Tests
+        run: |
+          helm unittest charts/descheduler --strict -d
 
       - name: Run chart-testing (list-changed)
         id: list-changed
         run: |
           changed=$(ct list-changed --config=.github/ci/ct.yaml)
           if [[ -n "$changed" ]]; then
-            echo "::set-output name=changed::true"
+            echo "changed=true" >> $GITHUB_OUTPUT
           fi
 
       - name: Run chart-testing (lint)
@@ -60,10 +68,3 @@ jobs:
       # helm-extra-set-args only available after ct 3.6.0
       - name: Run chart-testing (install)
         run: ct install --config=.github/ci/ct.yaml --helm-extra-set-args='--set=kind=Deployment'
-
-      - name: E2E after chart install
-        env:
-          KUBERNETES_VERSION: "v1.26.0"
-          KIND_E2E: true
-          SKIP_INSTALL: true
-        run: make test-e2e

.github/workflows/manifests.yaml

@@ -0,0 +1,42 @@
+name: manifests
+
+on:
+  pull_request:
+
+jobs:
+  deploy:
+    strategy:
+      matrix:
+        k8s-version: ["v1.32.0"]
+        descheduler-version: ["v0.32.2"]
+        descheduler-api: ["v1alpha2"]
+        manifest: ["deployment"]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout Repo
+        uses: actions/checkout@v4
+      - name: Create kind cluster
+        uses: helm/kind-action@v1.12.0
+        with:
+          node_image: kindest/node:${{ matrix.k8s-version }}
+          kubectl_version: ${{ matrix.k8s-version }}
+          config: test/kind-config.yaml
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - name: Build image
+        run: |
+          VERSION="dev" make dev-image
+          docker tag descheduler:dev registry.k8s.io/descheduler/descheduler:${{ matrix.descheduler-version }}
+      - name: Kind load image
+        run: |
+          kind load docker-image registry.k8s.io/descheduler/descheduler:${{ matrix.descheduler-version }} --name chart-testing
+      - name: Create k8s manifests
+        run: |
+          kubectl create -f kubernetes/base/rbac.yaml
+          kubectl create -f test/manifests/${{ matrix.descheduler-api }}/configmap.yaml
+          kubectl create -f kubernetes/${{ matrix.manifest }}/${{ matrix.manifest }}.yaml
+      - name: Wait for ready condition
+        run: |
+          kubectl wait --for=condition=Available --timeout=60s ${{ matrix.manifest }} descheduler -n kube-system

.github/workflows/release.yaml

@@ -5,6 +5,9 @@ on:
     branches:
       - release-*
 
+permissions:
+  contents: write # allow actions to update gh-pages branch
+
 jobs:
   release:
     runs-on: ubuntu-latest
@@ -20,12 +23,12 @@ jobs:
           git config user.email "$GITHUB_ACTOR@users.noreply.github.com"
 
       - name: Install Helm
-        uses: azure/setup-helm@v1
+        uses: azure/setup-helm@v4.2.0
         with:
-          version: v3.4.0
+          version: v3.15.1
 
       - name: Run chart-releaser
-        uses: helm/chart-releaser-action@v1.1.0
+        uses: helm/chart-releaser-action@v1.6.0
         env:
           CR_TOKEN: "${{ secrets.GITHUB_TOKEN }}"
           CR_RELEASE_NAME_TEMPLATE: "descheduler-helm-chart-{{ .Version }}"

.github/workflows/security.yaml

@@ -22,7 +22,7 @@ jobs:
       fail-fast: false
     steps:
       - name: Checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           fetch-depth: 0
 

.golangci.yml

@@ -1,5 +1,5 @@
 run:
-  deadline: 2m
+  timeout: 5m
 
 linters:
   disable-all: true
@@ -15,4 +15,4 @@ linters-settings:
   gofumpt:
     extra-rules: true
   goimports:
-local-prefixes: sigs.k8s.io/descheduler
+    local-prefixes: sigs.k8s.io/descheduler

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:1.19.3
+FROM golang:1.23.3
 
 WORKDIR /go/src/sigs.k8s.io/descheduler
 COPY . .
@@ -23,6 +23,8 @@ FROM scratch
 
 MAINTAINER Kubernetes SIG Scheduling <kubernetes-sig-scheduling@googlegroups.com>
 
+LABEL org.opencontainers.image.source https://github.com/kubernetes-sigs/descheduler
+
 USER 1000
 
 COPY --from=0 /go/src/sigs.k8s.io/descheduler/_output/bin/descheduler /bin/descheduler

Makefile

@@ -26,12 +26,14 @@ ARCHS = amd64 arm arm64
 
 LDFLAGS=-ldflags "-X ${LDFLAG_LOCATION}.version=${VERSION} -X ${LDFLAG_LOCATION}.buildDate=${BUILD} -X ${LDFLAG_LOCATION}.gitbranch=${BRANCH} -X ${LDFLAG_LOCATION}.gitsha1=${SHA1}"
 
-GOLANGCI_VERSION := v1.49.0
+GOLANGCI_VERSION := v1.62.2
 HAS_GOLANGCI := $(shell ls _output/bin/golangci-lint 2> /dev/null)
 
-GOFUMPT_VERSION := v0.4.0
+GOFUMPT_VERSION := v0.7.0
 HAS_GOFUMPT := $(shell command -v gofumpt 2> /dev/null)
 
+GO_VERSION := $(shell (command -v jq > /dev/null && (go mod edit -json | jq -r .Go)) || (sed -En 's/^go (.*)$$/\1/p' go.mod))
+
 # REGISTRY is the container registry to push
 # into. The default is to push to the staging
 # registry, not production.
@@ -43,6 +45,10 @@ IMAGE:=descheduler:$(VERSION)
 # IMAGE_GCLOUD is the image name of descheduler in the remote registry
 IMAGE_GCLOUD:=$(REGISTRY)/descheduler:$(VERSION)
 
+# CURRENT_DIR is the current dir where the Makefile exists
+CURRENT_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
+
+
 # TODO: upload binaries to GCS bucket
 #
 # In the future binaries can be uploaded to
@@ -100,7 +106,7 @@ clean:
 	rm -rf _output
 	rm -rf _tmp
 
-verify: verify-govet verify-spelling verify-gofmt verify-vendor lint lint-chart verify-toc verify-gen
+verify: verify-govet verify-spelling verify-gofmt verify-vendor lint lint-chart verify-gen
 
 verify-govet:
 	./hack/verify-govet.sh
@@ -114,8 +120,8 @@ verify-gofmt:
 verify-vendor:
 	./hack/verify-vendor.sh
 
-verify-toc:
+verify-docs:
-	./hack/verify-toc.sh
+	./hack/verify-docs.sh
 
 test-unit:
 	./test/run-unit-tests.sh
@@ -127,18 +133,22 @@ gen:
 	./hack/update-generated-conversions.sh
 	./hack/update-generated-deep-copies.sh
 	./hack/update-generated-defaulters.sh
-	./hack/update-toc.sh
+	./hack/update-docs.sh
+
+gen-docker:
+	$(CONTAINER_ENGINE) run --entrypoint make -it -v $(CURRENT_DIR):/go/src/sigs.k8s.io/descheduler -w /go/src/sigs.k8s.io/descheduler golang:$(GO_VERSION) gen
 
 verify-gen:
 	./hack/verify-conversions.sh
 	./hack/verify-deep-copies.sh
 	./hack/verify-defaulters.sh
+	./hack/verify-docs.sh
 
 lint:
 ifndef HAS_GOLANGCI
 	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
+	./_output/bin/golangci-lint run -v
 
 fmt:
 ifndef HAS_GOFUMPT

OWNERS

@@ -3,6 +3,7 @@ approvers:
 - ingvagabund
 - seanmalloy
 - a7i
+- knelasevero
 reviewers:
 - damemi
 - seanmalloy
@@ -11,6 +12,7 @@ reviewers:
 - a7i
 - janeliul
 - knelasevero
+- jklaw90
 emeritus_approvers:
 - aveshagarwal
 - k82cn

charts/descheduler/Chart.yaml

@@ -1,7 +1,7 @@
 apiVersion: v1
 name: descheduler
-version: 0.26.1
+version: 0.32.2
-appVersion: 0.26.1
+appVersion: 0.32.2
 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

@@ -52,9 +52,11 @@ The following table lists the configurable parameters of the _descheduler_ chart
 | `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                                                              | `""`                                      |
+| `namespaceOverride`                 | Override the deployment namespace; defaults to .Release.Namespace                                                     | `""`                                      |
 | `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`                                     |
+| `timeZone`                          | configure `timeZone` for CronJob                                                                                      | `nil`                                     |
 | `successfulJobsHistoryLimit`        | If set, configure `successfulJobsHistoryLimit` for the _descheduler_ job                                              | `3`                                       |
 | `failedJobsHistoryLimit`            | If set, configure `failedJobsHistoryLimit` for the _descheduler_ job                                                  | `1`                                       |
 | `ttlSecondsAfterFinished`           | If set, configure `ttlSecondsAfterFinished` for the _descheduler_ job                                                 | `nil`                                     |
@@ -62,7 +64,6 @@ The following table lists the configurable parameters of the _descheduler_ chart
 | `replicas`                          | The replica count for Deployment                                                                                      | `1`                                       |
 | `leaderElection`                    | The options for high availability when running replicated components                                                  | _see values.yaml_                         |
 | `cmdOptions`                        | The options to pass to the _descheduler_ command                                                                      | _see values.yaml_                         |
-| `deschedulerPolicy.strategies`      | The _descheduler_ strategies to apply                                                                                 | _see values.yaml_                         |
 | `priorityClassName`                 | The name of the priority class to add to pods                                                                         | `system-cluster-critical`                 |
 | `rbac.create`                       | If `true`, create & use RBAC resources                                                                                | `true`                                    |
 | `resources`                         | Descheduler container CPU and memory requests/limits                                                                  | _see values.yaml_                         |
@@ -75,6 +76,7 @@ The following table lists the configurable parameters of the _descheduler_ chart
 | `service.enabled`                   | If `true`, create a service for deployment                                                                            | `false`                                   |
 | `serviceMonitor.enabled`            | If `true`, create a ServiceMonitor for deployment                                                                     | `false`                                   |
 | `serviceMonitor.namespace`          | The namespace where Prometheus expects to find service monitors                                                       | `nil`                                     |
+| `serviceMonitor.additionalLabels`   | Add custom labels to the ServiceMonitor resource                                                                      | `{}`                                      |
 | `serviceMonitor.interval`           | The scrape interval. If not set, the Prometheus default scrape interval is used                                       | `nil`                                     |
 | `serviceMonitor.honorLabels`        | Keeps the scraped data's labels when labels are on collisions with target labels.                                     | `true`                                    |
 | `serviceMonitor.insecureSkipVerify` | Skip TLS certificate validation when scraping                                                                         | `true`                                    |
@@ -82,6 +84,7 @@ The following table lists the configurable parameters of the _descheduler_ chart
 | `serviceMonitor.metricRelabelings`  | MetricRelabelConfigs to apply to samples after scraping, but before ingestion                                         | `[]`                                      |
 | `serviceMonitor.relabelings`        | RelabelConfigs to apply to samples before scraping                                                                    | `[]`                                      |
 | `affinity`                          | Node affinity to run the descheduler cronjob/deployment on specific nodes                                             | `nil`                                     |
+| `topologySpreadConstraints`         | Topology Spread Constraints to spread the descheduler cronjob/deployment across the cluster                           | `[]`                                     |
 | `tolerations`                       | tolerations to run the descheduler cronjob/deployment on specific nodes                                               | `nil`                                     |
 | `suspend`                           | Set spec.suspend in descheduler cronjob                                                                               | `false`                                   |
 | `commonLabels`                      | Labels to apply to all resources                                                                                      | `{}`                                      |

charts/descheduler/templates/NOTES.txt

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

charts/descheduler/templates/_helpers.tpl

@@ -24,6 +24,14 @@ If release name contains chart name it will be used as a full name.
 {{- end -}}
 {{- end -}}
 
+{{/*
+Expand the namespace of the release.
+Allows overriding it for multi-namespace deployments in combined charts.
+*/}}
+{{- define "descheduler.namespace" -}}
+{{- default .Release.Namespace .Values.namespaceOverride | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
 {{/*
 Create chart name and version as used by the chart label.
 */}}
@@ -87,8 +95,10 @@ Leader Election
 {{- if .Values.leaderElection.resourceName }}
 - --leader-elect-resource-name={{ .Values.leaderElection.resourceName }}
 {{- end }}
-{{- if .Values.leaderElection.resourceNamescape }}
+{{/* resource namespace value starts with a typo so keeping resourceNamescape for backwards compatibility */}}
-- --leader-elect-resource-namespace={{ .Values.leaderElection.resourceNamescape }}
+{{- $resourceNamespace := default .Values.leaderElection.resourceNamespace .Values.leaderElection.resourceNamescape -}}
+{{- if $resourceNamespace -}}
+- --leader-elect-resource-namespace={{ $resourceNamespace }}
 {{- end -}}
 {{- end }}
 {{- end }}

charts/descheduler/templates/clusterrole.yaml

@@ -24,6 +24,9 @@ rules:
 - apiGroups: ["scheduling.k8s.io"]
   resources: ["priorityclasses"]
   verbs: ["get", "watch", "list"]
+- apiGroups: ["policy"]
+  resources: ["poddisruptionbudgets"]
+  verbs: ["get", "watch", "list"]
 {{- if .Values.leaderElection.enabled }}
 - apiGroups: ["coordination.k8s.io"]
   resources: ["leases"]
@@ -33,4 +36,9 @@ rules:
   resourceNames: ["{{ .Values.leaderElection.resourceName | default "descheduler" }}"]
   verbs: ["get", "patch", "delete"]
 {{- end }}
+{{- if and .Values.deschedulerPolicy .Values.deschedulerPolicy.metricsCollector .Values.deschedulerPolicy.metricsCollector.enabled }}
+- apiGroups: ["metrics.k8s.io"]
+  resources: ["pods", "nodes"]
+  verbs: ["get", "list"]
+{{- end }}
 {{- end -}}

charts/descheduler/templates/clusterrolebinding.yaml

@@ -12,5 +12,5 @@ roleRef:
 subjects:
   - kind: ServiceAccount
     name: {{ template "descheduler.serviceAccountName" . }}
-    namespace: {{ .Release.Namespace }}
+    namespace: {{ include "descheduler.namespace" . }}
 {{- end -}}

charts/descheduler/templates/configmap.yaml

@@ -1,12 +1,14 @@
+{{- if .Values.deschedulerPolicy }}
 apiVersion: v1
 kind: ConfigMap
 metadata:
   name: {{ template "descheduler.fullname" . }}
-  namespace: {{ .Release.Namespace }}
+  namespace: {{ include "descheduler.namespace" . }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
 data:
   policy.yaml: |
-    apiVersion: "descheduler/v1alpha1"
+    apiVersion: "{{ .Values.deschedulerPolicyAPIVersion }}"
     kind: "DeschedulerPolicy"
 {{ toYaml .Values.deschedulerPolicy | trim | indent 4 }}
+{{- end }}

charts/descheduler/templates/cronjob.yaml

@@ -3,7 +3,7 @@ apiVersion: {{ .Values.cronJobApiVersion | default "batch/v1" }}
 kind: CronJob
 metadata:
   name: {{ template "descheduler.fullname" . }}
-  namespace: {{ .Release.Namespace }}
+  namespace: {{ include "descheduler.namespace" . }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
 spec:
@@ -15,12 +15,15 @@ spec:
   {{- if .Values.startingDeadlineSeconds }}
   startingDeadlineSeconds: {{ .Values.startingDeadlineSeconds }}
   {{- end }}
-  {{- if .Values.successfulJobsHistoryLimit }}
+  {{- if ne .Values.successfulJobsHistoryLimit nil }}
   successfulJobsHistoryLimit: {{ .Values.successfulJobsHistoryLimit }}
   {{- end }}
-  {{- if .Values.failedJobsHistoryLimit }}
+  {{- if ne .Values.failedJobsHistoryLimit nil }}
   failedJobsHistoryLimit: {{ .Values.failedJobsHistoryLimit }}
   {{- end }}
+  {{- if .Values.timeZone }}
+  timeZone: {{ .Values.timeZone }}
+  {{- end }}
   jobTemplate:
     spec:
       {{- if .Values.ttlSecondsAfterFinished }}
@@ -48,6 +51,14 @@ spec:
           affinity:
             {{- toYaml . | nindent 12 }}
           {{- end }}
+          {{- with .Values.topologySpreadConstraints }}
+          topologySpreadConstraints:
+            {{- toYaml . | nindent 12 }}
+          {{- end }}
+          {{- if .Values.dnsConfig }}
+          dnsConfig:
+            {{- .Values.dnsConfig | toYaml | nindent 12 }}
+          {{- end }}
           {{- with .Values.tolerations }}
           tolerations:
             {{- toYaml . | nindent 12 }}
@@ -66,31 +77,33 @@ spec:
               image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default (printf "v%s" .Chart.AppVersion) }}"
               imagePullPolicy: {{ .Values.image.pullPolicy }}
               command:
-                - "/bin/descheduler"
+                {{- toYaml .Values.command | nindent 16 }}
               args:
-                - "--policy-config-file"
+                - --policy-config-file=/policy-dir/policy.yaml
-                - "/policy-dir/policy.yaml"
                 {{- range $key, $value := .Values.cmdOptions }}
-                - {{ printf "--%s" $key | quote }}
+                {{- if ne $value nil }}
-                {{- if $value }}
+                - {{ printf "--%s=%s" $key (toString $value) }}
-                - {{ $value | quote }}
+                {{- else }}
+                - {{ printf "--%s" $key }}
                 {{- end }}
                 {{- end }}
               livenessProbe:
                 {{- toYaml .Values.livenessProbe | nindent 16 }}
+              ports:
+                {{- toYaml .Values.ports | nindent 16 }}
               resources:
                 {{- toYaml .Values.resources | nindent 16 }}
+              {{- if .Values.securityContext }}
               securityContext:
-                allowPrivilegeEscalation: false
+                {{- toYaml .Values.securityContext | nindent 16 }}
-                capabilities:
+              {{- end }}
-                  drop:
-                    - ALL
-                privileged: false
-                readOnlyRootFilesystem: true
-                runAsNonRoot: true
               volumeMounts:
                 - mountPath: /policy-dir
                   name: policy-volume
+          {{- if .Values.podSecurityContext }}
+          securityContext:
+            {{- toYaml .Values.podSecurityContext | nindent 12 }}
+          {{- end }}
           volumes:
           - name: policy-volume
             configMap:

charts/descheduler/templates/deployment.yaml

@@ -3,11 +3,11 @@ apiVersion: apps/v1
 kind: Deployment
 metadata:
   name: {{ template "descheduler.fullname" . }}
-  namespace: {{ .Release.Namespace }}
+  namespace: {{ include "descheduler.namespace" . }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
 spec:
-  {{- if gt .Values.replicas 1.0}}
+  {{- if gt (.Values.replicas | int) 1 }}
   {{- if not .Values.leaderElection.enabled }}
   {{- fail "You must set leaderElection to use more than 1 replica"}}
   {{- end}}
@@ -31,50 +31,52 @@ spec:
         {{- .Values.podAnnotations | toYaml | nindent 8 }}
         {{- end }}
     spec:
+      {{- if .Values.dnsConfig }}
+      dnsConfig:
+        {{- .Values.dnsConfig | toYaml | nindent 8 }}
+      {{- end }}
       {{- if .Values.priorityClassName }}
       priorityClassName: {{ .Values.priorityClassName }}
       {{- end }}
       serviceAccountName: {{ template "descheduler.serviceAccountName" . }}
       {{- with .Values.imagePullSecrets }}
       imagePullSecrets:
-      {{- toYaml . | nindent 10 }}
+      {{- toYaml . | nindent 6 }}
       {{- end }}
       containers:
         - name: {{ .Chart.Name }}
           image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default (printf "v%s" .Chart.AppVersion) }}"
           imagePullPolicy: {{ .Values.image.pullPolicy }}
           command:
-            - "/bin/descheduler"
+            {{- toYaml .Values.command | nindent 12 }}
           args:
-            - "--policy-config-file"
+            - --policy-config-file=/policy-dir/policy.yaml
-            - "/policy-dir/policy.yaml"
+            - --descheduling-interval={{ required "deschedulingInterval required for running as Deployment" .Values.deschedulingInterval }}
-            - "--descheduling-interval"
-            - {{ required "deschedulingInterval required for running as Deployment" .Values.deschedulingInterval }}
             {{- range $key, $value := .Values.cmdOptions }}
-            - {{ printf "--%s" $key | quote }}
+            {{- if ne $value nil }}
-            {{- if $value }}
+            - {{ printf "--%s=%s" $key (toString $value) }}
-            - {{ $value | quote }}
+            {{- else }}
+            - {{ printf "--%s" $key }}
             {{- end }}
             {{- end }}
             {{- include "descheduler.leaderElection" . | nindent 12 }}
           ports:
-            - containerPort: 10258
+            {{- toYaml .Values.ports | nindent 12 }}
-              protocol: TCP
           livenessProbe:
             {{- toYaml .Values.livenessProbe | nindent 12 }}
           resources:
             {{- toYaml .Values.resources | nindent 12 }}
+          {{- if .Values.securityContext }}
           securityContext:
-            allowPrivilegeEscalation: false
+            {{- toYaml .Values.securityContext | nindent 12 }}
-            capabilities:
+          {{- end }}
-              drop:
-                - ALL
-            privileged: false
-            readOnlyRootFilesystem: true
-            runAsNonRoot: true
           volumeMounts:
             - mountPath: /policy-dir
               name: policy-volume
+      {{- if .Values.podSecurityContext }}
+      securityContext:
+        {{- toYaml .Values.podSecurityContext | nindent 8 }}
+      {{- end }}
       volumes:
         - name: policy-volume
           configMap:
@@ -87,6 +89,10 @@ spec:
       affinity:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      {{- with .Values.topologySpreadConstraints }}
+      topologySpreadConstraints:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
       {{- with .Values.tolerations }}
       tolerations:
         {{- toYaml . | nindent 8 }}

charts/descheduler/templates/service.yaml

@@ -6,9 +6,15 @@ metadata:
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
   name: {{ template "descheduler.fullname" . }}
-  namespace: {{ .Release.Namespace }}
+  namespace: {{ include "descheduler.namespace" . }}
 spec:
   clusterIP: None
+  {{- if .Values.service.ipFamilyPolicy }}
+  ipFamilyPolicy: {{ .Values.service.ipFamilyPolicy }}
+  {{- end }}
+  {{- if .Values.service.ipFamilies }}
+  ipFamilies: {{ toYaml .Values.service.ipFamilies | nindent 4 }}
+  {{- end }}
   ports:
   - name: http-metrics
     port: 10258

charts/descheduler/templates/serviceaccount.yaml

@@ -3,7 +3,7 @@ apiVersion: v1
 kind: ServiceAccount
 metadata:
   name: {{ template "descheduler.serviceAccountName" . }}
-  namespace: {{ .Release.Namespace }}
+  namespace: {{ include "descheduler.namespace" . }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
 {{- if .Values.serviceAccount.annotations }}

charts/descheduler/templates/servicemonitor.yaml

@@ -7,11 +7,14 @@ metadata:
   namespace: {{ .Values.serviceMonitor.namespace | default .Release.Namespace }}
   labels:
     {{- include "descheduler.labels" . | nindent 4 }}
+    {{- if .Values.serviceMonitor.additionalLabels }}
+    {{- toYaml .Values.serviceMonitor.additionalLabels | nindent 4 }}
+    {{- end }}
 spec:
   jobLabel: jobLabel
   namespaceSelector:
     matchNames:
-    - {{ .Release.Namespace }}
+    - {{ include "descheduler.namespace" . }}
   selector:
     matchLabels:
       {{- include "descheduler.selectorLabels" . | nindent 6 }}

charts/descheduler/tests/.gitignore

@@ -0,0 +1 @@
+__snapshot__

charts/descheduler/tests/cronjob_test.yaml

@@ -0,0 +1,17 @@
+suite: Test Descheduler CronJob
+
+templates:
+  - "*.yaml"
+
+release:
+  name: descheduler
+
+set:
+  kind: CronJob
+
+tests:
+  - it: creates CronJob when kind is set
+    template: templates/cronjob.yaml
+    asserts:
+      - isKind:
+          of: CronJob

charts/descheduler/tests/deployment_test.yaml

@@ -0,0 +1,49 @@
+suite: Test Descheduler Deployment
+
+templates:
+  - "*.yaml"
+
+release:
+  name: descheduler
+
+set:
+  kind: Deployment
+
+tests:
+  - it: creates Deployment when kind is set
+    template: templates/deployment.yaml
+    asserts:
+      - isKind:
+          of: Deployment
+  
+  - it: enables leader-election
+    set:
+      leaderElection:
+        enabled: true
+    template: templates/deployment.yaml
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --leader-elect=true
+
+  - it: support leader-election resourceNamespace
+    set:
+      leaderElection:
+        enabled: true
+        resourceNamespace: test
+    template: templates/deployment.yaml
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --leader-elect-resource-namespace=test
+
+  - it: support legacy leader-election resourceNamescape
+    set:
+      leaderElection:
+        enabled: true
+        resourceNamescape: typo
+    template: templates/deployment.yaml
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --leader-elect-resource-namespace=typo

charts/descheduler/values.yaml

@@ -18,13 +18,34 @@ resources:
   requests:
     cpu: 500m
     memory: 256Mi
-  # limits:
+  limits:
-  #   cpu: 100m
+    cpu: 500m
-  #   memory: 128Mi
+    memory: 256Mi
+
+ports:
+  - containerPort: 10258
+    protocol: TCP
+
+securityContext:
+  allowPrivilegeEscalation: false
+  capabilities:
+    drop:
+      - ALL
+  privileged: false
+  readOnlyRootFilesystem: true
+  runAsNonRoot: true
+  runAsUser: 1000
+
+# podSecurityContext -- [Security context for pod](https://kubernetes.io/docs/tasks/configure-pod-container/security-context/)
+podSecurityContext: {}
+  # fsGroup: 1000
 
 nameOverride: ""
 fullnameOverride: ""
 
+# -- Override the deployment namespace; defaults to .Release.Namespace
+namespaceOverride: ""
+
 # labels that'll be applied to all resources
 commonLabels: {}
 
@@ -35,6 +56,7 @@ suspend: false
 # successfulJobsHistoryLimit: 3
 # failedJobsHistoryLimit: 1
 # ttlSecondsAfterFinished 600
+# timeZone: Etc/UTC
 
 # Required when running as a Deployment
 deschedulingInterval: 5m
@@ -55,51 +77,78 @@ leaderElection: {}
 #  retryPeriod: 2s
 #  resourceLock: "leases"
 #  resourceName: "descheduler"
-#  resourceNamescape: "kube-system"
+#  resourceNamespace: "kube-system"
+
+command:
+- "/bin/descheduler"
 
 cmdOptions:
   v: 3
 
+# Recommended to use the latest Policy API version supported by the Descheduler app version
+deschedulerPolicyAPIVersion: "descheduler/v1alpha2"
+
+# deschedulerPolicy contains the policies the descheduler will execute.
+# To use policies stored in an existing configMap use:
+# NOTE: The name of the cm should comply to {{ template "descheduler.fullname" . }}
+# deschedulerPolicy: {}
 deschedulerPolicy:
   # nodeSelector: "key1=value1,key2=value2"
   # maxNoOfPodsToEvictPerNode: 10
   # maxNoOfPodsToEvictPerNamespace: 10
+  # metricsCollector:
+  #   enabled: true
   # ignorePvcPods: true
   # evictLocalStoragePods: true
-  strategies:
+  # evictDaemonSetPods: true
-    RemoveDuplicates:
+  # tracing:
-      enabled: true
+  #   collectorEndpoint: otel-collector.observability.svc.cluster.local:4317
-    RemovePodsHavingTooManyRestarts:
+  #   transportCert: ""
-      enabled: true
+  #   serviceName: ""
-      params:
+  #   serviceNamespace: ""
-        podsHavingTooManyRestarts:
+  #   sampleRate: 1.0
-          podRestartThreshold: 100
+  #   fallbackToNoOpProviderOnError: true
-          includingInitContainers: true
+  profiles:
-    RemovePodsViolatingNodeTaints:
+    - name: default
-      enabled: true
+      pluginConfig:
-    RemovePodsViolatingNodeAffinity:
+        - name: DefaultEvictor
-      enabled: true
+          args:
-      params:
+            ignorePvcPods: true
-        nodeAffinityType:
+            evictLocalStoragePods: true
-        - requiredDuringSchedulingIgnoredDuringExecution
+        - name: RemoveDuplicates
-    RemovePodsViolatingInterPodAntiAffinity:
+        - name: RemovePodsHavingTooManyRestarts
-      enabled: true
+          args:
-    RemovePodsViolatingTopologySpreadConstraint:
+            podRestartThreshold: 100
-      enabled: true
+            includingInitContainers: true
-      params:
+        - name: RemovePodsViolatingNodeAffinity
-        includeSoftConstraints: false
+          args:
-    LowNodeUtilization:
+            nodeAffinityType:
-      enabled: true
+            - requiredDuringSchedulingIgnoredDuringExecution
-      params:
+        - name: RemovePodsViolatingNodeTaints
-        nodeResourceUtilizationThresholds:
+        - name: RemovePodsViolatingInterPodAntiAffinity
-          thresholds:
+        - name: RemovePodsViolatingTopologySpreadConstraint
-            cpu: 20
+        - name: LowNodeUtilization
-            memory: 20
+          args:
-            pods: 20
+            thresholds:
-          targetThresholds:
+              cpu: 20
-            cpu: 50
+              memory: 20
-            memory: 50
+              pods: 20
-            pods: 50
+            targetThresholds:
+              cpu: 50
+              memory: 50
+              pods: 50
+      plugins:
+        balance:
+          enabled:
+            - RemoveDuplicates
+            - RemovePodsViolatingTopologySpreadConstraint
+            - LowNodeUtilization
+        deschedule:
+          enabled:
+            - RemovePodsHavingTooManyRestarts
+            - RemovePodsViolatingNodeTaints
+            - RemovePodsViolatingNodeAffinity
+            - RemovePodsViolatingInterPodAntiAffinity
 
 priorityClassName: system-cluster-critical
 
@@ -125,6 +174,13 @@ affinity: {}
 #              values:
 #                - descheduler
 #        topologyKey: "kubernetes.io/hostname"
+topologySpreadConstraints: []
+# - maxSkew: 1
+#   topologyKey: kubernetes.io/hostname
+#   whenUnsatisfiable: DoNotSchedule
+#   labelSelector:
+#     matchLabels:
+#       app.kubernetes.io/name: descheduler
 tolerations: []
 # - key: 'management'
 #   operator: 'Equal'
@@ -148,6 +204,8 @@ podAnnotations: {}
 
 podLabels: {}
 
+dnsConfig: {}
+
 livenessProbe:
   failureThreshold: 3
   httpGet:
@@ -159,11 +217,24 @@ livenessProbe:
 
 service:
   enabled: false
+  # @param service.ipFamilyPolicy [string], support SingleStack, PreferDualStack and RequireDualStack
+  #
+  ipFamilyPolicy: ""
+  # @param service.ipFamilies [array] List of IP families (e.g. IPv4, IPv6) assigned to the service.
+  # Ref: https://kubernetes.io/docs/concepts/services-networking/dual-stack/
+  # E.g.
+  # ipFamilies:
+  #   - IPv6
+  #   - IPv4
+  ipFamilies: []
 
 serviceMonitor:
   enabled: false
   # The namespace where Prometheus expects to find service monitors.
   # namespace: ""
+  # Add custom labels to the ServiceMonitor resource
+  additionalLabels: {}
+    # prometheus: kube-prometheus-stack
   interval: ""
   # honorLabels: true
   insecureSkipVerify: true

cmd/descheduler/app/options/options.go

@@ -18,17 +18,29 @@ limitations under the License.
 package options
 
 import (
+	"strings"
 	"time"
 
 	"github.com/spf13/pflag"
+
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	apiserver "k8s.io/apiserver/pkg/server"
 	apiserveroptions "k8s.io/apiserver/pkg/server/options"
 	clientset "k8s.io/client-go/kubernetes"
+
+	restclient "k8s.io/client-go/rest"
+	cliflag "k8s.io/component-base/cli/flag"
 	componentbaseconfig "k8s.io/component-base/config"
 	componentbaseoptions "k8s.io/component-base/config/options"
+	"k8s.io/component-base/featuregate"
+	"k8s.io/klog/v2"
+	metricsclient "k8s.io/metrics/pkg/client/clientset/versioned"
+
 	"sigs.k8s.io/descheduler/pkg/apis/componentconfig"
 	"sigs.k8s.io/descheduler/pkg/apis/componentconfig/v1alpha1"
 	deschedulerscheme "sigs.k8s.io/descheduler/pkg/descheduler/scheme"
+	"sigs.k8s.io/descheduler/pkg/features"
+	"sigs.k8s.io/descheduler/pkg/tracing"
 )
 
 const (
@@ -39,11 +51,17 @@ const (
 type DeschedulerServer struct {
 	componentconfig.DeschedulerConfiguration
 
-	Client         clientset.Interface
+	Client            clientset.Interface
-	EventClient    clientset.Interface
+	EventClient       clientset.Interface
-	SecureServing  *apiserveroptions.SecureServingOptionsWithLoopback
+	MetricsClient     metricsclient.Interface
-	DisableMetrics bool
+	SecureServing     *apiserveroptions.SecureServingOptionsWithLoopback
-	EnableHTTP2    bool
+	SecureServingInfo *apiserver.SecureServingInfo
+	DisableMetrics    bool
+	EnableHTTP2       bool
+	// FeatureGates enabled by the user
+	FeatureGates map[string]bool
+	// DefaultFeatureGates for internal accessing so unit tests can enable/disable specific features
+	DefaultFeatureGates featuregate.FeatureGate
 }
 
 // NewDeschedulerServer creates a new DeschedulerServer with default parameters
@@ -75,7 +93,9 @@ func newDefaultComponentConfig() (*componentconfig.DeschedulerConfiguration, err
 		},
 	}
 	deschedulerscheme.Scheme.Default(&versionedCfg)
-	cfg := componentconfig.DeschedulerConfiguration{}
+	cfg := componentconfig.DeschedulerConfiguration{
+		Tracing: componentconfig.TracingConfiguration{},
+	}
 	if err := deschedulerscheme.Scheme.Convert(&versionedCfg, &cfg, nil); err != nil {
 		return nil, err
 	}
@@ -84,15 +104,46 @@ 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", "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.ClientConnection.Kubeconfig, "kubeconfig", rs.ClientConnection.Kubeconfig, "File with kube configuration. Deprecated, use client-connection-kubeconfig instead.")
+	fs.StringVar(&rs.ClientConnection.Kubeconfig, "client-connection-kubeconfig", rs.ClientConnection.Kubeconfig, "File path to kube configuration for interacting with kubernetes apiserver.")
+	fs.Float32Var(&rs.ClientConnection.QPS, "client-connection-qps", rs.ClientConnection.QPS, "QPS to use for interacting with kubernetes apiserver.")
+	fs.Int32Var(&rs.ClientConnection.Burst, "client-connection-burst", rs.ClientConnection.Burst, "Burst to use for interacting with kubernetes apiserver.")
 	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.")
+	fs.BoolVar(&rs.DryRun, "dry-run", rs.DryRun, "Execute descheduler in dry run mode.")
 	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.")
+	fs.StringVar(&rs.Tracing.CollectorEndpoint, "otel-collector-endpoint", "", "Set this flag to the OpenTelemetry Collector Service Address")
+	fs.StringVar(&rs.Tracing.TransportCert, "otel-transport-ca-cert", "", "Path of the CA Cert that can be used to generate the client Certificate for establishing secure connection to the OTEL in gRPC mode")
+	fs.StringVar(&rs.Tracing.ServiceName, "otel-service-name", tracing.DefaultServiceName, "OTEL Trace name to be used with the resources")
+	fs.StringVar(&rs.Tracing.ServiceNamespace, "otel-trace-namespace", "", "OTEL Trace namespace to be used with the resources")
+	fs.Float64Var(&rs.Tracing.SampleRate, "otel-sample-rate", 1.0, "Sample rate to collect the Traces")
+	fs.BoolVar(&rs.Tracing.FallbackToNoOpProviderOnError, "otel-fallback-no-op-on-error", false, "Fallback to NoOp Tracer in case of error")
 	fs.BoolVar(&rs.EnableHTTP2, "enable-http2", false, "If http/2 should be enabled for the metrics and health check")
+	fs.Var(cliflag.NewMapStringBool(&rs.FeatureGates), "feature-gates", "A set of key=value pairs that describe feature gates for alpha/experimental features. "+
+		"Options are:\n"+strings.Join(features.DefaultMutableFeatureGate.KnownFeatures(), "\n"))
 
 	componentbaseoptions.BindLeaderElectionFlags(&rs.LeaderElection, fs)
 
 	rs.SecureServing.AddFlags(fs)
 }
+
+func (rs *DeschedulerServer) Apply() error {
+	err := features.DefaultMutableFeatureGate.SetFromMap(rs.FeatureGates)
+	if err != nil {
+		return err
+	}
+	rs.DefaultFeatureGates = features.DefaultMutableFeatureGate
+
+	// loopbackClientConfig is a config for a privileged loopback connection
+	var loopbackClientConfig *restclient.Config
+	var secureServing *apiserver.SecureServingInfo
+	if err := rs.SecureServing.ApplyTo(&secureServing, &loopbackClientConfig); err != nil {
+		klog.ErrorS(err, "failed to apply secure server configuration")
+		return err
+	}
+
+	secureServing.DisableHTTP2 = !rs.EnableHTTP2
+	rs.SecureServingInfo = secureServing
+
+	return nil
+}

cmd/descheduler/app/server.go

@@ -20,22 +20,22 @@ package app
 import (
 	"context"
 	"io"
-	"os"
 	"os/signal"
 	"syscall"
 
-	"k8s.io/apiserver/pkg/server/healthz"
+	"github.com/spf13/cobra"
 
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
 	"sigs.k8s.io/descheduler/pkg/descheduler"
+	"sigs.k8s.io/descheduler/pkg/tracing"
 
-	"github.com/spf13/cobra"
+	"k8s.io/apimachinery/pkg/util/runtime"
-
+	"k8s.io/apimachinery/pkg/watch"
-	apiserver "k8s.io/apiserver/pkg/server"
+	"k8s.io/apiserver/pkg/server/healthz"
 	"k8s.io/apiserver/pkg/server/mux"
-	restclient "k8s.io/client-go/rest"
+	"k8s.io/component-base/featuregate"
-	registry "k8s.io/component-base/logs/api/v1"
+	"k8s.io/component-base/logs"
-	jsonLog "k8s.io/component-base/logs/json"
+	logsapi "k8s.io/component-base/logs/api/v1"
 	_ "k8s.io/component-base/logs/json/register"
 	"k8s.io/component-base/metrics/legacyregistry"
 	"k8s.io/klog/v2"
@@ -48,76 +48,78 @@ func NewDeschedulerCommand(out io.Writer) *cobra.Command {
 		klog.ErrorS(err, "unable to initialize server")
 	}
 
+	featureGate := featuregate.NewFeatureGate()
+	logConfig := logsapi.NewLoggingConfiguration()
+
 	cmd := &cobra.Command{
 		Use:   "descheduler",
 		Short: "descheduler",
-		Long:  `The descheduler evicts pods which may be bound to less desired nodes`,
+		Long:  "The descheduler evicts pods which may be bound to less desired nodes",
-		Run: func(cmd *cobra.Command, args []string) {
+		PreRunE: func(cmd *cobra.Command, args []string) error {
-			// s.Logs.Config.Format = s.Logging.Format
+			logs.InitLogs()
-
+			if logsapi.ValidateAndApply(logConfig, featureGate); err != nil {
-			// LoopbackClientConfig is a config for a privileged loopback connection
+				return err
-			var LoopbackClientConfig *restclient.Config
+			}
-			var SecureServing *apiserver.SecureServingInfo
+			descheduler.SetupPlugins()
-			if err := s.SecureServing.ApplyTo(&SecureServing, &LoopbackClientConfig); err != nil {
+			return nil
-				klog.ErrorS(err, "failed to apply secure server configuration")
+		},
-				return
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err = s.Apply(); err != nil {
+				klog.ErrorS(err, "failed to apply")
+				return err
 			}
 
-			SecureServing.DisableHTTP2 = !s.EnableHTTP2
+			if err = Run(cmd.Context(), s); err != nil {
-
+				klog.ErrorS(err, "failed to run descheduler server")
-			var factory registry.LogFormatFactory
+				return err
-
-			if s.Logging.Format == "json" {
-				factory = jsonLog.Factory{}
 			}
 
-			if factory == nil {
+			return nil
-				klog.ClearLogger()
-			} else {
-				log, logrFlush := factory.Create(registry.LoggingConfiguration{
-					Format:    s.Logging.Format,
-					Verbosity: s.Logging.Verbosity,
-				})
-				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()
 	s.AddFlags(flags)
+
+	runtime.Must(logsapi.AddFeatureGates(featureGate))
+	logsapi.AddFlags(logConfig, flags)
+
 	return cmd
 }
 
-func Run(ctx context.Context, rs *options.DeschedulerServer) error {
+func Run(rootCtx context.Context, rs *options.DeschedulerServer) error {
-	return descheduler.Run(ctx, rs)
+	ctx, done := signal.NotifyContext(rootCtx, syscall.SIGINT, syscall.SIGTERM)
-}
 
-func SetupLogs() {
+	pathRecorderMux := mux.NewPathRecorderMux("descheduler")
-	klog.SetOutput(os.Stdout)
+	if !rs.DisableMetrics {
-	klog.InitFlags(nil)
+		pathRecorderMux.Handle("/metrics", legacyregistry.HandlerWithReset())
+	}
+
+	healthz.InstallHandler(pathRecorderMux, healthz.NamedCheck("Descheduler", healthz.PingHealthz.Check))
+
+	stoppedCh, _, err := rs.SecureServingInfo.Serve(pathRecorderMux, 0, ctx.Done())
+	if err != nil {
+		klog.Fatalf("failed to start secure server: %v", err)
+		return err
+	}
+
+	err = tracing.NewTracerProvider(ctx, rs.Tracing.CollectorEndpoint, rs.Tracing.TransportCert, rs.Tracing.ServiceName, rs.Tracing.ServiceNamespace, rs.Tracing.SampleRate, rs.Tracing.FallbackToNoOpProviderOnError)
+	if err != nil {
+		klog.ErrorS(err, "failed to create tracer provider")
+	}
+	defer tracing.Shutdown(ctx)
+
+	// increase the fake watch channel so the dry-run mode can be run
+	// over a cluster with thousands of pods
+	watch.DefaultChanSize = 100000
+	err = descheduler.Run(ctx, rs)
+	if err != nil {
+		return err
+	}
+
+	done()
+	// wait for metrics server to close
+	<-stoppedCh
+
+	return nil
 }

cmd/descheduler/descheduler.go

@@ -21,14 +21,8 @@ import (
 
 	"k8s.io/component-base/cli"
 	"sigs.k8s.io/descheduler/cmd/descheduler/app"
-	"sigs.k8s.io/descheduler/pkg/descheduler"
 )
 
-func init() {
-	app.SetupLogs()
-	descheduler.SetupPlugins()
-}
-
 func main() {
 	out := os.Stdout
 	cmd := app.NewDeschedulerCommand(out)

docs/cli/descheduler.md

@@ -0,0 +1,70 @@
+## descheduler
+
+descheduler
+
+### Synopsis
+
+The descheduler evicts pods which may be bound to less desired nodes
+
+```
+descheduler [flags]
+```
+
+### Options
+
+```
+      --bind-address ip                          The IP address on which to listen for the --secure-port port. The associated interface(s) must be reachable by the rest of the cluster, and by CLI/web clients. If blank or an unspecified address (0.0.0.0 or ::), all interfaces and IP address families will be used. (default 0.0.0.0)
+      --cert-dir string                          The directory where the TLS certs are located. If --tls-cert-file and --tls-private-key-file are provided, this flag will be ignored. (default "apiserver.local.config/certificates")
+      --client-connection-burst int32            Burst to use for interacting with kubernetes apiserver.
+      --client-connection-kubeconfig string      File path to kube configuration for interacting with kubernetes apiserver.
+      --client-connection-qps float32            QPS to use for interacting with kubernetes apiserver.
+      --descheduling-interval duration           Time interval between two consecutive descheduler executions. Setting this value instructs the descheduler to run in a continuous loop at the interval specified.
+      --disable-http2-serving                    If true, HTTP2 serving will be disabled [default=false]
+      --disable-metrics                          Disables metrics. The metrics are by default served through https://localhost:10258/metrics. Secure address, resp. port can be changed through --bind-address, resp. --secure-port flags.
+      --dry-run                                  Execute descheduler in dry run mode.
+      --enable-http2                             If http/2 should be enabled for the metrics and health check
+      --feature-gates mapStringBool              A set of key=value pairs that describe feature gates for alpha/experimental features. Options are:
+                                                 AllAlpha=true|false (ALPHA - default=false)
+                                                 AllBeta=true|false (BETA - default=false)
+                                                 EvictionsInBackground=true|false (ALPHA - default=false)
+  -h, --help                                     help for descheduler
+      --http2-max-streams-per-connection int     The limit that the server gives to clients for the maximum number of streams in an HTTP/2 connection. Zero means to use golang's default.
+      --kubeconfig string                        File with kube configuration. Deprecated, use client-connection-kubeconfig instead.
+      --leader-elect                             Start a leader election client and gain leadership before executing the main loop. Enable this when running replicated components for high availability.
+      --leader-elect-lease-duration duration     The duration that non-leader candidates will wait after observing a leadership renewal until attempting to acquire leadership of a led but unrenewed leader slot. This is effectively the maximum duration that a leader can be stopped before it is replaced by another candidate. This is only applicable if leader election is enabled. (default 2m17s)
+      --leader-elect-renew-deadline duration     The interval between attempts by the acting master to renew a leadership slot before it stops leading. This must be less than the lease duration. This is only applicable if leader election is enabled. (default 1m47s)
+      --leader-elect-resource-lock string        The type of resource object that is used for locking during leader election. Supported options are 'leases'. (default "leases")
+      --leader-elect-resource-name string        The name of resource object that is used for locking during leader election. (default "descheduler")
+      --leader-elect-resource-namespace string   The namespace of resource object that is used for locking during leader election. (default "kube-system")
+      --leader-elect-retry-period duration       The duration the clients should wait between attempting acquisition and renewal of a leadership. This is only applicable if leader election is enabled. (default 26s)
+      --log-flush-frequency duration             Maximum number of seconds between log flushes (default 5s)
+      --log-json-info-buffer-size quantity       [Alpha] In JSON format with split output streams, the info messages can be buffered for a while to increase performance. The default value of zero bytes disables buffering. The size can be specified as number of bytes (512), multiples of 1000 (1K), multiples of 1024 (2Ki), or powers of those (3M, 4G, 5Mi, 6Gi). Enable the LoggingAlphaOptions feature gate to use this.
+      --log-json-split-stream                    [Alpha] In JSON format, write error messages to stderr and info messages to stdout. The default is to write a single stream to stdout. Enable the LoggingAlphaOptions feature gate to use this.
+      --log-text-info-buffer-size quantity       [Alpha] In text format with split output streams, the info messages can be buffered for a while to increase performance. The default value of zero bytes disables buffering. The size can be specified as number of bytes (512), multiples of 1000 (1K), multiples of 1024 (2Ki), or powers of those (3M, 4G, 5Mi, 6Gi). Enable the LoggingAlphaOptions feature gate to use this.
+      --log-text-split-stream                    [Alpha] In text format, write error messages to stderr and info messages to stdout. The default is to write a single stream to stdout. Enable the LoggingAlphaOptions feature gate to use this.
+      --logging-format string                    Sets the log format. Permitted formats: "json" (gated by LoggingBetaOptions), "text". (default "text")
+      --otel-collector-endpoint string           Set this flag to the OpenTelemetry Collector Service Address
+      --otel-fallback-no-op-on-error             Fallback to NoOp Tracer in case of error
+      --otel-sample-rate float                   Sample rate to collect the Traces (default 1)
+      --otel-service-name string                 OTEL Trace name to be used with the resources (default "descheduler")
+      --otel-trace-namespace string              OTEL Trace namespace to be used with the resources
+      --otel-transport-ca-cert string            Path of the CA Cert that can be used to generate the client Certificate for establishing secure connection to the OTEL in gRPC mode
+      --permit-address-sharing                   If true, SO_REUSEADDR will be used when binding the port. This allows binding to wildcard IPs like 0.0.0.0 and specific IPs in parallel, and it avoids waiting for the kernel to release sockets in TIME_WAIT state. [default=false]
+      --permit-port-sharing                      If true, SO_REUSEPORT will be used when binding the port, which allows more than one instance to bind on the same address and port. [default=false]
+      --policy-config-file string                File with descheduler policy configuration.
+      --secure-port int                          The port on which to serve HTTPS with authentication and authorization. If 0, don't serve HTTPS at all. (default 10258)
+      --tls-cert-file string                     File containing the default x509 Certificate for HTTPS. (CA cert, if any, concatenated after server cert). If HTTPS serving is enabled, and --tls-cert-file and --tls-private-key-file are not provided, a self-signed certificate and key are generated for the public address and saved to the directory specified by --cert-dir.
+      --tls-cipher-suites strings                Comma-separated list of cipher suites for the server. If omitted, the default Go cipher suites will be used. 
+                                                 Preferred values: TLS_AES_128_GCM_SHA256, TLS_AES_256_GCM_SHA384, TLS_CHACHA20_POLY1305_SHA256, TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA, TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA, TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384, TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305, TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256, TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256, TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384, TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305, TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256. 
+                                                 Insecure values: TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, TLS_ECDHE_ECDSA_WITH_RC4_128_SHA, TLS_ECDHE_RSA_WITH_3DES_EDE_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, TLS_ECDHE_RSA_WITH_RC4_128_SHA, TLS_RSA_WITH_3DES_EDE_CBC_SHA, TLS_RSA_WITH_AES_128_CBC_SHA, TLS_RSA_WITH_AES_128_CBC_SHA256, TLS_RSA_WITH_AES_128_GCM_SHA256, TLS_RSA_WITH_AES_256_CBC_SHA, TLS_RSA_WITH_AES_256_GCM_SHA384, TLS_RSA_WITH_RC4_128_SHA.
+      --tls-min-version string                   Minimum TLS version supported. Possible values: VersionTLS10, VersionTLS11, VersionTLS12, VersionTLS13
+      --tls-private-key-file string              File containing the default x509 private key matching --tls-cert-file.
+      --tls-sni-cert-key namedCertKey            A pair of x509 certificate and private key file paths, optionally suffixed with a list of domain patterns which are fully qualified domain names, possibly with prefixed wildcard segments. The domain patterns also allow IP addresses, but IPs should only be used if the apiserver has visibility to the IP address requested by a client. If no domain patterns are provided, the names of the certificate are extracted. Non-wildcard matches trump over wildcard matches, explicit domain patterns trump over extracted names. For multiple key/certificate pairs, use the --tls-sni-cert-key multiple times. Examples: "example.crt,example.key" or "foo.crt,foo.key:*.foo.com,foo.com". (default [])
+  -v, --v Level                                  number for the log level verbosity
+      --vmodule pattern=N,...                    comma-separated list of pattern=N settings for file-filtered logging (only works for text log format)
+```
+
+### SEE ALSO
+
+* [descheduler version](descheduler_version.md)	 - Version of descheduler
+

docs/cli/descheduler_version.md

@@ -0,0 +1,22 @@
+## descheduler version
+
+Version of descheduler
+
+### Synopsis
+
+Prints the version of descheduler.
+
+```
+descheduler version [flags]
+```
+
+### Options
+
+```
+  -h, --help   help for version
+```
+
+### SEE ALSO
+
+* [descheduler](descheduler.md)	 - descheduler
+

docs/contributor-guide.md

@@ -3,7 +3,7 @@
 ## Required Tools
 
 - [Git](https://git-scm.com/downloads)
-- [Go 1.16+](https://golang.org/dl/)
+- [Go 1.23+](https://golang.org/dl/)
 - [Docker](https://docs.docker.com/install/)
 - [kubectl](https://kubernetes.io/docs/tasks/tools/install-kubectl)
 - [kind v0.10.0+](https://kind.sigs.k8s.io/)
@@ -20,7 +20,7 @@ make
 
 Run descheduler.
 ```sh
-./_output/bin/descheduler --kubeconfig <path to kubeconfig> --policy-config-file <path-to-policy-file> --v 1
+./_output/bin/descheduler --client-connection-kubeconfig <path to kubeconfig> --policy-config-file <path-to-policy-file> --v 1
 ```
 
 View all CLI options.

docs/user-guide.md

@@ -4,19 +4,13 @@ Starting with descheduler release v0.10.0 container images are available in the
 
 Descheduler Version | Container Image                                 | Architectures           |
 ------------------- |-------------------------------------------------|-------------------------|
-v0.26.1             | registry.k8s.io/descheduler/descheduler:v0.26.1 | AMD64<br>ARM64<br>ARMv7 |
+v0.32.2             | registry.k8s.io/descheduler/descheduler:v0.32.2 | AMD64<br>ARM64<br>ARMv7 |
-v0.26.0             | registry.k8s.io/descheduler/descheduler:v0.26.0 | AMD64<br>ARM64<br>ARMv7 |
+v0.32.1             | registry.k8s.io/descheduler/descheduler:v0.32.1 | AMD64<br>ARM64<br>ARMv7 |
-v0.25.1             | registry.k8s.io/descheduler/descheduler:v0.25.1 | AMD64<br>ARM64<br>ARMv7 |
+v0.32.0             | registry.k8s.io/descheduler/descheduler:v0.32.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.25.0             | registry.k8s.io/descheduler/descheduler:v0.25.0 | AMD64<br>ARM64<br>ARMv7 |
+v0.31.0             | registry.k8s.io/descheduler/descheduler:v0.31.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.24.1             | registry.k8s.io/descheduler/descheduler:v0.24.1 | AMD64<br>ARM64<br>ARMv7 |
+v0.30.1             | registry.k8s.io/descheduler/descheduler:v0.30.1 | AMD64<br>ARM64<br>ARMv7 |
-v0.24.0             | registry.k8s.io/descheduler/descheduler:v0.24.0 | AMD64<br>ARM64<br>ARMv7 |
+v0.30.0             | registry.k8s.io/descheduler/descheduler:v0.30.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.23.1             | registry.k8s.io/descheduler/descheduler:v0.23.1 | AMD64<br>ARM64<br>ARMv7 |
+v0.29.0             | registry.k8s.io/descheduler/descheduler:v0.29.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.22.0             | registry.k8s.io/descheduler/descheduler:v0.22.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.21.0             | registry.k8s.io/descheduler/descheduler:v0.21.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.20.0             | registry.k8s.io/descheduler/descheduler:v0.20.0 | AMD64<br>ARM64          |
-v0.19.0             | registry.k8s.io/descheduler/descheduler:v0.19.0 | AMD64                   |
-v0.18.0             | registry.k8s.io/descheduler/descheduler:v0.18.0 | AMD64                   |
-v0.10.0             | registry.k8s.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
@@ -31,65 +25,7 @@ kind load docker-image registry.k8s.io/descheduler/descheduler:v0.20.0
 The [examples](https://github.com/kubernetes-sigs/descheduler/tree/master/examples) directory has descheduler policy configuration examples.
 
 ## CLI Options
-The descheduler has many CLI options that can be used to override its default behavior.
+The descheduler has many CLI options that can be used to override its default behavior. Please check the [CLI Options](./cli/descheduler.md) documentation for details
-```
-descheduler --help
-The descheduler evicts pods which may be bound to less desired nodes
-
-Usage:
-  descheduler [flags]
-  descheduler [command]
-
-Available Commands:
-  completion  generate the autocompletion script for the specified shell
-  help        Help about any command
-  version     Version of descheduler
-
-Flags:
-      --add-dir-header                           If true, adds the file directory to the header of the log messages (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --alsologtostderr                          log to standard error as well as files (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --bind-address ip                          The IP address on which to listen for the --secure-port port. The associated interface(s) must be reachable by the rest of the cluster, and by CLI/web clients. If blank or an unspecified address (0.0.0.0 or ::), all interfaces will be used. (default 0.0.0.0)
-      --cert-dir string                          The directory where the TLS certs are located. If --tls-cert-file and --tls-private-key-file are provided, this flag will be ignored. (default "apiserver.local.config/certificates")
-      --descheduling-interval duration           Time interval between two consecutive descheduler executions. Setting this value instructs the descheduler to run in a continuous loop at the interval specified.
-      --disable-metrics                          Disables metrics. The metrics are by default served through https://localhost:10258/metrics. Secure address, resp. port can be changed through --bind-address, resp. --secure-port flags.
-      --dry-run                                  execute descheduler in dry run mode.
-  -h, --help                                     help for descheduler
-      --http2-max-streams-per-connection int     The limit that the server gives to clients for the maximum number of streams in an HTTP/2 connection. Zero means to use golang's default.
-      --kubeconfig string                        File with  kube configuration.
-      --leader-elect                             Start a leader election client and gain leadership before executing the main loop. Enable this when running replicated components for high availability.
-      --leader-elect-lease-duration duration     The duration that non-leader candidates will wait after observing a leadership renewal until attempting to acquire leadership of a led but unrenewed leader slot. This is effectively the maximum duration that a leader can be stopped before it is replaced by another candidate. This is only applicable if leader election is enabled. (default 15s)
-      --leader-elect-renew-deadline duration     The interval between attempts by the acting master to renew a leadership slot before it stops leading. This must be less than or equal to the lease duration. This is only applicable if leader election is enabled. (default 10s)
-      --leader-elect-resource-lock string        The type of resource object that is used for locking during leader election. Supported options are 'endpoints', 'configmaps', 'leases', 'endpointsleases' and 'configmapsleases'. (default "leases")
-      --leader-elect-resource-name string        The name of resource object that is used for locking during leader election. (default "descheduler")
-      --leader-elect-resource-namespace string   The namespace of resource object that is used for locking during leader election. (default "kube-system")
-      --leader-elect-retry-period duration       The duration the clients should wait between attempting acquisition and renewal of a leadership. This is only applicable if leader election is enabled. (default 2s)
-      --log-backtrace-at traceLocation           when logging hits line file:N, emit a stack trace (default :0) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --log_dir string                           If non-empty, write log files in this directory (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --log_file string                          If non-empty, use this log file (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --log_file_max_size uint                   Defines the maximum size a log file can grow to. Unit is megabytes. If the value is 0, the maximum file size is unlimited. (default 1800) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --log-flush-frequency duration             Maximum number of seconds between log flushes (default 5s)
-      --logging-format string                    Sets the log format. Permitted formats: "text", "json". Non-default formats don't honor these flags: --add-dir-header, --alsologtostderr, --log-backtrace-at, --log_dir, --log_file, --log_file_max_size, --logtostderr, --skip-headers, --skip-log-headers, --stderrthreshold, --log-flush-frequency.\nNon-default choices are currently alpha and subject to change without warning. (default "text")
-      --logtostderr                              log to standard error instead of files (default true) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --one-output                               If true, only write logs to their native severity level (vs also writing to each lower severity level) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --permit-address-sharing                   If true, SO_REUSEADDR will be used when binding the port. This allows binding to wildcard IPs like 0.0.0.0 and specific IPs in parallel, and it avoids waiting for the kernel to release sockets in TIME_WAIT state. [default=false]
-      --permit-port-sharing                      If true, SO_REUSEPORT will be used when binding the port, which allows more than one instance to bind on the same address and port. [default=false]
-      --policy-config-file string                File with descheduler policy configuration.
-      --secure-port int                          The port on which to serve HTTPS with authentication and authorization. If 0, don't serve HTTPS at all. (default 10258)
-      --skip-headers                             If true, avoid header prefixes in the log messages (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --skip-log-headers                         If true, avoid headers when opening log files (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --stderrthreshold severity                 logs at or above this threshold go to stderr (default 2) (DEPRECATED: will be removed in a future release, see https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/2845-deprecate-klog-specific-flags-in-k8s-components)
-      --tls-cert-file string                     File containing the default x509 Certificate for HTTPS. (CA cert, if any, concatenated after server cert). If HTTPS serving is enabled, and --tls-cert-file and --tls-private-key-file are not provided, a self-signed certificate and key are generated for the public address and saved to the directory specified by --cert-dir.
-      --tls-cipher-suites strings                Comma-separated list of cipher suites for the server. If omitted, the default Go cipher suites will be used. 
-                                                 Preferred values: TLS_AES_128_GCM_SHA256, TLS_AES_256_GCM_SHA384, TLS_CHACHA20_POLY1305_SHA256, TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA, TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA, TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384, TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305, TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256, TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256, TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384, TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305, TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256, TLS_RSA_WITH_AES_128_CBC_SHA, TLS_RSA_WITH_AES_128_GCM_SHA256, TLS_RSA_WITH_AES_256_CBC_SHA, TLS_RSA_WITH_AES_256_GCM_SHA384. 
-                                                 Insecure values: TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, TLS_ECDHE_ECDSA_WITH_RC4_128_SHA, TLS_ECDHE_RSA_WITH_3DES_EDE_CBC_SHA, TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, TLS_ECDHE_RSA_WITH_RC4_128_SHA, TLS_RSA_WITH_3DES_EDE_CBC_SHA, TLS_RSA_WITH_AES_128_CBC_SHA256, TLS_RSA_WITH_RC4_128_SHA.
-      --tls-min-version string                   Minimum TLS version supported. Possible values: VersionTLS10, VersionTLS11, VersionTLS12, VersionTLS13
-      --tls-private-key-file string              File containing the default x509 private key matching --tls-cert-file.
-      --tls-sni-cert-key namedCertKey            A pair of x509 certificate and private key file paths, optionally suffixed with a list of domain patterns which are fully qualified domain names, possibly with prefixed wildcard segments. The domain patterns also allow IP addresses, but IPs should only be used if the apiserver has visibility to the IP address requested by a client. If no domain patterns are provided, the names of the certificate are extracted. Non-wildcard matches trump over wildcard matches, explicit domain patterns trump over extracted names. For multiple key/certificate pairs, use the --tls-sni-cert-key multiple times. Examples: "example.crt,example.key" or "foo.crt,foo.key:*.foo.com,foo.com". (default [])
-  -v, --v Level                                  number for the log level verbosity
-      --vmodule moduleSpec                       comma-separated list of pattern=N settings for file-filtered logging
-
-Use "descheduler [command] --help" for more information about a command.
-```
 
 ## Production Use Cases
 This section contains descriptions of real world production use cases.
@@ -111,19 +47,23 @@ descheduler -v=3 --evict-local-storage-pods --policy-config-file=pod-life-time.y
 This policy configuration file ensures that pods created more than 7 days ago are evicted.
 ```
 ---
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "PodLifeTime":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "PodLifeTime"
-      podLifeTime:
+      args:
-        maxPodLifeTimeSeconds: 604800 # pods run for a maximum of 7 days
+        maxPodLifeTimeSeconds: 604800
+    plugins:
+      deschedule:
+        enabled:
+          - "PodLifeTime"
 ```
 
 ### 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 following two strategies 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`.
 
 #### Balance high utilization nodes
@@ -131,17 +71,21 @@ Using `LowNodeUtilization`, descheduler will rebalance the cluster based on memo
 from nodes with memory utilization over 70% to nodes with memory utilization below 20%.
 
 ```
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "LowNodeUtilization":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "LowNodeUtilization"
-      nodeResourceUtilizationThresholds:
+      args:
         thresholds:
           "memory": 20
         targetThresholds:
           "memory": 70
+    plugins:
+      balance:
+        enabled:
+          - "LowNodeUtilization"
 ```
 
 #### Balance low utilization nodes
@@ -150,15 +94,19 @@ from nodes with memory utilization lower than 20%. This should be use `NodeResou
 The evicted pods will be compacted into minimal set of nodes.
 
 ```
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "HighNodeUtilization":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "HighNodeUtilization"
-      nodeResourceUtilizationThresholds:
+      args:
         thresholds:
           "memory": 20
+    plugins:
+      balance:
+        enabled:
+          - "HighNodeUtilization"
 ```
 
 ### Autoheal Node Problems

examples/failed-pods.yaml

@@ -1,14 +1,18 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "RemoveFailedPods":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "RemoveFailedPods"
-      failedPods:
+      args:
         reasons:
         - "OutOfcpu"
         - "CreateContainerConfigError"
         includingInitContainers: true
         excludeOwnerKinds:
         - "Job"
-        minPodLifetimeSeconds: 3600 # 1 hour
+        minPodLifetimeSeconds: 3600
+    plugins:
+      deschedule:
+        enabled:
+          - "RemoveFailedPods"

examples/high-node-utilization.yml

@@ -1,9 +1,13 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "HighNodeUtilization":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "HighNodeUtilization"
-      nodeResourceUtilizationThresholds:
+      args:
         thresholds:
           "memory": 20
+    plugins:
+      balance:
+        enabled:
+          - "HighNodeUtilization"

examples/low-node-utilization.yml

@@ -1,11 +1,15 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "LowNodeUtilization":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "LowNodeUtilization"
-      nodeResourceUtilizationThresholds:
+      args:
         thresholds:
           "memory": 20
         targetThresholds:
           "memory": 70
+    plugins:
+      balance:
+        enabled:
+          - "LowNodeUtilization"

examples/node-affinity.yml

@@ -1,8 +1,13 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "RemovePodsViolatingNodeAffinity":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "RemovePodsViolatingNodeAffinity"
-      nodeAffinityType:
+      args:
-      - "requiredDuringSchedulingIgnoredDuringExecution"
+        nodeAffinityType:
+        - "requiredDuringSchedulingIgnoredDuringExecution"
+    plugins:
+      deschedule:
+        enabled:
+          - "RemovePodsViolatingNodeAffinity"

examples/pod-life-time.yml

@@ -1,11 +1,15 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "PodLifeTime":
+  - name: ProfileName
-    enabled: true
+    pluginConfig:
-    params:
+    - name: "PodLifeTime"
-      podLifeTime:
+      args:
         maxPodLifeTimeSeconds: 604800 # 7 days
         states:
         - "Pending"
         - "PodInitializing"
+    plugins:
+      deschedule:
+        enabled:
+          - "PodLifeTime"

examples/policy.yaml

@@ -1,29 +1,36 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "RemoveDuplicates":
+  - name: ProfileName
-     enabled: true
+    pluginConfig:
-  "RemovePodsViolatingInterPodAntiAffinity":
+    - name: "RemoveDuplicates"
-     enabled: true
+    - name: "RemovePodsViolatingInterPodAntiAffinity"
-  "LowNodeUtilization":
+    - name: "LowNodeUtilization"
-     enabled: true
+      args:
-     params:
+        thresholds:
-       nodeResourceUtilizationThresholds:
+          "cpu" : 20
-         thresholds:
+          "memory": 20
-           "cpu" : 20
+          "pods": 20
-           "memory": 20
+        targetThresholds:
-           "pods": 20
+          "cpu" : 50
-         targetThresholds:
+          "memory": 50
-           "cpu" : 50
+          "pods": 50
-           "memory": 50
+    - name: "RemovePodsHavingTooManyRestarts"
-           "pods": 50
+      args:
-  "RemovePodsHavingTooManyRestarts":
+        podRestartThreshold: 100
-     enabled: true
+        includingInitContainers: true
-     params:
+    - name: "RemovePodsViolatingTopologySpreadConstraint"
-       podsHavingTooManyRestarts:
+      args:
-         podRestartThreshold: 100
+        constraints:
-         includingInitContainers: true
+          - DoNotSchedule
-  "RemovePodsViolatingTopologySpreadConstraint":
+          - ScheduleAnyway
-     enabled: true
+    plugins:
-     params:
+      deschedule:
-       includeSoftConstraints: true
+        enabled:
+          - "RemovePodsViolatingInterPodAntiAffinity"
+          - "RemovePodsHavingTooManyRestarts"
+      balance:
+        enabled:
+          - "RemoveDuplicates"
+          - "LowNodeUtilization"
+          - "RemovePodsViolatingTopologySpreadConstraint"

examples/too-many-restarts.yml

@@ -0,0 +1,15 @@
+apiVersion: "descheduler/v1alpha2"
+kind: "DeschedulerPolicy"
+profiles:
+  - name: ProfileName
+    pluginConfig:
+    - name: "RemovePodsHavingTooManyRestarts"
+      args:
+        podRestartThreshold: 100
+        includingInitContainers: true
+        states:
+        - CrashLoopBackOff
+    plugins:
+      deschedule:
+        enabled:
+          - "RemovePodsHavingTooManyRestarts"

examples/topology-spread-constraint.yaml

@@ -1,8 +1,14 @@
-apiVersion: "descheduler/v1alpha1"
+apiVersion: "descheduler/v1alpha2"
 kind: "DeschedulerPolicy"
-strategies:
+profiles:
-  "RemovePodsViolatingTopologySpreadConstraint":
+  - name: ProfileName
-     enabled: true
+    pluginConfig:
-     params:
+    - name: "RemovePodsViolatingTopologySpreadConstraint"
-       nodeFit: true
+      args:
-       includeSoftConstraints: true # Include 'ScheduleAnyways' constraints
+        constraints:
+          - DoNotSchedule
+          - ScheduleAnyway
+    plugins:
+      balance:
+        enabled:
+          - "RemovePodsViolatingTopologySpreadConstraint"

go.mod

@@ -1,110 +1,131 @@
 module sigs.k8s.io/descheduler
 
-go 1.19
+go 1.23.3
 
 require (
 	github.com/client9/misspell v0.3.4
-	github.com/google/go-cmp v0.5.9
+	github.com/google/go-cmp v0.6.0
-	github.com/spf13/cobra v1.6.0
+	github.com/spf13/cobra v1.8.1
 	github.com/spf13/pflag v1.0.5
-	k8s.io/api v0.26.10
+	go.opentelemetry.io/otel v1.28.0
-	k8s.io/apimachinery v0.26.10
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.28.0
-	k8s.io/apiserver v0.26.10
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.28.0
-	k8s.io/client-go v0.26.10
+	go.opentelemetry.io/otel/sdk v1.28.0
-	k8s.io/code-generator v0.26.10
+	go.opentelemetry.io/otel/trace v1.28.0
-	k8s.io/component-base v0.26.10
+	google.golang.org/grpc v1.65.0
-	k8s.io/component-helpers v0.26.10
+	k8s.io/api v0.32.0
-	k8s.io/klog/v2 v2.80.1
+	k8s.io/apimachinery v0.32.0
-	k8s.io/utils v0.0.0-20221128185143-99ec85e7a448
+	k8s.io/apiserver v0.32.0
-	sigs.k8s.io/mdtoc v1.0.1
+	k8s.io/client-go v0.32.0
+	k8s.io/code-generator v0.32.0
+	k8s.io/component-base v0.32.0
+	k8s.io/component-helpers v0.32.0
+	k8s.io/klog/v2 v2.130.1
+	k8s.io/metrics v0.32.0
+	k8s.io/utils v0.0.0-20241210054802-24370beab758
+	kubevirt.io/api v1.3.0
+	kubevirt.io/client-go v1.3.0
+	kubevirt.io/containerized-data-importer-api v1.60.1 // indirect; drops dependency on o/api
+	sigs.k8s.io/mdtoc v1.1.0
+	sigs.k8s.io/yaml v1.4.0
 )
 
 require (
+	cel.dev/expr v0.18.0 // indirect
 	github.com/BurntSushi/toml v0.3.1 // indirect
 	github.com/NYTimes/gziphandler v1.1.1 // indirect
-	github.com/antlr/antlr4/runtime/Go/antlr v1.4.10 // indirect
+	github.com/antlr4-go/antlr/v4 v4.13.0 // indirect
+	github.com/asaskevich/govalidator v0.0.0-20190424111038-f61b66f89f4a // indirect
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/blang/semver/v4 v4.0.0 // indirect
-	github.com/cenkalti/backoff/v4 v4.1.3 // indirect
+	github.com/cenkalti/backoff/v4 v4.3.0 // indirect
-	github.com/cespare/xxhash/v2 v2.1.2 // indirect
+	github.com/cespare/xxhash/v2 v2.3.0 // indirect
-	github.com/coreos/go-semver v0.3.0 // indirect
+	github.com/coreos/go-semver v0.3.1 // indirect
-	github.com/coreos/go-systemd/v22 v22.3.2 // indirect
+	github.com/coreos/go-systemd/v22 v22.5.0 // indirect
-	github.com/davecgh/go-spew v1.1.1 // indirect
+	github.com/cpuguy83/go-md2man/v2 v2.0.4 // indirect
-	github.com/emicklei/go-restful/v3 v3.9.0 // indirect
+	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
-	github.com/evanphx/json-patch v4.12.0+incompatible // indirect
+	github.com/emicklei/go-restful/v3 v3.11.0 // indirect
-	github.com/felixge/httpsnoop v1.0.3 // indirect
+	github.com/felixge/httpsnoop v1.0.4 // indirect
-	github.com/fsnotify/fsnotify v1.6.0 // indirect
+	github.com/fsnotify/fsnotify v1.7.0 // indirect
-	github.com/go-logr/logr v1.2.3 // indirect
+	github.com/fxamacker/cbor/v2 v2.7.0 // indirect
+	github.com/go-kit/kit v0.13.0 // indirect
+	github.com/go-kit/log v0.2.1 // indirect
+	github.com/go-logfmt/logfmt v0.6.0 // indirect
+	github.com/go-logr/logr v1.4.2 // indirect
 	github.com/go-logr/stdr v1.2.2 // indirect
-	github.com/go-logr/zapr v1.2.3 // indirect
+	github.com/go-logr/zapr v1.3.0 // indirect
-	github.com/go-openapi/jsonpointer v0.19.5 // indirect
+	github.com/go-openapi/jsonpointer v0.21.0 // indirect
-	github.com/go-openapi/jsonreference v0.20.0 // indirect
+	github.com/go-openapi/jsonreference v0.20.2 // indirect
-	github.com/go-openapi/swag v0.19.14 // indirect
+	github.com/go-openapi/swag v0.23.0 // indirect
 	github.com/gogo/protobuf v1.3.2 // indirect
-	github.com/golang/groupcache v0.0.0-20210331224755-41bb18bfe9da // indirect
+	github.com/golang/glog v1.2.1 // indirect
-	github.com/golang/protobuf v1.5.2 // indirect
+	github.com/golang/protobuf v1.5.4 // indirect
-	github.com/gomarkdown/markdown v0.0.0-20200824053859-8c8b3816f167 // indirect
+	github.com/gomarkdown/markdown v0.0.0-20210514010506-3b9f47219fe7 // indirect
-	github.com/google/cel-go v0.12.7 // indirect
+	github.com/google/btree v1.0.1 // indirect
-	github.com/google/gnostic v0.5.7-v3refs // indirect
+	github.com/google/cel-go v0.22.0 // indirect
-	github.com/google/gofuzz v1.1.0 // indirect
+	github.com/google/gnostic-models v0.6.8 // indirect
-	github.com/google/uuid v1.3.0 // indirect
+	github.com/google/gofuzz v1.2.0 // indirect
+	github.com/google/uuid v1.6.0 // indirect
+	github.com/gorilla/websocket v1.5.0 // indirect
 	github.com/grpc-ecosystem/go-grpc-prometheus v1.2.0 // indirect
-	github.com/grpc-ecosystem/grpc-gateway/v2 v2.7.0 // indirect
+	github.com/grpc-ecosystem/grpc-gateway/v2 v2.20.0 // indirect
-	github.com/imdario/mergo v0.3.6 // indirect
+	github.com/inconshreveable/mousetrap v1.1.0 // indirect
-	github.com/inconshreveable/mousetrap v1.0.1 // 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/mailru/easyjson v0.7.7 // indirect
-	github.com/matttproud/golang_protobuf_extensions v1.0.2 // 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/openshift/custom-resource-status v1.1.2 // indirect
 	github.com/pkg/errors v0.9.1 // indirect
-	github.com/prometheus/client_golang v1.14.0 // indirect
+	github.com/prometheus/client_golang v1.19.1 // indirect
-	github.com/prometheus/client_model v0.3.0 // indirect
+	github.com/prometheus/client_model v0.6.1 // indirect
-	github.com/prometheus/common v0.37.0 // indirect
+	github.com/prometheus/common v0.55.0 // indirect
-	github.com/prometheus/procfs v0.8.0 // indirect
+	github.com/prometheus/procfs v0.15.1 // indirect
-	github.com/stoewer/go-strcase v1.2.0 // indirect
+	github.com/russross/blackfriday/v2 v2.1.0 // indirect
-	go.etcd.io/etcd/api/v3 v3.5.5 // indirect
+	github.com/stoewer/go-strcase v1.3.0 // indirect
-	go.etcd.io/etcd/client/pkg/v3 v3.5.5 // indirect
+	github.com/x448/float16 v0.8.4 // indirect
-	go.etcd.io/etcd/client/v3 v3.5.5 // indirect
+	go.etcd.io/etcd/api/v3 v3.5.16 // indirect
-	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.35.0 // indirect
+	go.etcd.io/etcd/client/pkg/v3 v3.5.16 // indirect
-	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.35.1 // indirect
+	go.etcd.io/etcd/client/v3 v3.5.16 // indirect
-	go.opentelemetry.io/otel v1.10.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.53.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/internal/retry v1.10.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.53.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.10.0 // indirect
+	go.opentelemetry.io/otel/metric v1.28.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.10.0 // indirect
+	go.opentelemetry.io/proto/otlp v1.3.1 // indirect
-	go.opentelemetry.io/otel/metric v0.31.0 // indirect
+	go.uber.org/multierr v1.11.0 // indirect
-	go.opentelemetry.io/otel/sdk v1.10.0 // indirect
+	go.uber.org/zap v1.27.0 // indirect
-	go.opentelemetry.io/otel/trace v1.10.0 // indirect
+	golang.org/x/crypto v0.31.0 // indirect
-	go.opentelemetry.io/proto/otlp v0.19.0 // indirect
+	golang.org/x/exp v0.0.0-20240719175910-8a7402abbf56 // indirect
-	go.uber.org/atomic v1.7.0 // indirect
+	golang.org/x/mod v0.21.0 // indirect
-	go.uber.org/multierr v1.6.0 // indirect
+	golang.org/x/net v0.30.0 // indirect
-	go.uber.org/zap v1.19.0 // indirect
+	golang.org/x/oauth2 v0.23.0 // indirect
-	golang.org/x/crypto v0.14.0 // indirect
+	golang.org/x/sync v0.10.0 // indirect
-	golang.org/x/mod v0.9.0 // indirect
+	golang.org/x/sys v0.28.0 // indirect
-	golang.org/x/net v0.17.0 // indirect
+	golang.org/x/term v0.27.0 // indirect
-	golang.org/x/oauth2 v0.0.0-20220223155221-ee480838109b // indirect
+	golang.org/x/text v0.21.0 // indirect
-	golang.org/x/sync v0.1.0 // indirect
+	golang.org/x/time v0.7.0 // indirect
-	golang.org/x/sys v0.13.0 // indirect
+	golang.org/x/tools v0.26.0 // indirect
-	golang.org/x/term v0.13.0 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20240826202546-f6391c0de4c7 // indirect
-	golang.org/x/text v0.13.0 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20240826202546-f6391c0de4c7 // indirect
-	golang.org/x/time v0.0.0-20220210224613-90d013bbcef8 // indirect
+	google.golang.org/protobuf v1.35.1 // indirect
-	golang.org/x/tools v0.6.0 // indirect
+	gopkg.in/evanphx/json-patch.v4 v4.12.0 // indirect
-	google.golang.org/appengine v1.6.7 // indirect
-	google.golang.org/genproto v0.0.0-20220502173005-c8bf987b8c21 // indirect
-	google.golang.org/grpc v1.49.0 // indirect
-	google.golang.org/protobuf v1.28.1 // indirect
 	gopkg.in/inf.v0 v0.9.1 // indirect
-	gopkg.in/natefinch/lumberjack.v2 v2.0.0 // indirect
+	gopkg.in/natefinch/lumberjack.v2 v2.2.1 // indirect
 	gopkg.in/yaml.v2 v2.4.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
-	k8s.io/gengo v0.0.0-20220902162205-c0856e24416d // indirect
+	k8s.io/apiextensions-apiserver v0.30.0 // indirect
-	k8s.io/kms v0.26.10 // indirect
+	k8s.io/gengo/v2 v2.0.0-20240911193312-2b36238f13e9 // indirect
-	k8s.io/kube-openapi v0.0.0-20221012153701-172d655c2280 // indirect
+	k8s.io/kms v0.32.0 // indirect
-	sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.0.37 // indirect
+	k8s.io/kube-openapi v0.30.0 // indirect
-	sigs.k8s.io/json v0.0.0-20220713155537-f223a00ba0e2 // indirect
+	kubevirt.io/controller-lifecycle-operator-sdk/api v0.0.0-20220329064328-f3cc58c6ed90 // indirect
-	sigs.k8s.io/structured-merge-diff/v4 v4.2.3 // indirect
+	sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.31.0 // indirect
-	sigs.k8s.io/yaml v1.3.0 // indirect
+	sigs.k8s.io/json v0.0.0-20241010143419-9aa6b5e7a4b3 // indirect
+	sigs.k8s.io/structured-merge-diff/v4 v4.4.2 // indirect
 )
+
+replace k8s.io/kube-openapi => k8s.io/kube-openapi v0.0.0-20240430033511-f0e62f92d13f
+
+replace golang.org/x/net => golang.org/x/net v0.33.0
+
+replace golang.org/x/crypto => golang.org/x/crypto v0.31.0

hack/cherry_pick_pull.sh

@@ -0,0 +1,257 @@
+#!/usr/bin/env bash
+
+# Copyright 2015 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.
+
+# Usage Instructions: https://git.k8s.io/community/contributors/devel/sig-release/cherry-picks.md
+
+# Checkout a PR from GitHub. (Yes, this is sitting in a Git tree. How
+# meta.) Assumes you care about pulls from remote "upstream" and
+# checks them out to a branch named:
+#  automated-cherry-pick-of-<pr>-<target branch>-<timestamp>
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+REPO_ROOT="$(git rev-parse --show-toplevel)"
+declare -r REPO_ROOT
+cd "${REPO_ROOT}"
+
+STARTINGBRANCH=$(git symbolic-ref --short HEAD)
+declare -r STARTINGBRANCH
+declare -r REBASEMAGIC="${REPO_ROOT}/.git/rebase-apply"
+DRY_RUN=${DRY_RUN:-""}
+REGENERATE_DOCS=${REGENERATE_DOCS:-""}
+UPSTREAM_REMOTE=${UPSTREAM_REMOTE:-upstream}
+FORK_REMOTE=${FORK_REMOTE:-origin}
+MAIN_REPO_ORG=${MAIN_REPO_ORG:-$(git remote get-url "$UPSTREAM_REMOTE" | awk '{gsub(/http[s]:\/\/|git@/,"")}1' | awk -F'[@:./]' 'NR==1{print $3}')}
+MAIN_REPO_NAME=${MAIN_REPO_NAME:-$(git remote get-url "$UPSTREAM_REMOTE" | awk '{gsub(/http[s]:\/\/|git@/,"")}1' | awk -F'[@:./]' 'NR==1{print $4}')}
+
+if [[ -z ${GITHUB_USER:-} ]]; then
+  echo "Please export GITHUB_USER=<your-user> (or GH organization, if that's where your fork lives)"
+  exit 1
+fi
+
+if ! command -v gh > /dev/null; then
+  echo "Can't find 'gh' tool in PATH, please install from https://github.com/cli/cli"
+  exit 1
+fi
+
+if [[ "$#" -lt 2 ]]; then
+  echo "${0} <remote branch> <pr-number>...: cherry pick one or more <pr> onto <remote branch> and leave instructions for proposing pull request"
+  echo
+  echo "  Checks out <remote branch> and handles the cherry-pick of <pr> (possibly multiple) for you."
+  echo "  Examples:"
+  echo "    $0 upstream/release-3.14 12345        # Cherry-picks PR 12345 onto upstream/release-3.14 and proposes that as a PR."
+  echo "    $0 upstream/release-3.14 12345 56789  # Cherry-picks PR 12345, then 56789 and proposes the combination as a single PR."
+  echo
+  echo "  Set the DRY_RUN environment var to skip git push and creating PR."
+  echo "  This is useful for creating patches to a release branch without making a PR."
+  echo "  When DRY_RUN is set the script will leave you in a branch containing the commits you cherry-picked."
+  echo
+  echo "  Set the REGENERATE_DOCS environment var to regenerate documentation for the target branch after picking the specified commits."
+  echo "  This is useful when picking commits containing changes to API documentation."
+  echo
+  echo "  Set UPSTREAM_REMOTE (default: upstream) and FORK_REMOTE (default: origin)"
+  echo "  to override the default remote names to what you have locally."
+  echo
+  echo "  For merge process info, see https://git.k8s.io/community/contributors/devel/sig-release/cherry-picks.md"
+  exit 2
+fi
+
+# Checks if you are logged in. Will error/bail if you are not.
+gh auth status
+
+if git_status=$(git status --porcelain --untracked=no 2>/dev/null) && [[ -n "${git_status}" ]]; then
+  echo "!!! Dirty tree. Clean up and try again."
+  exit 1
+fi
+
+if [[ -e "${REBASEMAGIC}" ]]; then
+  echo "!!! 'git rebase' or 'git am' in progress. Clean up and try again."
+  exit 1
+fi
+
+declare -r BRANCH="$1"
+shift 1
+declare -r PULLS=( "$@" )
+
+function join { local IFS="$1"; shift; echo "$*"; }
+PULLDASH=$(join - "${PULLS[@]/#/#}") # Generates something like "#12345-#56789"
+declare -r PULLDASH
+PULLSUBJ=$(join " " "${PULLS[@]/#/#}") # Generates something like "#12345 #56789"
+declare -r PULLSUBJ
+
+echo "+++ Updating remotes..."
+git remote update "${UPSTREAM_REMOTE}" "${FORK_REMOTE}"
+
+if ! git log -n1 --format=%H "${BRANCH}" >/dev/null 2>&1; then
+  echo "!!! '${BRANCH}' not found. The second argument should be something like ${UPSTREAM_REMOTE}/release-0.21."
+  echo "    (In particular, it needs to be a valid, existing remote branch that I can 'git checkout'.)"
+  exit 1
+fi
+
+NEWBRANCHREQ="automated-cherry-pick-of-${PULLDASH}" # "Required" portion for tools.
+declare -r NEWBRANCHREQ
+NEWBRANCH="$(echo "${NEWBRANCHREQ}-${BRANCH}" | sed 's/\//-/g')"
+declare -r NEWBRANCH
+NEWBRANCHUNIQ="${NEWBRANCH}-$(date +%s)"
+declare -r NEWBRANCHUNIQ
+echo "+++ Creating local branch ${NEWBRANCHUNIQ}"
+
+cleanbranch=""
+gitamcleanup=false
+function return_to_kansas {
+  if [[ "${gitamcleanup}" == "true" ]]; then
+    echo
+    echo "+++ Aborting in-progress git am."
+    git am --abort >/dev/null 2>&1 || true
+  fi
+
+  # return to the starting branch and delete the PR text file
+  if [[ -z "${DRY_RUN}" ]]; then
+    echo
+    echo "+++ Returning you to the ${STARTINGBRANCH} branch and cleaning up."
+    git checkout -f "${STARTINGBRANCH}" >/dev/null 2>&1 || true
+    if [[ -n "${cleanbranch}" ]]; then
+      git branch -D "${cleanbranch}" >/dev/null 2>&1 || true
+    fi
+  fi
+}
+trap return_to_kansas EXIT
+
+SUBJECTS=()
+function make-a-pr() {
+  local rel
+  rel="$(basename "${BRANCH}")"
+  echo
+  echo "+++ Creating a pull request on GitHub at ${GITHUB_USER}:${NEWBRANCH}"
+
+  local numandtitle
+  numandtitle=$(printf '%s\n' "${SUBJECTS[@]}")
+  prtext=$(cat <<EOF
+Cherry pick of ${PULLSUBJ} on ${rel}.
+
+${numandtitle}
+
+For details on the cherry pick process, see the [cherry pick requests](https://git.k8s.io/community/contributors/devel/sig-release/cherry-picks.md) page.
+
+\`\`\`release-note
+
+\`\`\`
+EOF
+)
+
+  gh pr create --title="Automated cherry pick of ${numandtitle}" --body="${prtext}" --head "${GITHUB_USER}:${NEWBRANCH}" --base "${rel}" --repo="${MAIN_REPO_ORG}/${MAIN_REPO_NAME}"
+}
+
+git checkout -b "${NEWBRANCHUNIQ}" "${BRANCH}"
+cleanbranch="${NEWBRANCHUNIQ}"
+
+gitamcleanup=true
+for pull in "${PULLS[@]}"; do
+  echo "+++ Downloading patch to /tmp/${pull}.patch (in case you need to do this again)"
+
+  curl -o "/tmp/${pull}.patch" -sSL "https://github.com/${MAIN_REPO_ORG}/${MAIN_REPO_NAME}/pull/${pull}.patch"
+  echo
+  echo "+++ About to attempt cherry pick of PR. To reattempt:"
+  echo "  $ git am -3 /tmp/${pull}.patch"
+  echo
+  git am -3 "/tmp/${pull}.patch" || {
+    conflicts=false
+    while unmerged=$(git status --porcelain | grep ^U) && [[ -n ${unmerged} ]] \
+      || [[ -e "${REBASEMAGIC}" ]]; do
+      conflicts=true # <-- We should have detected conflicts once
+      echo
+      echo "+++ Conflicts detected:"
+      echo
+      (git status --porcelain | grep ^U) || echo "!!! None. Did you git am --continue?"
+      echo
+      echo "+++ Please resolve the conflicts in another window (and remember to 'git add / git am --continue')"
+      read -p "+++ Proceed (anything other than 'y' aborts the cherry-pick)? [y/n] " -r
+      echo
+      if ! [[ "${REPLY}" =~ ^[yY]$ ]]; then
+        echo "Aborting." >&2
+        exit 1
+      fi
+    done
+
+    if [[ "${conflicts}" != "true" ]]; then
+      echo "!!! git am failed, likely because of an in-progress 'git am' or 'git rebase'"
+      exit 1
+    fi
+  }
+
+  # set the subject
+  subject=$(grep -m 1 "^Subject" "/tmp/${pull}.patch" | sed -e 's/Subject: \[PATCH//g' | sed 's/.*] //')
+  SUBJECTS+=("#${pull}: ${subject}")
+
+  # remove the patch file from /tmp
+  rm -f "/tmp/${pull}.patch"
+done
+gitamcleanup=false
+
+# Re-generate docs (if needed)
+if [[ -n "${REGENERATE_DOCS}" ]]; then
+  echo
+  echo "Regenerating docs..."
+  if ! hack/generate-docs.sh; then
+    echo
+    echo "hack/generate-docs.sh FAILED to complete."
+    exit 1
+  fi
+fi
+
+if [[ -n "${DRY_RUN}" ]]; then
+  echo "!!! Skipping git push and PR creation because you set DRY_RUN."
+  echo "To return to the branch you were in when you invoked this script:"
+  echo
+  echo "  git checkout ${STARTINGBRANCH}"
+  echo
+  echo "To delete this branch:"
+  echo
+  echo "  git branch -D ${NEWBRANCHUNIQ}"
+  exit 0
+fi
+
+if git remote -v | grep ^"${FORK_REMOTE}" | grep "${MAIN_REPO_ORG}/${MAIN_REPO_NAME}.git"; then
+  echo "!!! You have ${FORK_REMOTE} configured as your ${MAIN_REPO_ORG}/${MAIN_REPO_NAME}.git"
+  echo "This isn't normal. Leaving you with push instructions:"
+  echo
+  echo "+++ First manually push the branch this script created:"
+  echo
+  echo "  git push REMOTE ${NEWBRANCHUNIQ}:${NEWBRANCH}"
+  echo
+  echo "where REMOTE is your personal fork (maybe ${UPSTREAM_REMOTE}? Consider swapping those.)."
+  echo "OR consider setting UPSTREAM_REMOTE and FORK_REMOTE to different values."
+  echo
+  make-a-pr
+  cleanbranch=""
+  exit 0
+fi
+
+echo
+echo "+++ I'm about to do the following to push to GitHub (and I'm assuming ${FORK_REMOTE} is your personal fork):"
+echo
+echo "  git push ${FORK_REMOTE} ${NEWBRANCHUNIQ}:${NEWBRANCH}"
+echo
+read -p "+++ Proceed (anything other than 'y' aborts the cherry-pick)? [y/n] " -r
+if ! [[ "${REPLY}" =~ ^[yY]$ ]]; then
+  echo "Aborting." >&2
+  exit 1
+fi
+
+git push "${FORK_REMOTE}" -f "${NEWBRANCHUNIQ}:${NEWBRANCH}"
+make-a-pr

hack/doc-gen/main.go

@@ -0,0 +1,20 @@
+package main
+
+import (
+	"log"
+	"os"
+
+	"github.com/spf13/cobra/doc"
+	"sigs.k8s.io/descheduler/cmd/descheduler/app"
+)
+
+var docGenPath = "docs/cli"
+
+func main() {
+	cmd := app.NewDeschedulerCommand(os.Stdout)
+	cmd.AddCommand(app.NewVersionCommand())
+	cmd.DisableAutoGenTag = true // Disable this so that the diff wont track it
+	if err := doc.GenMarkdownTree(cmd, docGenPath); err != nil {
+		log.Fatal(err)
+	}
+}

hack/lib/generator-help.sh

@@ -20,6 +20,6 @@ function find_dirs_containing_comment_tags() {
      | LC_ALL=C sort -u \
      )
 
-   IFS=",";
+   IFS=" ";
    printf '%s' "${array[*]}";
 }

hack/lib/go.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
 
-# Copyright 2021 The Kubernetes Authors.
+# Copyright 2024 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.
@@ -14,16 +14,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-set -o errexit
+# go::verify_version verifies the go version is supported by the project.
-set -o nounset
+# descheduler actively supports 3 versions, therefore 3 go versions are supported.
-set -o pipefail
+go::verify_version() {
+  GO_VERSION=($(go version))
 
-source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+  if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.21|go1.22|go1.23') ]]; then
-
+    echo "Unknown go version '${GO_VERSION[2]}', skipping gofmt."
-go build -o "${OS_OUTPUT_BINPATH}/mdtoc" "sigs.k8s.io/mdtoc"
+    exit 1
-
+  fi
-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

hack/update-docs.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
 
-# Copyright 2021 The Kubernetes Authors.
+# Copyright 2023 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.
@@ -18,8 +18,6 @@ set -o errexit
 set -o nounset
 set -o pipefail
 
-source "$(dirname "${BASH_SOURCE}")/lib/init.sh"
+SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
 
-go build -o "${OS_OUTPUT_BINPATH}/mdtoc" "sigs.k8s.io/mdtoc"
+go run ${SCRIPT_DIR}/doc-gen
-
-${OS_OUTPUT_BINPATH}/mdtoc --inplace README.md

hack/update-generated-conversions.sh

@@ -6,5 +6,5 @@ go build -o "${OS_OUTPUT_BINPATH}/conversion-gen" "k8s.io/code-generator/cmd/con
 
 ${OS_OUTPUT_BINPATH}/conversion-gen \
 		--go-header-file "hack/boilerplate/boilerplate.go.txt" \
-		--input-dirs "$(find_dirs_containing_comment_tags "+k8s:conversion-gen=")" \
+		--output-file zz_generated.conversion.go \
-		--output-file-base zz_generated.conversion
+		$(find_dirs_containing_comment_tags "+k8s:conversion-gen=")

hack/update-generated-deep-copies.sh

@@ -6,6 +6,6 @@ go build -o "${OS_OUTPUT_BINPATH}/deepcopy-gen" "k8s.io/code-generator/cmd/deepc
 
 ${OS_OUTPUT_BINPATH}/deepcopy-gen \
                 --go-header-file "hack/boilerplate/boilerplate.go.txt" \
-                --input-dirs "$(find_dirs_containing_comment_tags "+k8s:deepcopy-gen=")" \
+                --output-file zz_generated.deepcopy.go \
-                --output-file-base zz_generated.deepcopy
+                $(find_dirs_containing_comment_tags "+k8s:deepcopy-gen=")
 

hack/update-generated-defaulters.sh

@@ -6,6 +6,6 @@ go build -o "${OS_OUTPUT_BINPATH}/defaulter-gen" "k8s.io/code-generator/cmd/defa
 
 ${OS_OUTPUT_BINPATH}/defaulter-gen \
                 --go-header-file "hack/boilerplate/boilerplate.go.txt" \
-                --input-dirs "$(find_dirs_containing_comment_tags "+k8s:defaulter-gen=")" \
+                --extra-peer-dirs "${PRJ_PREFIX}/pkg/apis/componentconfig/v1alpha1,${PRJ_PREFIX}/pkg/api/v1alpha2" \
-		--extra-peer-dirs "${PRJ_PREFIX}/pkg/apis/componentconfig/v1alpha1,${PRJ_PREFIX}/pkg/api/v1alpha1" \
+                --output-file zz_generated.defaults.go \
-                --output-file-base zz_generated.defaults
+                $(find_dirs_containing_comment_tags "+k8s:defaulter-gen=")

hack/update-gofmt.sh

@@ -20,13 +20,9 @@ set -o nounset
 set -o pipefail
 
 DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
+source "${DESCHEDULER_ROOT}/hack/lib/go.sh"
 
-GO_VERSION=($(go version))
+go::verify_version
-
-if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.17|go1.18|go1.19') ]]; then
-  echo "Unknown go version '${GO_VERSION[2]}', skipping gofmt."
-  exit 1
-fi
 
 cd "${DESCHEDULER_ROOT}"
 

hack/verify-docs.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+# Copyright 2023 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
+
+SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
+
+temp_dir=$(mktemp -d)
+
+go run -ldflags "-X main.docGenPath=${temp_dir}" ${SCRIPT_DIR}/doc-gen
+
+if ! _out="$(diff -Naupr ${SCRIPT_DIR}/../docs/cli "${temp_dir}")"; then
+    echo "Generated output differs:" >&2
+    echo "${_out}" >&2
+    echo "Generated conversions verify failed. Please run ./hack/update-docs.sh"
+    rm -rf ${temp_dir}
+    exit 1
+fi
+
+rm -rf ${temp_dir}

hack/verify-gofmt.sh

@@ -20,13 +20,9 @@ set -o nounset
 set -o pipefail
 
 DESCHEDULER_ROOT=$(dirname "${BASH_SOURCE}")/..
+source "${DESCHEDULER_ROOT}/hack/lib/go.sh"
 
-GO_VERSION=($(go version))
+go::verify_version
-
-if [[ -z $(echo "${GO_VERSION[2]}" | grep -E 'go1.17|go1.18|go1.19') ]]; then
-  echo "Unknown go version '${GO_VERSION[2]}', skipping gofmt."
-  exit 1
-fi
 
 cd "${DESCHEDULER_ROOT}"
 

hack/verify-vendor.sh

@@ -70,7 +70,7 @@ pushd "${DESCHEDULER_ROOT}" > /dev/null 2>&1
     ret=1
   fi
 
-  if ! _out="$(diff -Naupr -x "BUILD" -x "AUTHORS*" -x "CONTRIBUTORS*" vendor "${_deschedulertmp}/vendor")"; then
+  if ! _out="$(diff -Naupr -x "BUILD" -x "AUTHORS*" -x "CONTRIBUTORS*" -x "README*" vendor "${_deschedulertmp}/vendor")"; then
     echo "Your vendored results are different:" >&2
     echo "${_out}" >&2
     echo "Vendor Verify failed." >&2

keps/1397-evictions-in-background/kep.yaml

@@ -0,0 +1,16 @@
+title: descheduler integration with evacuation API as an alternative to eviction API
+kep-number: 1397
+authors:
+  - "@ingvagabund"
+owning-sig: sig-scheduling
+participating-sigs:
+  - sig-apps
+status: provisional
+creation-date: 2024-04-14
+reviewers:
+  - atiratree
+approvers:
+  - TBD
+feature-gates:
+  - TBD
+stage: alpha

keps/753-descheduling-framework/kep.yaml

@@ -0,0 +1,29 @@
+title: Descheduling framework
+kep-number: 753
+authors:
+  - "@ingvagabund"
+  - "@damemi"
+owning-sig: sig-scheduling
+status: provisional
+creation-date: 2024-04-08
+reviewers:
+  - "@ingvagabund"
+  - "@damemi"
+  - "@a7i"
+  - "@knelasevero"
+approvers:
+  - "@ingvagabund"
+  - "@damemi"
+  - "@a7i"
+  - "@knelasevero"
+#replaces:
+
+# The target maturity stage in the current dev cycle for this KEP.
+stage: alpha
+
+# The most recent milestone for which work toward delivery of this KEP has been
+# done. This can be the current (upcoming) milestone, if it is being actively
+# worked on.
+# latest-milestone: "v1.19"
+
+# disable-supported: true

kubernetes/base/configmap.yaml

@@ -6,23 +6,29 @@ metadata:
   namespace: kube-system
 data:
   policy.yaml: |
-    apiVersion: "descheduler/v1alpha1"
+    apiVersion: "descheduler/v1alpha2"
     kind: "DeschedulerPolicy"
-    strategies:
+    profiles:
-      "RemoveDuplicates":
+      - name: ProfileName
-         enabled: true
+        pluginConfig:
-      "RemovePodsViolatingInterPodAntiAffinity":
+        - name: "DefaultEvictor"
-         enabled: true
+        - name: "RemovePodsViolatingInterPodAntiAffinity"
-      "LowNodeUtilization":
+        - name: "RemoveDuplicates"
-         enabled: true
+        - name: "LowNodeUtilization"
-         params:
+          args:
-           nodeResourceUtilizationThresholds:
+            thresholds:
-             thresholds:
+              "cpu" : 20
-               "cpu" : 20
+              "memory": 20
-               "memory": 20
+              "pods": 20
-               "pods": 20
+            targetThresholds:
-             targetThresholds:
+              "cpu" : 50
-               "cpu" : 50
+              "memory": 50
-               "memory": 50
+              "pods": 50
-               "pods": 50
+        plugins:
-
+          balance:
+            enabled:
+              - "LowNodeUtilization"
+              - "RemoveDuplicates"
+          deschedule:
+            enabled:
+              - "RemovePodsViolatingInterPodAntiAffinity"

kubernetes/base/rbac.yaml

@@ -22,13 +22,19 @@ rules:
 - apiGroups: ["scheduling.k8s.io"]
   resources: ["priorityclasses"]
   verbs: ["get", "watch", "list"]
+- apiGroups: ["policy"]
+  resources: ["poddisruptionbudgets"]
+  verbs: ["get", "watch", "list"]
 - apiGroups: ["coordination.k8s.io"]
   resources: ["leases"]
-  verbs: ["create"]
+  verbs: ["create", "update"]
 - apiGroups: ["coordination.k8s.io"]
   resources: ["leases"]
   resourceNames: ["descheduler"]
   verbs: ["get", "patch", "delete"]
+- apiGroups: ["metrics.k8s.io"]
+  resources: ["nodes", "pods"]
+  verbs: ["get", "list"]
 ---
 apiVersion: v1
 kind: ServiceAccount

kubernetes/cronjob/cronjob.yaml

@@ -16,7 +16,7 @@ spec:
           priorityClassName: system-cluster-critical
           containers:
           - name: descheduler
-            image: registry.k8s.io/descheduler/descheduler:v0.26.1
+            image: registry.k8s.io/descheduler/descheduler:v0.32.2
             volumeMounts:
             - mountPath: /policy-dir
               name: policy-volume

kubernetes/deployment/deployment.yaml

@@ -19,7 +19,7 @@ spec:
       serviceAccountName: descheduler-sa
       containers:
         - name: descheduler
-          image: registry.k8s.io/descheduler/descheduler:v0.26.1
+          image: registry.k8s.io/descheduler/descheduler:v0.32.2
           imagePullPolicy: IfNotPresent
           command:
             - "/bin/descheduler"

kubernetes/job/job.yaml

@@ -14,7 +14,7 @@ spec:
       priorityClassName: system-cluster-critical
       containers:
         - name: descheduler
-          image: registry.k8s.io/descheduler/descheduler:v0.26.1
+          image: registry.k8s.io/descheduler/descheduler:v0.32.2
           volumeMounts:
           - mountPath: /policy-dir
             name: policy-volume

metrics/metrics.go

@@ -36,7 +36,7 @@ var (
 			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"})
+		}, []string{"result", "strategy", "profile", "namespace", "node"})
 
 	buildInfo = metrics.NewGauge(
 		&metrics.GaugeOpts{
@@ -48,9 +48,29 @@ var (
 		},
 	)
 
+	DeschedulerLoopDuration = metrics.NewHistogramVec(
+		&metrics.HistogramOpts{
+			Subsystem:      DeschedulerSubsystem,
+			Name:           "descheduler_loop_duration_seconds",
+			Help:           "Time taken to complete a full descheduling cycle",
+			StabilityLevel: metrics.ALPHA,
+			Buckets:        []float64{0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 25, 50, 100, 250, 500},
+		}, []string{})
+
+	DeschedulerStrategyDuration = metrics.NewHistogramVec(
+		&metrics.HistogramOpts{
+			Subsystem:      DeschedulerSubsystem,
+			Name:           "descheduler_strategy_duration_seconds",
+			Help:           "Time taken to complete Each strategy of the descheduling operation",
+			StabilityLevel: metrics.ALPHA,
+			Buckets:        []float64{0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 25, 50, 100},
+		}, []string{"strategy", "profile"})
+
 	metricsList = []metrics.Registerable{
 		PodsEvicted,
 		buildInfo,
+		DeschedulerLoopDuration,
+		DeschedulerStrategyDuration,
 	}
 )
 

pkg/api/conversion.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2018 The Kubernetes Authors.
+Copyright 2023 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.
@@ -14,6 +14,4 @@ See the License for the specific language governing permissions and
 limitations under the License.
 */
 
-// +k8s:deepcopy-gen=package
+package api
-
-package config // import "k8s.io/apiserver/pkg/apis/config"

pkg/api/sort.go

@@ -15,7 +15,7 @@ package api
 
 import "sort"
 
-func SortProfilesByName(profiles []Profile) []Profile {
+func SortDeschedulerProfileByName(profiles []DeschedulerProfile) []DeschedulerProfile {
 	sort.Slice(profiles, func(i, j int) bool {
 		return profiles[i].Name < profiles[j].Name
 	})

pkg/api/types.go

@@ -28,7 +28,7 @@ type DeschedulerPolicy struct {
 	metav1.TypeMeta
 
 	// Profiles
-	Profiles []Profile
+	Profiles []DeschedulerProfile
 
 	// NodeSelector for a set of nodes to operate over
 	NodeSelector *string
@@ -38,13 +38,23 @@ type DeschedulerPolicy struct {
 
 	// MaxNoOfPodsToEvictPerNamespace restricts maximum of pods to be evicted per namespace.
 	MaxNoOfPodsToEvictPerNamespace *uint
+
+	// MaxNoOfPodsToTotal restricts maximum of pods to be evicted total.
+	MaxNoOfPodsToEvictTotal *uint
+
+	// EvictionFailureEventNotification should be set to true to enable eviction failure event notification.
+	// Default is false.
+	EvictionFailureEventNotification *bool
+
+	// MetricsCollector configures collection of metrics about actual resource utilization
+	MetricsCollector MetricsCollector
 }
 
 // Namespaces carries a list of included/excluded namespaces
 // for which a given strategy is applicable
 type Namespaces struct {
-	Include []string
+	Include []string `json:"include,omitempty"`
-	Exclude []string
+	Exclude []string `json:"exclude,omitempty"`
 }
 
 type (
@@ -53,11 +63,11 @@ type (
 )
 
 type PriorityThreshold struct {
-	Value *int32
+	Value *int32 `json:"value"`
-	Name  string
+	Name  string `json:"name"`
 }
 
-type Profile struct {
+type DeschedulerProfile struct {
 	Name          string
 	PluginConfigs []PluginConfig
 	Plugins       Plugins
@@ -73,7 +83,6 @@ type Plugins struct {
 	Sort              PluginSet
 	Deschedule        PluginSet
 	Balance           PluginSet
-	Evict             PluginSet
 	Filter            PluginSet
 	PreEvictionFilter PluginSet
 }
@@ -82,3 +91,10 @@ type PluginSet struct {
 	Enabled  []string
 	Disabled []string
 }
+
+// MetricsCollector configures collection of metrics about actual resource utilization
+type MetricsCollector struct {
+	// Enabled metrics collection from kubernetes metrics.
+	// Later, the collection can be extended to other providers.
+	Enabled bool
+}

pkg/api/v1alpha1/types.go

@@ -1,129 +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 v1alpha1
-
-import (
-	v1 "k8s.io/api/core/v1"
-	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-)
-
-// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
-
-type DeschedulerPolicy struct {
-	metav1.TypeMeta `json:",inline"`
-
-	// Strategies
-	Strategies StrategyList `json:"strategies,omitempty"`
-
-	// 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 *uint `json:"maxNoOfPodsToEvictPerNode,omitempty"`
-
-	// MaxNoOfPodsToEvictPerNamespace restricts maximum of pods to be evicted per namespace.
-	MaxNoOfPodsToEvictPerNamespace *uint `json:"maxNoOfPodsToEvictPerNamespace,omitempty"`
-}
-
-type (
-	StrategyName string
-	StrategyList map[StrategyName]DeschedulerStrategy
-)
-
-type DeschedulerStrategy struct {
-	// Enabled or disabled
-	Enabled bool `json:"enabled,omitempty"`
-
-	// Weight
-	Weight int `json:"weight,omitempty"`
-
-	// Strategy parameters
-	Params *StrategyParameters `json:"params,omitempty"`
-}
-
-// Namespaces carries a list of included/excluded namespaces
-// for which a given strategy is applicable.
-type Namespaces struct {
-	Include []string `json:"include"`
-	Exclude []string `json:"exclude"`
-}
-
-// Besides Namespaces ThresholdPriority and ThresholdPriorityClassName only one of its members may be specified
-type StrategyParameters struct {
-	NodeResourceUtilizationThresholds *NodeResourceUtilizationThresholds `json:"nodeResourceUtilizationThresholds,omitempty"`
-	NodeAffinityType                  []string                           `json:"nodeAffinityType,omitempty"`
-	PodsHavingTooManyRestarts         *PodsHavingTooManyRestarts         `json:"podsHavingTooManyRestarts,omitempty"`
-	PodLifeTime                       *PodLifeTime                       `json:"podLifeTime,omitempty"`
-	RemoveDuplicates                  *RemoveDuplicates                  `json:"removeDuplicates,omitempty"`
-	FailedPods                        *FailedPods                        `json:"failedPods,omitempty"`
-	IncludeSoftConstraints            bool                               `json:"includeSoftConstraints"`
-	Namespaces                        *Namespaces                        `json:"namespaces"`
-	ThresholdPriority                 *int32                             `json:"thresholdPriority"`
-	ThresholdPriorityClassName        string                             `json:"thresholdPriorityClassName"`
-	LabelSelector                     *metav1.LabelSelector              `json:"labelSelector"`
-	NodeFit                           bool                               `json:"nodeFit"`
-	IncludePreferNoSchedule           bool                               `json:"includePreferNoSchedule"`
-	ExcludedTaints                    []string                           `json:"excludedTaints,omitempty"`
-}
-
-type (
-	Percentage         float64
-	ResourceThresholds map[v1.ResourceName]Percentage
-)
-
-type NodeResourceUtilizationThresholds struct {
-	UseDeviationThresholds bool               `json:"useDeviationThresholds,omitempty"`
-	Thresholds             ResourceThresholds `json:"thresholds,omitempty"`
-	TargetThresholds       ResourceThresholds `json:"targetThresholds,omitempty"`
-	NumberOfNodes          int                `json:"numberOfNodes,omitempty"`
-}
-
-type PodsHavingTooManyRestarts struct {
-	PodRestartThreshold     int32 `json:"podRestartThreshold,omitempty"`
-	IncludingInitContainers bool  `json:"includingInitContainers,omitempty"`
-}
-
-type RemoveDuplicates struct {
-	ExcludeOwnerKinds []string `json:"excludeOwnerKinds,omitempty"`
-}
-
-type PodLifeTime struct {
-	MaxPodLifeTimeSeconds *uint    `json:"maxPodLifeTimeSeconds,omitempty"`
-	States                []string `json:"states,omitempty"`
-
-	// Deprecated: Use States instead.
-	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.deepcopy.go

@@ -1,380 +0,0 @@
-//go:build !ignore_autogenerated
-// +build !ignore_autogenerated
-
-/*
-Copyright 2023 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.
-*/
-
-// Code generated by deepcopy-gen. DO NOT EDIT.
-
-package v1alpha1
-
-import (
-	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	runtime "k8s.io/apimachinery/pkg/runtime"
-)
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
-	*out = *in
-	out.TypeMeta = in.TypeMeta
-	if in.Strategies != nil {
-		in, out := &in.Strategies, &out.Strategies
-		*out = make(StrategyList, len(*in))
-		for key, val := range *in {
-			(*out)[key] = *val.DeepCopy()
-		}
-	}
-	if in.NodeSelector != nil {
-		in, out := &in.NodeSelector, &out.NodeSelector
-		*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(uint)
-		**out = **in
-	}
-	if in.MaxNoOfPodsToEvictPerNamespace != nil {
-		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
-		*out = new(uint)
-		**out = **in
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeschedulerPolicy.
-func (in *DeschedulerPolicy) DeepCopy() *DeschedulerPolicy {
-	if in == nil {
-		return nil
-	}
-	out := new(DeschedulerPolicy)
-	in.DeepCopyInto(out)
-	return out
-}
-
-// DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.
-func (in *DeschedulerPolicy) DeepCopyObject() runtime.Object {
-	if c := in.DeepCopy(); c != nil {
-		return c
-	}
-	return nil
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *DeschedulerStrategy) DeepCopyInto(out *DeschedulerStrategy) {
-	*out = *in
-	if in.Params != nil {
-		in, out := &in.Params, &out.Params
-		*out = new(StrategyParameters)
-		(*in).DeepCopyInto(*out)
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeschedulerStrategy.
-func (in *DeschedulerStrategy) DeepCopy() *DeschedulerStrategy {
-	if in == nil {
-		return nil
-	}
-	out := new(DeschedulerStrategy)
-	in.DeepCopyInto(out)
-	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
-	if in.Include != nil {
-		in, out := &in.Include, &out.Include
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	if in.Exclude != nil {
-		in, out := &in.Exclude, &out.Exclude
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Namespaces.
-func (in *Namespaces) DeepCopy() *Namespaces {
-	if in == nil {
-		return nil
-	}
-	out := new(Namespaces)
-	in.DeepCopyInto(out)
-	return out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *NodeResourceUtilizationThresholds) DeepCopyInto(out *NodeResourceUtilizationThresholds) {
-	*out = *in
-	if in.Thresholds != nil {
-		in, out := &in.Thresholds, &out.Thresholds
-		*out = make(ResourceThresholds, len(*in))
-		for key, val := range *in {
-			(*out)[key] = val
-		}
-	}
-	if in.TargetThresholds != nil {
-		in, out := &in.TargetThresholds, &out.TargetThresholds
-		*out = make(ResourceThresholds, len(*in))
-		for key, val := range *in {
-			(*out)[key] = val
-		}
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new NodeResourceUtilizationThresholds.
-func (in *NodeResourceUtilizationThresholds) DeepCopy() *NodeResourceUtilizationThresholds {
-	if in == nil {
-		return nil
-	}
-	out := new(NodeResourceUtilizationThresholds)
-	in.DeepCopyInto(out)
-	return out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *PodLifeTime) DeepCopyInto(out *PodLifeTime) {
-	*out = *in
-	if in.MaxPodLifeTimeSeconds != nil {
-		in, out := &in.MaxPodLifeTimeSeconds, &out.MaxPodLifeTimeSeconds
-		*out = new(uint)
-		**out = **in
-	}
-	if in.States != nil {
-		in, out := &in.States, &out.States
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	if in.PodStatusPhases != nil {
-		in, out := &in.PodStatusPhases, &out.PodStatusPhases
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new PodLifeTime.
-func (in *PodLifeTime) DeepCopy() *PodLifeTime {
-	if in == nil {
-		return nil
-	}
-	out := new(PodLifeTime)
-	in.DeepCopyInto(out)
-	return out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *PodsHavingTooManyRestarts) DeepCopyInto(out *PodsHavingTooManyRestarts) {
-	*out = *in
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new PodsHavingTooManyRestarts.
-func (in *PodsHavingTooManyRestarts) DeepCopy() *PodsHavingTooManyRestarts {
-	if in == nil {
-		return nil
-	}
-	out := new(PodsHavingTooManyRestarts)
-	in.DeepCopyInto(out)
-	return out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *RemoveDuplicates) DeepCopyInto(out *RemoveDuplicates) {
-	*out = *in
-	if in.ExcludeOwnerKinds != nil {
-		in, out := &in.ExcludeOwnerKinds, &out.ExcludeOwnerKinds
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new RemoveDuplicates.
-func (in *RemoveDuplicates) DeepCopy() *RemoveDuplicates {
-	if in == nil {
-		return nil
-	}
-	out := new(RemoveDuplicates)
-	in.DeepCopyInto(out)
-	return out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in ResourceThresholds) DeepCopyInto(out *ResourceThresholds) {
-	{
-		in := &in
-		*out = make(ResourceThresholds, len(*in))
-		for key, val := range *in {
-			(*out)[key] = val
-		}
-		return
-	}
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new ResourceThresholds.
-func (in ResourceThresholds) DeepCopy() ResourceThresholds {
-	if in == nil {
-		return nil
-	}
-	out := new(ResourceThresholds)
-	in.DeepCopyInto(out)
-	return *out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in StrategyList) DeepCopyInto(out *StrategyList) {
-	{
-		in := &in
-		*out = make(StrategyList, len(*in))
-		for key, val := range *in {
-			(*out)[key] = *val.DeepCopy()
-		}
-		return
-	}
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new StrategyList.
-func (in StrategyList) DeepCopy() StrategyList {
-	if in == nil {
-		return nil
-	}
-	out := new(StrategyList)
-	in.DeepCopyInto(out)
-	return *out
-}
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *StrategyParameters) DeepCopyInto(out *StrategyParameters) {
-	*out = *in
-	if in.NodeResourceUtilizationThresholds != nil {
-		in, out := &in.NodeResourceUtilizationThresholds, &out.NodeResourceUtilizationThresholds
-		*out = new(NodeResourceUtilizationThresholds)
-		(*in).DeepCopyInto(*out)
-	}
-	if in.NodeAffinityType != nil {
-		in, out := &in.NodeAffinityType, &out.NodeAffinityType
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	if in.PodsHavingTooManyRestarts != nil {
-		in, out := &in.PodsHavingTooManyRestarts, &out.PodsHavingTooManyRestarts
-		*out = new(PodsHavingTooManyRestarts)
-		**out = **in
-	}
-	if in.PodLifeTime != nil {
-		in, out := &in.PodLifeTime, &out.PodLifeTime
-		*out = new(PodLifeTime)
-		(*in).DeepCopyInto(*out)
-	}
-	if in.RemoveDuplicates != nil {
-		in, out := &in.RemoveDuplicates, &out.RemoveDuplicates
-		*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)
-		(*in).DeepCopyInto(*out)
-	}
-	if in.ThresholdPriority != nil {
-		in, out := &in.ThresholdPriority, &out.ThresholdPriority
-		*out = new(int32)
-		**out = **in
-	}
-	if in.LabelSelector != nil {
-		in, out := &in.LabelSelector, &out.LabelSelector
-		*out = new(v1.LabelSelector)
-		(*in).DeepCopyInto(*out)
-	}
-	if in.ExcludedTaints != nil {
-		in, out := &in.ExcludedTaints, &out.ExcludedTaints
-		*out = make([]string, len(*in))
-		copy(*out, *in)
-	}
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new StrategyParameters.
-func (in *StrategyParameters) DeepCopy() *StrategyParameters {
-	if in == nil {
-		return nil
-	}
-	out := new(StrategyParameters)
-	in.DeepCopyInto(out)
-	return out
-}

pkg/api/v1alpha2/conversion.go

@@ -0,0 +1,136 @@
+/*
+Copyright 2023 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 v1alpha2
+
+import (
+	"fmt"
+	"sync"
+
+	"k8s.io/apimachinery/pkg/conversion"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/serializer"
+	utilruntime "k8s.io/apimachinery/pkg/util/runtime"
+	api "sigs.k8s.io/descheduler/pkg/api"
+
+	"sigs.k8s.io/descheduler/pkg/framework/pluginregistry"
+)
+
+var (
+	// pluginArgConversionScheme is a scheme with internal and v1alpha2 registered,
+	// used for defaulting/converting typed PluginConfig Args.
+	// Access via getPluginArgConversionScheme()
+	pluginArgConversionScheme     *runtime.Scheme
+	initPluginArgConversionScheme sync.Once
+
+	Scheme = runtime.NewScheme()
+	Codecs = serializer.NewCodecFactory(Scheme, serializer.EnableStrict)
+)
+
+func GetPluginArgConversionScheme() *runtime.Scheme {
+	initPluginArgConversionScheme.Do(func() {
+		// set up the scheme used for plugin arg conversion
+		pluginArgConversionScheme = runtime.NewScheme()
+		utilruntime.Must(AddToScheme(pluginArgConversionScheme))
+		utilruntime.Must(api.AddToScheme(pluginArgConversionScheme))
+	})
+	return pluginArgConversionScheme
+}
+
+func Convert_v1alpha2_DeschedulerPolicy_To_api_DeschedulerPolicy(in *DeschedulerPolicy, out *api.DeschedulerPolicy, s conversion.Scope) error {
+	if err := autoConvert_v1alpha2_DeschedulerPolicy_To_api_DeschedulerPolicy(in, out, s); err != nil {
+		return err
+	}
+	return convertToInternalPluginConfigArgs(out)
+}
+
+// convertToInternalPluginConfigArgs converts PluginConfig#Args into internal
+// types using a scheme, after applying defaults.
+func convertToInternalPluginConfigArgs(out *api.DeschedulerPolicy) error {
+	scheme := GetPluginArgConversionScheme()
+	for i := range out.Profiles {
+		prof := &out.Profiles[i]
+		for j := range prof.PluginConfigs {
+			args := prof.PluginConfigs[j].Args
+			if args == nil {
+				continue
+			}
+			if _, isUnknown := args.(*runtime.Unknown); isUnknown {
+				continue
+			}
+			internalArgs, err := scheme.ConvertToVersion(args, api.SchemeGroupVersion)
+			if err != nil {
+				err = nil
+				internalArgs = args
+				if err != nil {
+					return fmt.Errorf("converting .Profiles[%d].PluginConfigs[%d].Args into internal type: %w", i, j, err)
+				}
+			}
+			prof.PluginConfigs[j].Args = internalArgs
+		}
+	}
+	return nil
+}
+
+func Convert_v1alpha2_PluginConfig_To_api_PluginConfig(in *PluginConfig, out *api.PluginConfig, s conversion.Scope) error {
+	out.Name = in.Name
+	if _, ok := pluginregistry.PluginRegistry[in.Name]; ok {
+		out.Args = pluginregistry.PluginRegistry[in.Name].PluginArgInstance.DeepCopyObject()
+		if in.Args.Raw != nil {
+			_, _, err := Codecs.UniversalDecoder().Decode(in.Args.Raw, nil, out.Args)
+			if err != nil {
+				return err
+			}
+		} else if in.Args.Object != nil {
+			out.Args = in.Args.Object
+		}
+	} else {
+		if err := runtime.Convert_runtime_RawExtension_To_runtime_Object(&in.Args, &out.Args, s); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func Convert_api_DeschedulerPolicy_To_v1alpha2_DeschedulerPolicy(in *api.DeschedulerPolicy, out *DeschedulerPolicy, s conversion.Scope) error {
+	if err := autoConvert_api_DeschedulerPolicy_To_v1alpha2_DeschedulerPolicy(in, out, s); err != nil {
+		return err
+	}
+	return convertToExternalPluginConfigArgs(out)
+}
+
+// convertToExternalPluginConfigArgs converts PluginConfig#Args into
+// external (versioned) types using a scheme.
+func convertToExternalPluginConfigArgs(out *DeschedulerPolicy) error {
+	scheme := GetPluginArgConversionScheme()
+	for i := range out.Profiles {
+		for j := range out.Profiles[i].PluginConfigs {
+			args := out.Profiles[i].PluginConfigs[j].Args
+			if args.Object == nil {
+				continue
+			}
+			if _, isUnknown := args.Object.(*runtime.Unknown); isUnknown {
+				continue
+			}
+			externalArgs, err := scheme.ConvertToVersion(args.Object, SchemeGroupVersion)
+			if err != nil {
+				return err
+			}
+			out.Profiles[i].PluginConfigs[j].Args.Object = externalArgs
+		}
+	}
+	return nil
+}

pkg/api/v1alpha2/defaults.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2017 The Kubernetes Authors.
+Copyright 2023 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 v1alpha1
+package v1alpha2
 
 import "k8s.io/apimachinery/pkg/runtime"
 

pkg/api/v1alpha2/doc.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2017 The Kubernetes Authors.
+Copyright 2023 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.
@@ -15,9 +15,10 @@ limitations under the License.
 */
 
 // +k8s:deepcopy-gen=package,register
+// +k8s:conversion-gen=sigs.k8s.io/descheduler/pkg/api
 // +k8s:defaulter-gen=TypeMeta
 
-// Package v1alpha1 is the v1alpha1 version of the descheduler API
+// Package v1alpha2 is the v1alpha2 version of the descheduler API
 // +groupName=descheduler
 
-package v1alpha1 // import "sigs.k8s.io/descheduler/pkg/api/v1alpha1"
+package v1alpha2 // import "sigs.k8s.io/descheduler/pkg/api/v1alpha2"

pkg/api/v1alpha2/register.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2017 The Kubernetes Authors.
+Copyright 2023 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 v1alpha1
+package v1alpha2
 
 import (
 	"k8s.io/apimachinery/pkg/runtime"
@@ -30,7 +30,7 @@ var (
 // GroupName is the group name used in this package
 const (
 	GroupName    = "descheduler"
-	GroupVersion = "v1alpha1"
+	GroupVersion = "v1alpha2"
 )
 
 // SchemeGroupVersion is group version used to register these objects

pkg/api/v1alpha2/sort.go

@@ -0,0 +1,26 @@
+/*
+Copyright 2023 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 v1alpha2
+
+import "sort"
+
+func SortDeschedulerProfileByName(profiles []DeschedulerProfile) []DeschedulerProfile {
+	sort.Slice(profiles, func(i, j int) bool {
+		return profiles[i].Name < profiles[j].Name
+	})
+	return profiles
+}

pkg/api/v1alpha2/types.go

@@ -0,0 +1,82 @@
+/*
+Copyright 2023 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 v1alpha2
+
+import (
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+)
+
+// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
+
+type DeschedulerPolicy struct {
+	metav1.TypeMeta `json:",inline"`
+
+	// Profiles
+	Profiles []DeschedulerProfile `json:"profiles,omitempty"`
+
+	// NodeSelector for a set of nodes to operate over
+	NodeSelector *string `json:"nodeSelector,omitempty"`
+
+	// MaxNoOfPodsToEvictPerNode restricts maximum of pods to be evicted per node.
+	MaxNoOfPodsToEvictPerNode *uint `json:"maxNoOfPodsToEvictPerNode,omitempty"`
+
+	// MaxNoOfPodsToEvictPerNamespace restricts maximum of pods to be evicted per namespace.
+	MaxNoOfPodsToEvictPerNamespace *uint `json:"maxNoOfPodsToEvictPerNamespace,omitempty"`
+
+	// MaxNoOfPodsToTotal restricts maximum of pods to be evicted total.
+	MaxNoOfPodsToEvictTotal *uint `json:"maxNoOfPodsToEvictTotal,omitempty"`
+
+	// EvictionFailureEventNotification should be set to true to enable eviction failure event notification.
+	// Default is false.
+	EvictionFailureEventNotification *bool
+
+	// MetricsCollector configures collection of metrics for actual resource utilization
+	MetricsCollector MetricsCollector `json:"metricsCollector,omitempty"`
+}
+
+type DeschedulerProfile struct {
+	Name          string         `json:"name"`
+	PluginConfigs []PluginConfig `json:"pluginConfig"`
+	Plugins       Plugins        `json:"plugins"`
+}
+
+type Plugins struct {
+	PreSort           PluginSet `json:"presort"`
+	Sort              PluginSet `json:"sort"`
+	Deschedule        PluginSet `json:"deschedule"`
+	Balance           PluginSet `json:"balance"`
+	Filter            PluginSet `json:"filter"`
+	PreEvictionFilter PluginSet `json:"preevictionfilter"`
+}
+
+type PluginConfig struct {
+	Name string               `json:"name"`
+	Args runtime.RawExtension `json:"args"`
+}
+
+type PluginSet struct {
+	Enabled  []string `json:"enabled"`
+	Disabled []string `json:"disabled"`
+}
+
+// MetricsCollector configures collection of metrics about actual resource utilization
+type MetricsCollector struct {
+	// Enabled metrics collection from kubernetes metrics.
+	// Later, the collection can be extended to other providers.
+	Enabled bool `json:"enabled,omitempty"`
+}

pkg/api/v1alpha2/zz_generated.conversion.go

@@ -0,0 +1,311 @@
+//go:build !ignore_autogenerated
+// +build !ignore_autogenerated
+
+/*
+Copyright 2025 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.
+*/
+
+// Code generated by conversion-gen. DO NOT EDIT.
+
+package v1alpha2
+
+import (
+	unsafe "unsafe"
+
+	conversion "k8s.io/apimachinery/pkg/conversion"
+	runtime "k8s.io/apimachinery/pkg/runtime"
+	api "sigs.k8s.io/descheduler/pkg/api"
+)
+
+func init() {
+	localSchemeBuilder.Register(RegisterConversions)
+}
+
+// RegisterConversions adds conversion functions to the given scheme.
+// Public to allow building arbitrary schemes.
+func RegisterConversions(s *runtime.Scheme) error {
+	if err := s.AddGeneratedConversionFunc((*DeschedulerProfile)(nil), (*api.DeschedulerProfile)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha2_DeschedulerProfile_To_api_DeschedulerProfile(a.(*DeschedulerProfile), b.(*api.DeschedulerProfile), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*api.DeschedulerProfile)(nil), (*DeschedulerProfile)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_DeschedulerProfile_To_v1alpha2_DeschedulerProfile(a.(*api.DeschedulerProfile), b.(*DeschedulerProfile), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*MetricsCollector)(nil), (*api.MetricsCollector)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha2_MetricsCollector_To_api_MetricsCollector(a.(*MetricsCollector), b.(*api.MetricsCollector), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*api.MetricsCollector)(nil), (*MetricsCollector)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_MetricsCollector_To_v1alpha2_MetricsCollector(a.(*api.MetricsCollector), b.(*MetricsCollector), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*api.PluginConfig)(nil), (*PluginConfig)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_PluginConfig_To_v1alpha2_PluginConfig(a.(*api.PluginConfig), b.(*PluginConfig), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*PluginSet)(nil), (*api.PluginSet)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha2_PluginSet_To_api_PluginSet(a.(*PluginSet), b.(*api.PluginSet), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*api.PluginSet)(nil), (*PluginSet)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_PluginSet_To_v1alpha2_PluginSet(a.(*api.PluginSet), b.(*PluginSet), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*Plugins)(nil), (*api.Plugins)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha2_Plugins_To_api_Plugins(a.(*Plugins), b.(*api.Plugins), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*api.Plugins)(nil), (*Plugins)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_Plugins_To_v1alpha2_Plugins(a.(*api.Plugins), b.(*Plugins), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddConversionFunc((*api.DeschedulerPolicy)(nil), (*DeschedulerPolicy)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_api_DeschedulerPolicy_To_v1alpha2_DeschedulerPolicy(a.(*api.DeschedulerPolicy), b.(*DeschedulerPolicy), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddConversionFunc((*DeschedulerPolicy)(nil), (*api.DeschedulerPolicy)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha2_DeschedulerPolicy_To_api_DeschedulerPolicy(a.(*DeschedulerPolicy), b.(*api.DeschedulerPolicy), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddConversionFunc((*PluginConfig)(nil), (*api.PluginConfig)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha2_PluginConfig_To_api_PluginConfig(a.(*PluginConfig), b.(*api.PluginConfig), scope)
+	}); err != nil {
+		return err
+	}
+	return nil
+}
+
+func autoConvert_v1alpha2_DeschedulerPolicy_To_api_DeschedulerPolicy(in *DeschedulerPolicy, out *api.DeschedulerPolicy, s conversion.Scope) error {
+	if in.Profiles != nil {
+		in, out := &in.Profiles, &out.Profiles
+		*out = make([]api.DeschedulerProfile, len(*in))
+		for i := range *in {
+			if err := Convert_v1alpha2_DeschedulerProfile_To_api_DeschedulerProfile(&(*in)[i], &(*out)[i], s); err != nil {
+				return err
+			}
+		}
+	} else {
+		out.Profiles = nil
+	}
+	out.NodeSelector = (*string)(unsafe.Pointer(in.NodeSelector))
+	out.MaxNoOfPodsToEvictPerNode = (*uint)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNode))
+	out.MaxNoOfPodsToEvictPerNamespace = (*uint)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNamespace))
+	out.MaxNoOfPodsToEvictTotal = (*uint)(unsafe.Pointer(in.MaxNoOfPodsToEvictTotal))
+	out.EvictionFailureEventNotification = (*bool)(unsafe.Pointer(in.EvictionFailureEventNotification))
+	if err := Convert_v1alpha2_MetricsCollector_To_api_MetricsCollector(&in.MetricsCollector, &out.MetricsCollector, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+func autoConvert_api_DeschedulerPolicy_To_v1alpha2_DeschedulerPolicy(in *api.DeschedulerPolicy, out *DeschedulerPolicy, s conversion.Scope) error {
+	if in.Profiles != nil {
+		in, out := &in.Profiles, &out.Profiles
+		*out = make([]DeschedulerProfile, len(*in))
+		for i := range *in {
+			if err := Convert_api_DeschedulerProfile_To_v1alpha2_DeschedulerProfile(&(*in)[i], &(*out)[i], s); err != nil {
+				return err
+			}
+		}
+	} else {
+		out.Profiles = nil
+	}
+	out.NodeSelector = (*string)(unsafe.Pointer(in.NodeSelector))
+	out.MaxNoOfPodsToEvictPerNode = (*uint)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNode))
+	out.MaxNoOfPodsToEvictPerNamespace = (*uint)(unsafe.Pointer(in.MaxNoOfPodsToEvictPerNamespace))
+	out.MaxNoOfPodsToEvictTotal = (*uint)(unsafe.Pointer(in.MaxNoOfPodsToEvictTotal))
+	out.EvictionFailureEventNotification = (*bool)(unsafe.Pointer(in.EvictionFailureEventNotification))
+	if err := Convert_api_MetricsCollector_To_v1alpha2_MetricsCollector(&in.MetricsCollector, &out.MetricsCollector, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+func autoConvert_v1alpha2_DeschedulerProfile_To_api_DeschedulerProfile(in *DeschedulerProfile, out *api.DeschedulerProfile, s conversion.Scope) error {
+	out.Name = in.Name
+	if in.PluginConfigs != nil {
+		in, out := &in.PluginConfigs, &out.PluginConfigs
+		*out = make([]api.PluginConfig, len(*in))
+		for i := range *in {
+			if err := Convert_v1alpha2_PluginConfig_To_api_PluginConfig(&(*in)[i], &(*out)[i], s); err != nil {
+				return err
+			}
+		}
+	} else {
+		out.PluginConfigs = nil
+	}
+	if err := Convert_v1alpha2_Plugins_To_api_Plugins(&in.Plugins, &out.Plugins, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Convert_v1alpha2_DeschedulerProfile_To_api_DeschedulerProfile is an autogenerated conversion function.
+func Convert_v1alpha2_DeschedulerProfile_To_api_DeschedulerProfile(in *DeschedulerProfile, out *api.DeschedulerProfile, s conversion.Scope) error {
+	return autoConvert_v1alpha2_DeschedulerProfile_To_api_DeschedulerProfile(in, out, s)
+}
+
+func autoConvert_api_DeschedulerProfile_To_v1alpha2_DeschedulerProfile(in *api.DeschedulerProfile, out *DeschedulerProfile, s conversion.Scope) error {
+	out.Name = in.Name
+	if in.PluginConfigs != nil {
+		in, out := &in.PluginConfigs, &out.PluginConfigs
+		*out = make([]PluginConfig, len(*in))
+		for i := range *in {
+			if err := Convert_api_PluginConfig_To_v1alpha2_PluginConfig(&(*in)[i], &(*out)[i], s); err != nil {
+				return err
+			}
+		}
+	} else {
+		out.PluginConfigs = nil
+	}
+	if err := Convert_api_Plugins_To_v1alpha2_Plugins(&in.Plugins, &out.Plugins, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Convert_api_DeschedulerProfile_To_v1alpha2_DeschedulerProfile is an autogenerated conversion function.
+func Convert_api_DeschedulerProfile_To_v1alpha2_DeschedulerProfile(in *api.DeschedulerProfile, out *DeschedulerProfile, s conversion.Scope) error {
+	return autoConvert_api_DeschedulerProfile_To_v1alpha2_DeschedulerProfile(in, out, s)
+}
+
+func autoConvert_v1alpha2_MetricsCollector_To_api_MetricsCollector(in *MetricsCollector, out *api.MetricsCollector, s conversion.Scope) error {
+	out.Enabled = in.Enabled
+	return nil
+}
+
+// Convert_v1alpha2_MetricsCollector_To_api_MetricsCollector is an autogenerated conversion function.
+func Convert_v1alpha2_MetricsCollector_To_api_MetricsCollector(in *MetricsCollector, out *api.MetricsCollector, s conversion.Scope) error {
+	return autoConvert_v1alpha2_MetricsCollector_To_api_MetricsCollector(in, out, s)
+}
+
+func autoConvert_api_MetricsCollector_To_v1alpha2_MetricsCollector(in *api.MetricsCollector, out *MetricsCollector, s conversion.Scope) error {
+	out.Enabled = in.Enabled
+	return nil
+}
+
+// Convert_api_MetricsCollector_To_v1alpha2_MetricsCollector is an autogenerated conversion function.
+func Convert_api_MetricsCollector_To_v1alpha2_MetricsCollector(in *api.MetricsCollector, out *MetricsCollector, s conversion.Scope) error {
+	return autoConvert_api_MetricsCollector_To_v1alpha2_MetricsCollector(in, out, s)
+}
+
+func autoConvert_v1alpha2_PluginConfig_To_api_PluginConfig(in *PluginConfig, out *api.PluginConfig, s conversion.Scope) error {
+	out.Name = in.Name
+	if err := runtime.Convert_runtime_RawExtension_To_runtime_Object(&in.Args, &out.Args, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+func autoConvert_api_PluginConfig_To_v1alpha2_PluginConfig(in *api.PluginConfig, out *PluginConfig, s conversion.Scope) error {
+	out.Name = in.Name
+	if err := runtime.Convert_runtime_Object_To_runtime_RawExtension(&in.Args, &out.Args, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Convert_api_PluginConfig_To_v1alpha2_PluginConfig is an autogenerated conversion function.
+func Convert_api_PluginConfig_To_v1alpha2_PluginConfig(in *api.PluginConfig, out *PluginConfig, s conversion.Scope) error {
+	return autoConvert_api_PluginConfig_To_v1alpha2_PluginConfig(in, out, s)
+}
+
+func autoConvert_v1alpha2_PluginSet_To_api_PluginSet(in *PluginSet, out *api.PluginSet, s conversion.Scope) error {
+	out.Enabled = *(*[]string)(unsafe.Pointer(&in.Enabled))
+	out.Disabled = *(*[]string)(unsafe.Pointer(&in.Disabled))
+	return nil
+}
+
+// Convert_v1alpha2_PluginSet_To_api_PluginSet is an autogenerated conversion function.
+func Convert_v1alpha2_PluginSet_To_api_PluginSet(in *PluginSet, out *api.PluginSet, s conversion.Scope) error {
+	return autoConvert_v1alpha2_PluginSet_To_api_PluginSet(in, out, s)
+}
+
+func autoConvert_api_PluginSet_To_v1alpha2_PluginSet(in *api.PluginSet, out *PluginSet, s conversion.Scope) error {
+	out.Enabled = *(*[]string)(unsafe.Pointer(&in.Enabled))
+	out.Disabled = *(*[]string)(unsafe.Pointer(&in.Disabled))
+	return nil
+}
+
+// Convert_api_PluginSet_To_v1alpha2_PluginSet is an autogenerated conversion function.
+func Convert_api_PluginSet_To_v1alpha2_PluginSet(in *api.PluginSet, out *PluginSet, s conversion.Scope) error {
+	return autoConvert_api_PluginSet_To_v1alpha2_PluginSet(in, out, s)
+}
+
+func autoConvert_v1alpha2_Plugins_To_api_Plugins(in *Plugins, out *api.Plugins, s conversion.Scope) error {
+	if err := Convert_v1alpha2_PluginSet_To_api_PluginSet(&in.PreSort, &out.PreSort, s); err != nil {
+		return err
+	}
+	if err := Convert_v1alpha2_PluginSet_To_api_PluginSet(&in.Sort, &out.Sort, s); err != nil {
+		return err
+	}
+	if err := Convert_v1alpha2_PluginSet_To_api_PluginSet(&in.Deschedule, &out.Deschedule, s); err != nil {
+		return err
+	}
+	if err := Convert_v1alpha2_PluginSet_To_api_PluginSet(&in.Balance, &out.Balance, s); err != nil {
+		return err
+	}
+	if err := Convert_v1alpha2_PluginSet_To_api_PluginSet(&in.Filter, &out.Filter, s); err != nil {
+		return err
+	}
+	if err := Convert_v1alpha2_PluginSet_To_api_PluginSet(&in.PreEvictionFilter, &out.PreEvictionFilter, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Convert_v1alpha2_Plugins_To_api_Plugins is an autogenerated conversion function.
+func Convert_v1alpha2_Plugins_To_api_Plugins(in *Plugins, out *api.Plugins, s conversion.Scope) error {
+	return autoConvert_v1alpha2_Plugins_To_api_Plugins(in, out, s)
+}
+
+func autoConvert_api_Plugins_To_v1alpha2_Plugins(in *api.Plugins, out *Plugins, s conversion.Scope) error {
+	if err := Convert_api_PluginSet_To_v1alpha2_PluginSet(&in.PreSort, &out.PreSort, s); err != nil {
+		return err
+	}
+	if err := Convert_api_PluginSet_To_v1alpha2_PluginSet(&in.Sort, &out.Sort, s); err != nil {
+		return err
+	}
+	if err := Convert_api_PluginSet_To_v1alpha2_PluginSet(&in.Deschedule, &out.Deschedule, s); err != nil {
+		return err
+	}
+	if err := Convert_api_PluginSet_To_v1alpha2_PluginSet(&in.Balance, &out.Balance, s); err != nil {
+		return err
+	}
+	if err := Convert_api_PluginSet_To_v1alpha2_PluginSet(&in.Filter, &out.Filter, s); err != nil {
+		return err
+	}
+	if err := Convert_api_PluginSet_To_v1alpha2_PluginSet(&in.PreEvictionFilter, &out.PreEvictionFilter, s); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Convert_api_Plugins_To_v1alpha2_Plugins is an autogenerated conversion function.
+func Convert_api_Plugins_To_v1alpha2_Plugins(in *api.Plugins, out *Plugins, s conversion.Scope) error {
+	return autoConvert_api_Plugins_To_v1alpha2_Plugins(in, out, s)
+}

pkg/api/v1alpha2/zz_generated.deepcopy.go

@@ -0,0 +1,189 @@
+//go:build !ignore_autogenerated
+// +build !ignore_autogenerated
+
+/*
+Copyright 2025 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.
+*/
+
+// Code generated by deepcopy-gen. DO NOT EDIT.
+
+package v1alpha2
+
+import (
+	runtime "k8s.io/apimachinery/pkg/runtime"
+)
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
+	*out = *in
+	out.TypeMeta = in.TypeMeta
+	if in.Profiles != nil {
+		in, out := &in.Profiles, &out.Profiles
+		*out = make([]DeschedulerProfile, len(*in))
+		for i := range *in {
+			(*in)[i].DeepCopyInto(&(*out)[i])
+		}
+	}
+	if in.NodeSelector != nil {
+		in, out := &in.NodeSelector, &out.NodeSelector
+		*out = new(string)
+		**out = **in
+	}
+	if in.MaxNoOfPodsToEvictPerNode != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNode, &out.MaxNoOfPodsToEvictPerNode
+		*out = new(uint)
+		**out = **in
+	}
+	if in.MaxNoOfPodsToEvictPerNamespace != nil {
+		in, out := &in.MaxNoOfPodsToEvictPerNamespace, &out.MaxNoOfPodsToEvictPerNamespace
+		*out = new(uint)
+		**out = **in
+	}
+	if in.MaxNoOfPodsToEvictTotal != nil {
+		in, out := &in.MaxNoOfPodsToEvictTotal, &out.MaxNoOfPodsToEvictTotal
+		*out = new(uint)
+		**out = **in
+	}
+	if in.EvictionFailureEventNotification != nil {
+		in, out := &in.EvictionFailureEventNotification, &out.EvictionFailureEventNotification
+		*out = new(bool)
+		**out = **in
+	}
+	out.MetricsCollector = in.MetricsCollector
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeschedulerPolicy.
+func (in *DeschedulerPolicy) DeepCopy() *DeschedulerPolicy {
+	if in == nil {
+		return nil
+	}
+	out := new(DeschedulerPolicy)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.
+func (in *DeschedulerPolicy) DeepCopyObject() runtime.Object {
+	if c := in.DeepCopy(); c != nil {
+		return c
+	}
+	return nil
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *DeschedulerProfile) DeepCopyInto(out *DeschedulerProfile) {
+	*out = *in
+	if in.PluginConfigs != nil {
+		in, out := &in.PluginConfigs, &out.PluginConfigs
+		*out = make([]PluginConfig, len(*in))
+		for i := range *in {
+			(*in)[i].DeepCopyInto(&(*out)[i])
+		}
+	}
+	in.Plugins.DeepCopyInto(&out.Plugins)
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeschedulerProfile.
+func (in *DeschedulerProfile) DeepCopy() *DeschedulerProfile {
+	if in == nil {
+		return nil
+	}
+	out := new(DeschedulerProfile)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *MetricsCollector) DeepCopyInto(out *MetricsCollector) {
+	*out = *in
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new MetricsCollector.
+func (in *MetricsCollector) DeepCopy() *MetricsCollector {
+	if in == nil {
+		return nil
+	}
+	out := new(MetricsCollector)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *PluginConfig) DeepCopyInto(out *PluginConfig) {
+	*out = *in
+	in.Args.DeepCopyInto(&out.Args)
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new PluginConfig.
+func (in *PluginConfig) DeepCopy() *PluginConfig {
+	if in == nil {
+		return nil
+	}
+	out := new(PluginConfig)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *PluginSet) DeepCopyInto(out *PluginSet) {
+	*out = *in
+	if in.Enabled != nil {
+		in, out := &in.Enabled, &out.Enabled
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
+	if in.Disabled != nil {
+		in, out := &in.Disabled, &out.Disabled
+		*out = make([]string, len(*in))
+		copy(*out, *in)
+	}
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new PluginSet.
+func (in *PluginSet) DeepCopy() *PluginSet {
+	if in == nil {
+		return nil
+	}
+	out := new(PluginSet)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *Plugins) DeepCopyInto(out *Plugins) {
+	*out = *in
+	in.PreSort.DeepCopyInto(&out.PreSort)
+	in.Sort.DeepCopyInto(&out.Sort)
+	in.Deschedule.DeepCopyInto(&out.Deschedule)
+	in.Balance.DeepCopyInto(&out.Balance)
+	in.Filter.DeepCopyInto(&out.Filter)
+	in.PreEvictionFilter.DeepCopyInto(&out.PreEvictionFilter)
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Plugins.
+func (in *Plugins) DeepCopy() *Plugins {
+	if in == nil {
+		return nil
+	}
+	out := new(Plugins)
+	in.DeepCopyInto(out)
+	return out
+}

pkg/api/v1alpha2/zz_generated.defaults.go

@@ -2,7 +2,7 @@
 // +build !ignore_autogenerated
 
 /*
-Copyright The Kubernetes Authors.
+Copyright 2025 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.
@@ -19,7 +19,7 @@ limitations under the License.
 
 // Code generated by defaulter-gen. DO NOT EDIT.
 
-package v1
+package v1alpha2
 
 import (
 	runtime "k8s.io/apimachinery/pkg/runtime"
@@ -29,18 +29,5 @@ import (
 // Public to allow building arbitrary schemes.
 // All generated defaulters are covering - they call all nested defaulters.
 func RegisterDefaults(scheme *runtime.Scheme) error {
-	scheme.AddTypeDefaultingFunc(&EncryptionConfiguration{}, func(obj interface{}) { SetObjectDefaults_EncryptionConfiguration(obj.(*EncryptionConfiguration)) })
 	return nil
 }
-
-func SetObjectDefaults_EncryptionConfiguration(in *EncryptionConfiguration) {
-	for i := range in.Resources {
-		a := &in.Resources[i]
-		for j := range a.Providers {
-			b := &a.Providers[j]
-			if b.KMS != nil {
-				SetDefaults_KMSConfiguration(b.KMS)
-			}
-		}
-	}
-}

pkg/api/zz_generated.deepcopy.go

@@ -2,7 +2,7 @@
 // +build !ignore_autogenerated
 
 /*
-Copyright 2023 The Kubernetes Authors.
+Copyright 2025 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.
@@ -31,7 +31,7 @@ func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
 	out.TypeMeta = in.TypeMeta
 	if in.Profiles != nil {
 		in, out := &in.Profiles, &out.Profiles
-		*out = make([]Profile, len(*in))
+		*out = make([]DeschedulerProfile, len(*in))
 		for i := range *in {
 			(*in)[i].DeepCopyInto(&(*out)[i])
 		}
@@ -51,6 +51,17 @@ func (in *DeschedulerPolicy) DeepCopyInto(out *DeschedulerPolicy) {
 		*out = new(uint)
 		**out = **in
 	}
+	if in.MaxNoOfPodsToEvictTotal != nil {
+		in, out := &in.MaxNoOfPodsToEvictTotal, &out.MaxNoOfPodsToEvictTotal
+		*out = new(uint)
+		**out = **in
+	}
+	if in.EvictionFailureEventNotification != nil {
+		in, out := &in.EvictionFailureEventNotification, &out.EvictionFailureEventNotification
+		*out = new(bool)
+		**out = **in
+	}
+	out.MetricsCollector = in.MetricsCollector
 	return
 }
 
@@ -72,6 +83,46 @@ func (in *DeschedulerPolicy) DeepCopyObject() runtime.Object {
 	return nil
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *DeschedulerProfile) DeepCopyInto(out *DeschedulerProfile) {
+	*out = *in
+	if in.PluginConfigs != nil {
+		in, out := &in.PluginConfigs, &out.PluginConfigs
+		*out = make([]PluginConfig, len(*in))
+		for i := range *in {
+			(*in)[i].DeepCopyInto(&(*out)[i])
+		}
+	}
+	in.Plugins.DeepCopyInto(&out.Plugins)
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeschedulerProfile.
+func (in *DeschedulerProfile) DeepCopy() *DeschedulerProfile {
+	if in == nil {
+		return nil
+	}
+	out := new(DeschedulerProfile)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *MetricsCollector) DeepCopyInto(out *MetricsCollector) {
+	*out = *in
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new MetricsCollector.
+func (in *MetricsCollector) DeepCopy() *MetricsCollector {
+	if in == nil {
+		return nil
+	}
+	out := new(MetricsCollector)
+	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
@@ -150,7 +201,6 @@ func (in *Plugins) DeepCopyInto(out *Plugins) {
 	in.Sort.DeepCopyInto(&out.Sort)
 	in.Deschedule.DeepCopyInto(&out.Deschedule)
 	in.Balance.DeepCopyInto(&out.Balance)
-	in.Evict.DeepCopyInto(&out.Evict)
 	in.Filter.DeepCopyInto(&out.Filter)
 	in.PreEvictionFilter.DeepCopyInto(&out.PreEvictionFilter)
 	return
@@ -187,30 +237,6 @@ func (in *PriorityThreshold) DeepCopy() *PriorityThreshold {
 	return out
 }
 
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *Profile) DeepCopyInto(out *Profile) {
-	*out = *in
-	if in.PluginConfigs != nil {
-		in, out := &in.PluginConfigs, &out.PluginConfigs
-		*out = make([]PluginConfig, len(*in))
-		for i := range *in {
-			(*in)[i].DeepCopyInto(&(*out)[i])
-		}
-	}
-	in.Plugins.DeepCopyInto(&out.Plugins)
-	return
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Profile.
-func (in *Profile) DeepCopy() *Profile {
-	if in == nil {
-		return nil
-	}
-	out := new(Profile)
-	in.DeepCopyInto(out)
-	return out
-}
-
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in ResourceThresholds) DeepCopyInto(out *ResourceThresholds) {
 	{

pkg/apis/componentconfig/types.go

@@ -21,7 +21,6 @@ import (
 
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	componentbaseconfig "k8s.io/component-base/config"
-	registry "k8s.io/component-base/logs/api/v1"
 )
 
 // +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
@@ -34,6 +33,7 @@ type DeschedulerConfiguration struct {
 
 	// KubeconfigFile is path to kubeconfig file with authorization and master
 	// location information.
+	// Deprecated: Use clientConnection.kubeConfig instead.
 	KubeconfigFile string
 
 	// PolicyConfigFile is the filepath to the descheduler policy configuration.
@@ -51,13 +51,41 @@ type DeschedulerConfiguration struct {
 	// EvictLocalStoragePods allows pods using local storage to be evicted.
 	EvictLocalStoragePods bool
 
+	// EvictDaemonSetPods allows pods owned by a DaemonSet resource to be evicted.
+	EvictDaemonSetPods bool
+
 	// IgnorePVCPods sets whether PVC pods should be allowed to be evicted
 	IgnorePVCPods bool
 
+	// Tracing specifies the options for tracing.
+	Tracing TracingConfiguration
+
 	// LeaderElection starts Deployment using leader election loop
 	LeaderElection componentbaseconfig.LeaderElectionConfiguration
 
-	// Logging specifies the options of logging.
+	// ClientConnection specifies the kubeconfig file and client connection settings to use when communicating with the apiserver.
-	// Refer [Logs Options](https://github.com/kubernetes/component-base/blob/master/logs/api/v1/options.go) for more information.
+	// Refer to [ClientConnection](https://pkg.go.dev/k8s.io/kubernetes/pkg/apis/componentconfig#ClientConnectionConfiguration) for more information.
-	Logging registry.LoggingConfiguration
+	ClientConnection componentbaseconfig.ClientConnectionConfiguration
+}
+
+type TracingConfiguration struct {
+	// CollectorEndpoint is the address of the OpenTelemetry collector.
+	// If not specified, tracing will be used NoopTraceProvider.
+	CollectorEndpoint string
+	// TransportCert is the path to the certificate file for the OpenTelemetry collector.
+	// If not specified, provider will start in insecure mode.
+	TransportCert string
+	// ServiceName is the name of the service to be used in the OpenTelemetry collector.
+	// If not specified, the default value is "descheduler".
+	ServiceName string
+	// ServiceNamespace is the namespace of the service to be used in the OpenTelemetry collector.
+	// If not specified, tracing will be used default namespace.
+	ServiceNamespace string
+	// SampleRate is used to configure the sample rate of the OTEL trace collection. This value will
+	// be used as the Base value with sample ratio. A value >= 1.0 will sample everything and < 0 will
+	// not sample anything. Everything else is a percentage value.
+	SampleRate float64
+	// FallbackToNoOpProviderOnError can be set in case if you want your trace provider to fallback to
+	// no op provider in case if the configured end point based provider can't be setup.
+	FallbackToNoOpProviderOnError bool
 }

pkg/apis/componentconfig/v1alpha1/types.go

@@ -21,7 +21,6 @@ import (
 
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	componentbaseconfig "k8s.io/component-base/config"
-	registry "k8s.io/component-base/logs/api/v1"
 )
 
 // +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
@@ -34,6 +33,7 @@ type DeschedulerConfiguration struct {
 
 	// KubeconfigFile is path to kubeconfig file with authorization and master
 	// location information.
+	// Deprecated: Use clientConnection.kubeConfig instead.
 	KubeconfigFile string `json:"kubeconfigFile"`
 
 	// PolicyConfigFile is the filepath to the descheduler policy configuration.
@@ -51,13 +51,41 @@ type DeschedulerConfiguration struct {
 	// EvictLocalStoragePods allows pods using local storage to be evicted.
 	EvictLocalStoragePods bool `json:"evictLocalStoragePods,omitempty"`
 
+	// EvictDaemonSetPods allows pods owned by a DaemonSet resource to be evicted.
+	EvictDaemonSetPods bool `json:"evictDaemonSetPods,omitempty"`
+
 	// IgnorePVCPods sets whether PVC pods should be allowed to be evicted
 	IgnorePVCPods bool `json:"ignorePvcPods,omitempty"`
 
+	// Tracing is used to setup the required OTEL tracing configuration
+	Tracing TracingConfiguration `json:"tracing,omitempty"`
+
 	// LeaderElection starts Deployment using leader election loop
 	LeaderElection componentbaseconfig.LeaderElectionConfiguration `json:"leaderElection,omitempty"`
 
-	// Logging specifies the options of logging.
+	// ClientConnection specifies the kubeconfig file and client connection settings to use when communicating with the apiserver.
-	// Refer [Logs Options](https://github.com/kubernetes/component-base/blob/master/logs/api/v1/options.go) for more information.
+	// Refer to [ClientConnection](https://pkg.go.dev/k8s.io/kubernetes/pkg/apis/componentconfig#ClientConnectionConfiguration) for more information.
-	Logging registry.LoggingConfiguration `json:"logging,omitempty"`
+	ClientConnection componentbaseconfig.ClientConnectionConfiguration `json:"clientConnection,omitempty"`
+}
+
+type TracingConfiguration struct {
+	// CollectorEndpoint is the address of the OpenTelemetry collector.
+	// If not specified, tracing will be used NoopTraceProvider.
+	CollectorEndpoint string `json:"collectorEndpoint"`
+	// TransportCert is the path to the certificate file for the OpenTelemetry collector.
+	// If not specified, provider will start in insecure mode.
+	TransportCert string `json:"transportCert,omitempty"`
+	// ServiceName is the name of the service to be used in the OpenTelemetry collector.
+	// If not specified, the default value is "descheduler".
+	ServiceName string `json:"serviceName,omitempty"`
+	// ServiceNamespace is the namespace of the service to be used in the OpenTelemetry collector.
+	// If not specified, tracing will be used default namespace.
+	ServiceNamespace string `json:"serviceNamespace,omitempty"`
+	// SampleRate is used to configure the sample rate of the OTEL trace collection. This value will
+	// be used as the Base value with sample ratio. A value >= 1.0 will sample everything and < 0 will
+	// not sample anything. Everything else is a percentage value.
+	SampleRate float64 `json:"sampleRate"`
+	// FallbackToNoOpProviderOnError can be set in case if you want your trace provider to fallback to
+	// no op provider in case if the configured end point based provider can't be setup.
+	FallbackToNoOpProviderOnError bool `json:"fallbackToNoOpProviderOnError"`
 }

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

@@ -2,7 +2,7 @@
 // +build !ignore_autogenerated
 
 /*
-Copyright 2023 The Kubernetes Authors.
+Copyright 2025 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.
@@ -46,6 +46,16 @@ func RegisterConversions(s *runtime.Scheme) error {
 	}); err != nil {
 		return err
 	}
+	if err := s.AddGeneratedConversionFunc((*TracingConfiguration)(nil), (*componentconfig.TracingConfiguration)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_v1alpha1_TracingConfiguration_To_componentconfig_TracingConfiguration(a.(*TracingConfiguration), b.(*componentconfig.TracingConfiguration), scope)
+	}); err != nil {
+		return err
+	}
+	if err := s.AddGeneratedConversionFunc((*componentconfig.TracingConfiguration)(nil), (*TracingConfiguration)(nil), func(a, b interface{}, scope conversion.Scope) error {
+		return Convert_componentconfig_TracingConfiguration_To_v1alpha1_TracingConfiguration(a.(*componentconfig.TracingConfiguration), b.(*TracingConfiguration), scope)
+	}); err != nil {
+		return err
+	}
 	return nil
 }
 
@@ -57,9 +67,13 @@ func autoConvert_v1alpha1_DeschedulerConfiguration_To_componentconfig_Deschedule
 	out.NodeSelector = in.NodeSelector
 	out.MaxNoOfPodsToEvictPerNode = in.MaxNoOfPodsToEvictPerNode
 	out.EvictLocalStoragePods = in.EvictLocalStoragePods
+	out.EvictDaemonSetPods = in.EvictDaemonSetPods
 	out.IgnorePVCPods = in.IgnorePVCPods
+	if err := Convert_v1alpha1_TracingConfiguration_To_componentconfig_TracingConfiguration(&in.Tracing, &out.Tracing, s); err != nil {
+		return err
+	}
 	out.LeaderElection = in.LeaderElection
-	out.Logging = in.Logging
+	out.ClientConnection = in.ClientConnection
 	return nil
 }
 
@@ -76,9 +90,13 @@ func autoConvert_componentconfig_DeschedulerConfiguration_To_v1alpha1_Deschedule
 	out.NodeSelector = in.NodeSelector
 	out.MaxNoOfPodsToEvictPerNode = in.MaxNoOfPodsToEvictPerNode
 	out.EvictLocalStoragePods = in.EvictLocalStoragePods
+	out.EvictDaemonSetPods = in.EvictDaemonSetPods
 	out.IgnorePVCPods = in.IgnorePVCPods
+	if err := Convert_componentconfig_TracingConfiguration_To_v1alpha1_TracingConfiguration(&in.Tracing, &out.Tracing, s); err != nil {
+		return err
+	}
 	out.LeaderElection = in.LeaderElection
-	out.Logging = in.Logging
+	out.ClientConnection = in.ClientConnection
 	return nil
 }
 
@@ -86,3 +104,33 @@ func autoConvert_componentconfig_DeschedulerConfiguration_To_v1alpha1_Deschedule
 func Convert_componentconfig_DeschedulerConfiguration_To_v1alpha1_DeschedulerConfiguration(in *componentconfig.DeschedulerConfiguration, out *DeschedulerConfiguration, s conversion.Scope) error {
 	return autoConvert_componentconfig_DeschedulerConfiguration_To_v1alpha1_DeschedulerConfiguration(in, out, s)
 }
+
+func autoConvert_v1alpha1_TracingConfiguration_To_componentconfig_TracingConfiguration(in *TracingConfiguration, out *componentconfig.TracingConfiguration, s conversion.Scope) error {
+	out.CollectorEndpoint = in.CollectorEndpoint
+	out.TransportCert = in.TransportCert
+	out.ServiceName = in.ServiceName
+	out.ServiceNamespace = in.ServiceNamespace
+	out.SampleRate = in.SampleRate
+	out.FallbackToNoOpProviderOnError = in.FallbackToNoOpProviderOnError
+	return nil
+}
+
+// Convert_v1alpha1_TracingConfiguration_To_componentconfig_TracingConfiguration is an autogenerated conversion function.
+func Convert_v1alpha1_TracingConfiguration_To_componentconfig_TracingConfiguration(in *TracingConfiguration, out *componentconfig.TracingConfiguration, s conversion.Scope) error {
+	return autoConvert_v1alpha1_TracingConfiguration_To_componentconfig_TracingConfiguration(in, out, s)
+}
+
+func autoConvert_componentconfig_TracingConfiguration_To_v1alpha1_TracingConfiguration(in *componentconfig.TracingConfiguration, out *TracingConfiguration, s conversion.Scope) error {
+	out.CollectorEndpoint = in.CollectorEndpoint
+	out.TransportCert = in.TransportCert
+	out.ServiceName = in.ServiceName
+	out.ServiceNamespace = in.ServiceNamespace
+	out.SampleRate = in.SampleRate
+	out.FallbackToNoOpProviderOnError = in.FallbackToNoOpProviderOnError
+	return nil
+}
+
+// Convert_componentconfig_TracingConfiguration_To_v1alpha1_TracingConfiguration is an autogenerated conversion function.
+func Convert_componentconfig_TracingConfiguration_To_v1alpha1_TracingConfiguration(in *componentconfig.TracingConfiguration, out *TracingConfiguration, s conversion.Scope) error {
+	return autoConvert_componentconfig_TracingConfiguration_To_v1alpha1_TracingConfiguration(in, out, s)
+}

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

@@ -2,7 +2,7 @@
 // +build !ignore_autogenerated
 
 /*
-Copyright 2023 The Kubernetes Authors.
+Copyright 2025 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.
@@ -29,8 +29,9 @@ import (
 func (in *DeschedulerConfiguration) DeepCopyInto(out *DeschedulerConfiguration) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
+	out.Tracing = in.Tracing
 	out.LeaderElection = in.LeaderElection
-	in.Logging.DeepCopyInto(&out.Logging)
+	out.ClientConnection = in.ClientConnection
 	return
 }
 
@@ -51,3 +52,19 @@ func (in *DeschedulerConfiguration) DeepCopyObject() runtime.Object {
 	}
 	return nil
 }
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *TracingConfiguration) DeepCopyInto(out *TracingConfiguration) {
+	*out = *in
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new TracingConfiguration.
+func (in *TracingConfiguration) DeepCopy() *TracingConfiguration {
+	if in == nil {
+		return nil
+	}
+	out := new(TracingConfiguration)
+	in.DeepCopyInto(out)
+	return out
+}

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

@@ -2,7 +2,7 @@
 // +build !ignore_autogenerated
 
 /*
-Copyright 2023 The Kubernetes Authors.
+Copyright 2025 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

@@ -2,7 +2,7 @@
 // +build !ignore_autogenerated
 
 /*
-Copyright 2023 The Kubernetes Authors.
+Copyright 2025 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.
@@ -29,8 +29,9 @@ import (
 func (in *DeschedulerConfiguration) DeepCopyInto(out *DeschedulerConfiguration) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
+	out.Tracing = in.Tracing
 	out.LeaderElection = in.LeaderElection
-	in.Logging.DeepCopyInto(&out.Logging)
+	out.ClientConnection = in.ClientConnection
 	return
 }
 
@@ -51,3 +52,19 @@ func (in *DeschedulerConfiguration) DeepCopyObject() runtime.Object {
 	}
 	return nil
 }
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *TracingConfiguration) DeepCopyInto(out *TracingConfiguration) {
+	*out = *in
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new TracingConfiguration.
+func (in *TracingConfiguration) DeepCopy() *TracingConfiguration {
+	if in == nil {
+		return nil
+	}
+	out := new(TracingConfiguration)
+	in.DeepCopyInto(out)
+	return out
+}

pkg/descheduler/client/client.go

@@ -19,39 +19,63 @@ package client
 import (
 	"fmt"
 
-	clientset "k8s.io/client-go/kubernetes"
 	// Ensure to load all auth plugins.
+	clientset "k8s.io/client-go/kubernetes"
 	_ "k8s.io/client-go/plugin/pkg/client/auth"
 	"k8s.io/client-go/rest"
 	"k8s.io/client-go/tools/clientcmd"
+	componentbaseconfig "k8s.io/component-base/config"
+	metricsclient "k8s.io/metrics/pkg/client/clientset/versioned"
 )
 
-func CreateClient(kubeconfig, userAgt string) (clientset.Interface, error) {
+func createConfig(clientConnection componentbaseconfig.ClientConnectionConfiguration, userAgt string) (*rest.Config, error) {
 	var cfg *rest.Config
-	if len(kubeconfig) != 0 {
+	if len(clientConnection.Kubeconfig) != 0 {
-		master, err := GetMasterFromKubeconfig(kubeconfig)
+		master, err := GetMasterFromKubeconfig(clientConnection.Kubeconfig)
 		if err != nil {
-			return nil, fmt.Errorf("Failed to parse kubeconfig file: %v ", err)
+			return nil, fmt.Errorf("failed to parse kubeconfig file: %v ", err)
 		}
 
-		cfg, err = clientcmd.BuildConfigFromFlags(master, kubeconfig)
+		cfg, err = clientcmd.BuildConfigFromFlags(master, clientConnection.Kubeconfig)
 		if err != nil {
-			return nil, fmt.Errorf("Unable to build config: %v", err)
+			return nil, fmt.Errorf("unable to build config: %v", err)
 		}
 
 	} else {
 		var err error
 		cfg, err = rest.InClusterConfig()
 		if err != nil {
-			return nil, fmt.Errorf("Unable to build in cluster config: %v", err)
+			return nil, fmt.Errorf("unable to build in cluster config: %v", err)
 		}
 	}
 
+	cfg.Burst = int(clientConnection.Burst)
+	cfg.QPS = clientConnection.QPS
+
 	if len(userAgt) != 0 {
-		return clientset.NewForConfig(rest.AddUserAgent(cfg, userAgt))
+		cfg = rest.AddUserAgent(cfg, userAgt)
-	} else {
-		return clientset.NewForConfig(cfg)
 	}
+
+	return cfg, nil
+}
+
+func CreateClient(clientConnection componentbaseconfig.ClientConnectionConfiguration, userAgt string) (clientset.Interface, error) {
+	cfg, err := createConfig(clientConnection, userAgt)
+	if err != nil {
+		return nil, fmt.Errorf("unable to create config: %v", err)
+	}
+
+	return clientset.NewForConfig(cfg)
+}
+
+func CreateMetricsClient(clientConnection componentbaseconfig.ClientConnectionConfiguration, userAgt string) (metricsclient.Interface, error) {
+	cfg, err := createConfig(clientConnection, userAgt)
+	if err != nil {
+		return nil, fmt.Errorf("unable to create config: %v", err)
+	}
+
+	// Create the metrics clientset to access the metrics.k8s.io API
+	return metricsclient.NewForConfig(cfg)
 }
 
 func GetMasterFromKubeconfig(filename string) (string, error) {
@@ -62,11 +86,11 @@ func GetMasterFromKubeconfig(filename string) (string, error) {
 
 	context, ok := config.Contexts[config.CurrentContext]
 	if !ok {
-		return "", fmt.Errorf("Failed to get master address from kubeconfig")
+		return "", fmt.Errorf("failed to get master address from kubeconfig: current context not found")
 	}
 
 	if val, ok := config.Clusters[context.Cluster]; ok {
 		return val.Server, nil
 	}
-	return "", fmt.Errorf("Failed to get master address from kubeconfig")
+	return "", fmt.Errorf("failed to get master address from kubeconfig: cluster information not found")
 }

pkg/descheduler/descheduler.go

@@ -19,21 +19,31 @@ package descheduler
 import (
 	"context"
 	"fmt"
+	"math"
+	"strconv"
+	"time"
 
-	"k8s.io/klog/v2"
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/trace"
 
 	v1 "k8s.io/api/core/v1"
 	policy "k8s.io/api/policy/v1"
-	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	policyv1 "k8s.io/api/policy/v1"
+	schedulingv1 "k8s.io/api/scheduling/v1"
+	"k8s.io/apimachinery/pkg/api/meta"
 	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	utilversion "k8s.io/apimachinery/pkg/util/version"
 	"k8s.io/apimachinery/pkg/util/wait"
+	"k8s.io/client-go/discovery"
 	"k8s.io/client-go/informers"
 	clientset "k8s.io/client-go/kubernetes"
 	fakeclientset "k8s.io/client-go/kubernetes/fake"
-	listersv1 "k8s.io/client-go/listers/core/v1"
-	schedulingv1 "k8s.io/client-go/listers/scheduling/v1"
 	core "k8s.io/client-go/testing"
+	"k8s.io/client-go/tools/events"
+	componentbaseconfig "k8s.io/component-base/config"
+	"k8s.io/klog/v2"
 
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
 	"sigs.k8s.io/descheduler/metrics"
@@ -41,25 +51,269 @@ import (
 	"sigs.k8s.io/descheduler/pkg/descheduler/client"
 	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
 	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
+	"sigs.k8s.io/descheduler/pkg/descheduler/metricscollector"
 	nodeutil "sigs.k8s.io/descheduler/pkg/descheduler/node"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
-	"sigs.k8s.io/descheduler/pkg/framework"
+	"sigs.k8s.io/descheduler/pkg/framework/pluginregistry"
-	"sigs.k8s.io/descheduler/pkg/framework/plugins/defaultevictor"
+	frameworkprofile "sigs.k8s.io/descheduler/pkg/framework/profile"
-	"sigs.k8s.io/descheduler/pkg/framework/plugins/pluginbuilder"
+	frameworktypes "sigs.k8s.io/descheduler/pkg/framework/types"
+	"sigs.k8s.io/descheduler/pkg/tracing"
 	"sigs.k8s.io/descheduler/pkg/utils"
+	"sigs.k8s.io/descheduler/pkg/version"
 )
 
+type eprunner func(ctx context.Context, nodes []*v1.Node) *frameworktypes.Status
+
+type profileRunner struct {
+	name                      string
+	descheduleEPs, balanceEPs eprunner
+}
+
+type descheduler struct {
+	rs                     *options.DeschedulerServer
+	ir                     *informerResources
+	getPodsAssignedToNode  podutil.GetPodsAssignedToNodeFunc
+	sharedInformerFactory  informers.SharedInformerFactory
+	deschedulerPolicy      *api.DeschedulerPolicy
+	eventRecorder          events.EventRecorder
+	podEvictor             *evictions.PodEvictor
+	podEvictionReactionFnc func(*fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error)
+	metricsCollector       *metricscollector.MetricsCollector
+}
+
+type informerResources struct {
+	sharedInformerFactory informers.SharedInformerFactory
+	resourceToInformer    map[schema.GroupVersionResource]informers.GenericInformer
+}
+
+func newInformerResources(sharedInformerFactory informers.SharedInformerFactory) *informerResources {
+	return &informerResources{
+		sharedInformerFactory: sharedInformerFactory,
+		resourceToInformer:    make(map[schema.GroupVersionResource]informers.GenericInformer),
+	}
+}
+
+func (ir *informerResources) Uses(resources ...schema.GroupVersionResource) error {
+	for _, resource := range resources {
+		informer, err := ir.sharedInformerFactory.ForResource(resource)
+		if err != nil {
+			return err
+		}
+
+		ir.resourceToInformer[resource] = informer
+	}
+	return nil
+}
+
+// CopyTo Copy informer subscriptions to the new factory and objects to the fake client so that the backing caches are populated for when listers are used.
+func (ir *informerResources) CopyTo(fakeClient *fakeclientset.Clientset, newFactory informers.SharedInformerFactory) error {
+	for resource, informer := range ir.resourceToInformer {
+		_, err := newFactory.ForResource(resource)
+		if err != nil {
+			return fmt.Errorf("error getting resource %s: %w", resource, err)
+		}
+
+		objects, err := informer.Lister().List(labels.Everything())
+		if err != nil {
+			return fmt.Errorf("error listing %s: %w", informer, err)
+		}
+
+		for _, object := range objects {
+			fakeClient.Tracker().Add(object)
+		}
+	}
+	return nil
+}
+
+func newDescheduler(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string, eventRecorder events.EventRecorder, sharedInformerFactory informers.SharedInformerFactory,
+) (*descheduler, error) {
+	podInformer := sharedInformerFactory.Core().V1().Pods().Informer()
+
+	ir := newInformerResources(sharedInformerFactory)
+	ir.Uses(v1.SchemeGroupVersion.WithResource("pods"),
+		v1.SchemeGroupVersion.WithResource("nodes"),
+		// Future work could be to let each plugin declare what type of resources it needs; that way dry runs would stay
+		// consistent with the real runs without having to keep the list here in sync.
+		v1.SchemeGroupVersion.WithResource("namespaces"),                 // Used by the defaultevictor plugin
+		schedulingv1.SchemeGroupVersion.WithResource("priorityclasses"),  // Used by the defaultevictor plugin
+		policyv1.SchemeGroupVersion.WithResource("poddisruptionbudgets"), // Used by the defaultevictor plugin
+
+	) // Used by the defaultevictor plugin
+
+	getPodsAssignedToNode, err := podutil.BuildGetPodsAssignedToNodeFunc(podInformer)
+	if err != nil {
+		return nil, fmt.Errorf("build get pods assigned to node function error: %v", err)
+	}
+
+	podEvictor, err := evictions.NewPodEvictor(
+		ctx,
+		rs.Client,
+		eventRecorder,
+		podInformer,
+		rs.DefaultFeatureGates,
+		evictions.NewOptions().
+			WithPolicyGroupVersion(evictionPolicyGroupVersion).
+			WithMaxPodsToEvictPerNode(deschedulerPolicy.MaxNoOfPodsToEvictPerNode).
+			WithMaxPodsToEvictPerNamespace(deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace).
+			WithMaxPodsToEvictTotal(deschedulerPolicy.MaxNoOfPodsToEvictTotal).
+			WithEvictionFailureEventNotification(deschedulerPolicy.EvictionFailureEventNotification).
+			WithDryRun(rs.DryRun).
+			WithMetricsEnabled(!rs.DisableMetrics),
+	)
+	if err != nil {
+		return nil, err
+	}
+
+	var metricsCollector *metricscollector.MetricsCollector
+	if deschedulerPolicy.MetricsCollector.Enabled {
+		nodeSelector := labels.Everything()
+		if deschedulerPolicy.NodeSelector != nil {
+			sel, err := labels.Parse(*deschedulerPolicy.NodeSelector)
+			if err != nil {
+				return nil, err
+			}
+			nodeSelector = sel
+		}
+		metricsCollector = metricscollector.NewMetricsCollector(sharedInformerFactory.Core().V1().Nodes().Lister(), rs.MetricsClient, nodeSelector)
+	}
+
+	return &descheduler{
+		rs:                     rs,
+		ir:                     ir,
+		getPodsAssignedToNode:  getPodsAssignedToNode,
+		sharedInformerFactory:  sharedInformerFactory,
+		deschedulerPolicy:      deschedulerPolicy,
+		eventRecorder:          eventRecorder,
+		podEvictor:             podEvictor,
+		podEvictionReactionFnc: podEvictionReactionFnc,
+		metricsCollector:       metricsCollector,
+	}, nil
+}
+
+func (d *descheduler) runDeschedulerLoop(ctx context.Context, nodes []*v1.Node) error {
+	var span trace.Span
+	ctx, span = tracing.Tracer().Start(ctx, "runDeschedulerLoop")
+	defer span.End()
+	defer func(loopStartDuration time.Time) {
+		metrics.DeschedulerLoopDuration.With(map[string]string{}).Observe(time.Since(loopStartDuration).Seconds())
+	}(time.Now())
+
+	// if len is still <= 1 error out
+	if len(nodes) <= 1 {
+		klog.V(1).InfoS("The cluster size is 0 or 1 meaning eviction causes service disruption or degradation. So aborting..")
+		return fmt.Errorf("the cluster size is 0 or 1")
+	}
+
+	var client 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 d.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 := fakeclientset.NewSimpleClientset()
+		// simulate a pod eviction by deleting a pod
+		fakeClient.PrependReactor("create", "pods", d.podEvictionReactionFnc(fakeClient))
+		fakeSharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+
+		err := d.ir.CopyTo(fakeClient, fakeSharedInformerFactory)
+		if err != nil {
+			return err
+		}
+
+		// create a new instance of the shared informer factor from the cached client
+		// register the pod informer, otherwise it will not get running
+		d.getPodsAssignedToNode, err = podutil.BuildGetPodsAssignedToNodeFunc(fakeSharedInformerFactory.Core().V1().Pods().Informer())
+		if err != nil {
+			return fmt.Errorf("build get pods assigned to node function error: %v", err)
+		}
+
+		fakeCtx, cncl := context.WithCancel(context.TODO())
+		defer cncl()
+		fakeSharedInformerFactory.Start(fakeCtx.Done())
+		fakeSharedInformerFactory.WaitForCacheSync(fakeCtx.Done())
+
+		client = fakeClient
+		d.sharedInformerFactory = fakeSharedInformerFactory
+	} else {
+		client = d.rs.Client
+	}
+
+	klog.V(3).Infof("Setting up the pod evictor")
+	d.podEvictor.SetClient(client)
+	d.podEvictor.ResetCounters()
+
+	d.runProfiles(ctx, client, nodes)
+
+	klog.V(1).InfoS("Number of evictions/requests", "totalEvicted", d.podEvictor.TotalEvicted(), "evictionRequests", d.podEvictor.TotalEvictionRequests())
+
+	return nil
+}
+
+// runProfiles runs all the deschedule plugins of all profiles and
+// later runs through all balance plugins of all profiles. (All Balance plugins should come after all Deschedule plugins)
+// see https://github.com/kubernetes-sigs/descheduler/issues/979
+func (d *descheduler) runProfiles(ctx context.Context, client clientset.Interface, nodes []*v1.Node) {
+	var span trace.Span
+	ctx, span = tracing.Tracer().Start(ctx, "runProfiles")
+	defer span.End()
+	var profileRunners []profileRunner
+	for _, profile := range d.deschedulerPolicy.Profiles {
+		currProfile, err := frameworkprofile.NewProfile(
+			profile,
+			pluginregistry.PluginRegistry,
+			frameworkprofile.WithClientSet(client),
+			frameworkprofile.WithSharedInformerFactory(d.sharedInformerFactory),
+			frameworkprofile.WithPodEvictor(d.podEvictor),
+			frameworkprofile.WithGetPodsAssignedToNodeFnc(d.getPodsAssignedToNode),
+			frameworkprofile.WithMetricsCollector(d.metricsCollector),
+		)
+		if err != nil {
+			klog.ErrorS(err, "unable to create a profile", "profile", profile.Name)
+			continue
+		}
+		profileRunners = append(profileRunners, profileRunner{profile.Name, currProfile.RunDeschedulePlugins, currProfile.RunBalancePlugins})
+	}
+
+	for _, profileR := range profileRunners {
+		// First deschedule
+		status := profileR.descheduleEPs(ctx, nodes)
+		if status != nil && status.Err != nil {
+			span.AddEvent("failed to perform deschedule operations", trace.WithAttributes(attribute.String("err", status.Err.Error()), attribute.String("profile", profileR.name), attribute.String("operation", tracing.DescheduleOperation)))
+			klog.ErrorS(status.Err, "running deschedule extension point failed with error", "profile", profileR.name)
+			continue
+		}
+	}
+
+	for _, profileR := range profileRunners {
+		// Balance Later
+		status := profileR.balanceEPs(ctx, nodes)
+		if status != nil && status.Err != nil {
+			span.AddEvent("failed to perform balance operations", trace.WithAttributes(attribute.String("err", status.Err.Error()), attribute.String("profile", profileR.name), attribute.String("operation", tracing.BalanceOperation)))
+			klog.ErrorS(status.Err, "running balance extension point failed with error", "profile", profileR.name)
+			continue
+		}
+	}
+}
+
 func Run(ctx context.Context, rs *options.DeschedulerServer) error {
+	var span trace.Span
+	ctx, span = tracing.Tracer().Start(ctx, "Run")
+	defer span.End()
 	metrics.Register()
 
-	rsclient, eventClient, err := createClients(rs.KubeconfigFile)
+	clientConnection := rs.ClientConnection
+	if rs.KubeconfigFile != "" && clientConnection.Kubeconfig == "" {
+		clientConnection.Kubeconfig = rs.KubeconfigFile
+	}
+	rsclient, eventClient, err := createClients(clientConnection)
 	if err != nil {
 		return err
 	}
 	rs.Client = rsclient
 	rs.EventClient = eventClient
 
-	deschedulerPolicy, err := LoadPolicyConfig(rs.PolicyConfigFile, rs.Client, pluginbuilder.PluginRegistry)
+	deschedulerPolicy, err := LoadPolicyConfig(rs.PolicyConfigFile, rs.Client, pluginregistry.PluginRegistry)
 	if err != nil {
 		return err
 	}
@@ -67,25 +321,40 @@ func Run(ctx context.Context, rs *options.DeschedulerServer) error {
 		return fmt.Errorf("deschedulerPolicy is nil")
 	}
 
+	// Add k8s compatibility warnings to logs
+	if err := validateVersionCompatibility(rs.Client.Discovery(), version.Get()); err != nil {
+		klog.Warning(err.Error())
+	}
+
 	evictionPolicyGroupVersion, err := eutils.SupportEviction(rs.Client)
 	if err != nil || len(evictionPolicyGroupVersion) == 0 {
 		return err
 	}
 
+	if deschedulerPolicy.MetricsCollector.Enabled {
+		metricsClient, err := client.CreateMetricsClient(clientConnection, "descheduler")
+		if err != nil {
+			return err
+		}
+		rs.MetricsClient = metricsClient
+	}
+
 	runFn := func() error {
 		return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion)
 	}
 
 	if rs.LeaderElection.LeaderElect && rs.DeschedulingInterval.Seconds() == 0 {
+		span.AddEvent("Validation Failure", trace.WithAttributes(attribute.String("err", "leaderElection must be used with deschedulingInterval")))
 		return fmt.Errorf("leaderElection must be used with deschedulingInterval")
 	}
 
 	if rs.LeaderElection.LeaderElect && rs.DryRun {
-		klog.V(1).InfoS("Warning: DryRun is set to True. You need to disable it to use Leader Election.")
+		klog.V(1).Info("Warning: DryRun is set to True. You need to disable it to use Leader Election.")
 	}
 
 	if rs.LeaderElection.LeaderElect && !rs.DryRun {
 		if err := NewLeaderElection(runFn, rsclient, &rs.LeaderElection, ctx); err != nil {
+			span.AddEvent("Leader Election Failure", trace.WithAttributes(attribute.String("err", err.Error())))
 			return fmt.Errorf("leaderElection: %w", err)
 		}
 		return nil
@@ -94,16 +363,38 @@ func Run(ctx context.Context, rs *options.DeschedulerServer) error {
 	return runFn()
 }
 
-func cachedClient(
+func validateVersionCompatibility(discovery discovery.DiscoveryInterface, deschedulerVersionInfo version.Info) error {
-	realClient clientset.Interface,
+	kubeServerVersionInfo, err := discovery.ServerVersion()
-	podLister listersv1.PodLister,
+	if err != nil {
-	nodeLister listersv1.NodeLister,
+		return fmt.Errorf("failed to discover Kubernetes server version: %v", err)
-	namespaceLister listersv1.NamespaceLister,
+	}
-	priorityClassLister schedulingv1.PriorityClassLister,
+
-) (clientset.Interface, error) {
+	kubeServerVersion, err := utilversion.ParseSemantic(kubeServerVersionInfo.String())
-	fakeClient := fakeclientset.NewSimpleClientset()
+	if err != nil {
-	// simulate a pod eviction by deleting a pod
+		return fmt.Errorf("failed to parse Kubernetes server version '%s': %v", kubeServerVersionInfo.String(), err)
-	fakeClient.PrependReactor("create", "pods", func(action core.Action) (bool, runtime.Object, error) {
+	}
+
+	deschedulerMinor, err := strconv.ParseFloat(deschedulerVersionInfo.Minor, 64)
+	if err != nil {
+		return fmt.Errorf("failed to convert Descheduler minor version '%s' to float: %v", deschedulerVersionInfo.Minor, err)
+	}
+
+	kubeServerMinor := float64(kubeServerVersion.Minor())
+	if math.Abs(deschedulerMinor-kubeServerMinor) > 3 {
+		return fmt.Errorf(
+			"descheduler version %s.%s may not be supported on your version of Kubernetes %v."+
+				"See compatibility docs for more info: https://github.com/kubernetes-sigs/descheduler#compatibility-matrix",
+			deschedulerVersionInfo.Major,
+			deschedulerVersionInfo.Minor,
+			kubeServerVersionInfo.String(),
+		)
+	}
+
+	return nil
+}
+
+func podEvictionReactionFnc(fakeClient *fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error) {
+	return func(action core.Action) (bool, runtime.Object, error) {
 		if action.GetSubresource() == "eviction" {
 			createAct, matched := action.(core.CreateActionImpl)
 			if !matched {
@@ -120,132 +411,15 @@ func cachedClient(
 		}
 		// fallback to the default reactor
 		return false, nil, nil
-	})
-
-	klog.V(3).Infof("Pulling resources for the cached client from the cluster")
-	pods, err := podLister.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 := nodeLister.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 := namespaceLister.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 := priorityClassLister.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
-}
-
-// evictorImpl implements the Evictor interface so plugins
-// can evict a pod without importing a specific pod evictor
-type evictorImpl struct {
-	podEvictor    *evictions.PodEvictor
-	evictorFilter framework.EvictorPlugin
-}
-
-var _ framework.Evictor = &evictorImpl{}
-
-// Filter checks if a pod can be evicted
-func (ei *evictorImpl) Filter(pod *v1.Pod) bool {
-	return ei.evictorFilter.Filter(pod)
-}
-
-// PreEvictionFilter checks if pod can be evicted right before eviction
-func (ei *evictorImpl) PreEvictionFilter(pod *v1.Pod) bool {
-	return ei.evictorFilter.PreEvictionFilter(pod)
-}
-
-// Evict evicts a pod (no pre-check performed)
-func (ei *evictorImpl) Evict(ctx context.Context, pod *v1.Pod, opts evictions.EvictOptions) bool {
-	return ei.podEvictor.EvictPod(ctx, pod, opts)
-}
-
-func (ei *evictorImpl) NodeLimitExceeded(node *v1.Node) bool {
-	return ei.podEvictor.NodeLimitExceeded(node)
-}
-
-// handleImpl implements the framework handle which gets passed to plugins
-type handleImpl struct {
-	clientSet                 clientset.Interface
-	getPodsAssignedToNodeFunc podutil.GetPodsAssignedToNodeFunc
-	sharedInformerFactory     informers.SharedInformerFactory
-	evictor                   *evictorImpl
-}
-
-var _ framework.Handle = &handleImpl{}
-
-// ClientSet retrieves kube client set
-func (hi *handleImpl) ClientSet() clientset.Interface {
-	return hi.clientSet
-}
-
-// GetPodsAssignedToNodeFunc retrieves GetPodsAssignedToNodeFunc implementation
-func (hi *handleImpl) GetPodsAssignedToNodeFunc() podutil.GetPodsAssignedToNodeFunc {
-	return hi.getPodsAssignedToNodeFunc
-}
-
-// SharedInformerFactory retrieves shared informer factory
-func (hi *handleImpl) SharedInformerFactory() informers.SharedInformerFactory {
-	return hi.sharedInformerFactory
-}
-
-// Evictor retrieves evictor so plugins can filter and evict pods
-func (hi *handleImpl) Evictor() framework.Evictor {
-	return hi.evictor
 }
 
 func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string) error {
-	sharedInformerFactory := informers.NewSharedInformerFactory(rs.Client, 0)
+	var span trace.Span
-	podInformer := sharedInformerFactory.Core().V1().Pods().Informer()
+	ctx, span = tracing.Tracer().Start(ctx, "RunDeschedulerStrategies")
-	podLister := sharedInformerFactory.Core().V1().Pods().Lister()
+	defer span.End()
-	nodeLister := sharedInformerFactory.Core().V1().Nodes().Lister()
-	namespaceLister := sharedInformerFactory.Core().V1().Namespaces().Lister()
-	priorityClassLister := sharedInformerFactory.Scheduling().V1().PriorityClasses().Lister()
 
-	ctx, cancel := context.WithCancel(ctx)
+	sharedInformerFactory := informers.NewSharedInformerFactoryWithOptions(rs.Client, 0, informers.WithTransform(trimManagedFields))
-	defer cancel()
-
-	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())
 
 	var nodeSelector string
 	if deschedulerPolicy.NodeSelector != nil {
@@ -258,155 +432,54 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 	} else {
 		eventClient = rs.Client
 	}
-
 	eventBroadcaster, eventRecorder := utils.GetRecorderAndBroadcaster(ctx, eventClient)
 	defer eventBroadcaster.Shutdown()
 
+	descheduler, err := newDescheduler(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion, eventRecorder, sharedInformerFactory)
+	if err != nil {
+		span.AddEvent("Failed to create new descheduler", trace.WithAttributes(attribute.String("err", err.Error())))
+		return err
+	}
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	sharedInformerFactory.Start(ctx.Done())
+	sharedInformerFactory.WaitForCacheSync(ctx.Done())
+	descheduler.podEvictor.WaitForEventHandlersSync(ctx)
+
+	if deschedulerPolicy.MetricsCollector.Enabled {
+		go func() {
+			klog.V(2).Infof("Starting metrics collector")
+			descheduler.metricsCollector.Run(ctx)
+			klog.V(2).Infof("Stopped metrics collector")
+		}()
+		klog.V(2).Infof("Waiting for metrics collector to sync")
+		if err := wait.PollWithContext(ctx, time.Second, time.Minute, func(context.Context) (done bool, err error) {
+			return descheduler.metricsCollector.HasSynced(), nil
+		}); err != nil {
+			return fmt.Errorf("unable to wait for metrics collector to sync: %v", err)
+		}
+	}
+
 	wait.NonSlidingUntil(func() {
-		nodes, err := nodeutil.ReadyNodes(ctx, rs.Client, nodeLister, nodeSelector)
+		// A next context is created here intentionally to avoid nesting the spans via context.
+		sCtx, sSpan := tracing.Tracer().Start(ctx, "NonSlidingUntil")
+		defer sSpan.End()
+
+		nodes, err := nodeutil.ReadyNodes(sCtx, rs.Client, descheduler.sharedInformerFactory.Core().V1().Nodes().Lister(), nodeSelector)
 		if err != nil {
-			klog.V(1).InfoS("Unable to get ready nodes", "err", err)
+			sSpan.AddEvent("Failed to detect ready nodes", trace.WithAttributes(attribute.String("err", err.Error())))
+			klog.Error(err)
 			cancel()
 			return
 		}
-
+		err = descheduler.runDeschedulerLoop(sCtx, nodes)
-		if len(nodes) <= 1 {
+		if err != nil {
-			klog.V(1).InfoS("The cluster size is 0 or 1 meaning eviction causes service disruption or degradation. So aborting..")
+			sSpan.AddEvent("Failed to run descheduler loop", trace.WithAttributes(attribute.String("err", err.Error())))
+			klog.Error(err)
 			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, podLister, nodeLister, namespaceLister, priorityClassLister)
-			if err != nil {
-				klog.Error(err)
-				return
-			}
-
-			fakeSharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
-			getPodsAssignedToNode, err = podutil.BuildGetPodsAssignedToNodeFunc(fakeSharedInformerFactory.Core().V1().Pods().Informer())
-			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(
-			podEvictorClient,
-			evictionPolicyGroupVersion,
-			rs.DryRun,
-			deschedulerPolicy.MaxNoOfPodsToEvictPerNode,
-			deschedulerPolicy.MaxNoOfPodsToEvictPerNamespace,
-			nodes,
-			!rs.DisableMetrics,
-			eventRecorder,
-		)
-
-		var enabledDeschedulePlugins []framework.DeschedulePlugin
-		var enabledBalancePlugins []framework.BalancePlugin
-
-		// Build plugins
-		for _, profile := range deschedulerPolicy.Profiles {
-			pc := getPluginConfig(defaultevictor.PluginName, profile.PluginConfigs)
-			if pc == nil {
-				klog.ErrorS(fmt.Errorf("unable to get plugin config"), "skipping plugin", "plugin", defaultevictor.PluginName, "profile", profile.Name)
-				continue
-			}
-			evictorFilter, err := defaultevictor.New(
-				pc.Args,
-				&handleImpl{
-					clientSet:                 rs.Client,
-					getPodsAssignedToNodeFunc: getPodsAssignedToNode,
-					sharedInformerFactory:     sharedInformerFactory,
-				},
-			)
-			if err != nil {
-				klog.ErrorS(fmt.Errorf("unable to construct a plugin"), "skipping plugin", "plugin", defaultevictor.PluginName)
-				continue
-			}
-			handle := &handleImpl{
-				clientSet:                 rs.Client,
-				getPodsAssignedToNodeFunc: getPodsAssignedToNode,
-				sharedInformerFactory:     sharedInformerFactory,
-				evictor: &evictorImpl{
-					podEvictor:    podEvictor,
-					evictorFilter: evictorFilter.(framework.EvictorPlugin),
-				},
-			}
-			// Assuming only a list of enabled extension points.
-			// Later, when a default list of plugins and their extension points is established,
-			// compute the list of enabled extension points as (DefaultEnabled + Enabled - Disabled)
-			for _, plugin := range append(profile.Plugins.Deschedule.Enabled, profile.Plugins.Balance.Enabled...) {
-				pc := getPluginConfig(plugin, profile.PluginConfigs)
-				if pc == nil {
-					klog.ErrorS(fmt.Errorf("unable to get plugin config"), "skipping plugin", "plugin", plugin)
-					continue
-				}
-				registryPlugin, ok := pluginbuilder.PluginRegistry[plugin]
-				pgFnc := registryPlugin.PluginBuilder
-				if !ok {
-					klog.ErrorS(fmt.Errorf("unable to find plugin in the pluginsMap"), "skipping plugin", "plugin", plugin)
-				}
-				pg, err := pgFnc(pc.Args, handle)
-				if err != nil {
-					klog.ErrorS(err, "unable to initialize a plugin", "pluginName", plugin)
-				}
-				if pg != nil {
-					switch v := pg.(type) {
-					case framework.DeschedulePlugin:
-						enabledDeschedulePlugins = append(enabledDeschedulePlugins, v)
-					case framework.BalancePlugin:
-						enabledBalancePlugins = append(enabledBalancePlugins, v)
-					default:
-						klog.ErrorS(fmt.Errorf("unknown plugin extension point"), "skipping plugin", "plugin", plugin)
-					}
-				}
-			}
-		}
-
-		// Execute extension points
-		for _, pg := range enabledDeschedulePlugins {
-			// TODO: strategyName should be accessible from within the strategy using a framework
-			// handle or function which the Evictor has access to. For migration/in-progress framework
-			// work, we are currently passing this via context. To be removed
-			// (See discussion thread https://github.com/kubernetes-sigs/descheduler/pull/885#discussion_r919962292)
-			childCtx := context.WithValue(ctx, "strategyName", pg.Name())
-			status := pg.Deschedule(childCtx, nodes)
-			if status != nil && status.Err != nil {
-				klog.ErrorS(status.Err, "plugin finished with error", "pluginName", pg.Name())
-			}
-		}
-
-		for _, pg := range enabledBalancePlugins {
-			// TODO: strategyName should be accessible from within the strategy using a framework
-			// handle or function which the Evictor has access to. For migration/in-progress framework
-			// work, we are currently passing this via context. To be removed
-			// (See discussion thread https://github.com/kubernetes-sigs/descheduler/pull/885#discussion_r919962292)
-			childCtx := context.WithValue(ctx, "strategyName", pg.Name())
-			status := pg.Balance(childCtx, nodes)
-			if status != nil && status.Err != nil {
-				klog.ErrorS(status.Err, "plugin finished with error", "pluginName", pg.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 {
 			cancel()
@@ -416,25 +489,32 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 	return nil
 }
 
-func getPluginConfig(pluginName string, pluginConfigs []api.PluginConfig) *api.PluginConfig {
+func GetPluginConfig(pluginName string, pluginConfigs []api.PluginConfig) (*api.PluginConfig, int) {
-	for _, pluginConfig := range pluginConfigs {
+	for idx, pluginConfig := range pluginConfigs {
 		if pluginConfig.Name == pluginName {
-			return &pluginConfig
+			return &pluginConfig, idx
 		}
 	}
-	return nil
+	return nil, 0
 }
 
-func createClients(kubeconfig string) (clientset.Interface, clientset.Interface, error) {
+func createClients(clientConnection componentbaseconfig.ClientConnectionConfiguration) (clientset.Interface, clientset.Interface, error) {
-	kClient, err := client.CreateClient(kubeconfig, "descheduler")
+	kClient, err := client.CreateClient(clientConnection, "descheduler")
 	if err != nil {
 		return nil, nil, err
 	}
 
-	eventClient, err := client.CreateClient(kubeconfig, "")
+	eventClient, err := client.CreateClient(clientConnection, "")
 	if err != nil {
 		return nil, nil, err
 	}
 
 	return kClient, eventClient, nil
 }
+
+func trimManagedFields(obj interface{}) (interface{}, error) {
+	if accessor, err := meta.Accessor(obj); err == nil {
+		accessor.SetManagedFields(nil)
+	}
+	return obj, nil
+}

pkg/descheduler/descheduler_test.go

@@ -2,46 +2,180 @@ package descheduler
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"math/rand"
+	"net/http"
 	"testing"
 	"time"
 
 	v1 "k8s.io/api/core/v1"
 	policy "k8s.io/api/policy/v1"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	apiversion "k8s.io/apimachinery/pkg/version"
+	fakediscovery "k8s.io/client-go/discovery/fake"
+	"k8s.io/client-go/informers"
 	fakeclientset "k8s.io/client-go/kubernetes/fake"
 	core "k8s.io/client-go/testing"
+	"k8s.io/component-base/featuregate"
+	"k8s.io/klog/v2"
+	"k8s.io/metrics/pkg/apis/metrics/v1beta1"
+	metricsclient "k8s.io/metrics/pkg/client/clientset/versioned"
+	fakemetricsclient "k8s.io/metrics/pkg/client/clientset/versioned/fake"
+	utilptr "k8s.io/utils/ptr"
+
 	"sigs.k8s.io/descheduler/cmd/descheduler/app/options"
 	"sigs.k8s.io/descheduler/pkg/api"
-	"sigs.k8s.io/descheduler/pkg/api/v1alpha1"
+	"sigs.k8s.io/descheduler/pkg/descheduler/evictions"
-	"sigs.k8s.io/descheduler/pkg/framework/plugins/pluginbuilder"
+	"sigs.k8s.io/descheduler/pkg/features"
+	"sigs.k8s.io/descheduler/pkg/framework/pluginregistry"
+	"sigs.k8s.io/descheduler/pkg/framework/plugins/defaultevictor"
+	"sigs.k8s.io/descheduler/pkg/framework/plugins/nodeutilization"
 	"sigs.k8s.io/descheduler/pkg/framework/plugins/removeduplicates"
 	"sigs.k8s.io/descheduler/pkg/framework/plugins/removepodsviolatingnodetaints"
+	"sigs.k8s.io/descheduler/pkg/utils"
+	deschedulerversion "sigs.k8s.io/descheduler/pkg/version"
 	"sigs.k8s.io/descheduler/test"
 )
 
-func TestTaintsUpdated(t *testing.T) {
+var (
-	pluginbuilder.PluginRegistry = pluginbuilder.NewRegistry()
+	podEvictionError     = errors.New("PodEvictionError")
-	pluginbuilder.Register(removepodsviolatingnodetaints.PluginName, removepodsviolatingnodetaints.New, &removepodsviolatingnodetaints.RemovePodsViolatingNodeTaintsArgs{}, pluginbuilder.PluginRegistry)
+	tooManyRequestsError = &apierrors.StatusError{
-	ctx := context.Background()
+		ErrStatus: metav1.Status{
-	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+			Status:  metav1.StatusFailure,
-	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+			Code:    http.StatusTooManyRequests,
-
+			Reason:  metav1.StatusReasonTooManyRequests,
-	p1 := test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, nil)
+			Message: "admission webhook \"virt-launcher-eviction-interceptor.kubevirt.io\" denied the request: Eviction triggered evacuation of VMI",
-	p1.ObjectMeta.OwnerReferences = []metav1.OwnerReference{
+		},
-		{},
 	}
+	nodesgvr = schema.GroupVersionResource{Group: "metrics.k8s.io", Version: "v1beta1", Resource: "nodes"}
+	podsgvr  = schema.GroupVersionResource{Group: "metrics.k8s.io", Version: "v1beta1", Resource: "pods"}
+)
 
-	client := fakeclientset.NewSimpleClientset(n1, n2, p1)
+func initFeatureGates() featuregate.FeatureGate {
-	eventClient := fakeclientset.NewSimpleClientset(n1, n2, p1)
+	featureGates := featuregate.NewFeatureGate()
-	dp := &v1alpha1.DeschedulerPolicy{
+	featureGates.Add(map[featuregate.Feature]featuregate.FeatureSpec{
-		Strategies: v1alpha1.StrategyList{
+		features.EvictionsInBackground: {Default: false, PreRelease: featuregate.Alpha},
-			"RemovePodsViolatingNodeTaints": v1alpha1.DeschedulerStrategy{
+	})
-				Enabled: true,
+	return featureGates
+}
+
+func initPluginRegistry() {
+	pluginregistry.PluginRegistry = pluginregistry.NewRegistry()
+	pluginregistry.Register(removeduplicates.PluginName, removeduplicates.New, &removeduplicates.RemoveDuplicates{}, &removeduplicates.RemoveDuplicatesArgs{}, removeduplicates.ValidateRemoveDuplicatesArgs, removeduplicates.SetDefaults_RemoveDuplicatesArgs, pluginregistry.PluginRegistry)
+	pluginregistry.Register(defaultevictor.PluginName, defaultevictor.New, &defaultevictor.DefaultEvictor{}, &defaultevictor.DefaultEvictorArgs{}, defaultevictor.ValidateDefaultEvictorArgs, defaultevictor.SetDefaults_DefaultEvictorArgs, pluginregistry.PluginRegistry)
+	pluginregistry.Register(removepodsviolatingnodetaints.PluginName, removepodsviolatingnodetaints.New, &removepodsviolatingnodetaints.RemovePodsViolatingNodeTaints{}, &removepodsviolatingnodetaints.RemovePodsViolatingNodeTaintsArgs{}, removepodsviolatingnodetaints.ValidateRemovePodsViolatingNodeTaintsArgs, removepodsviolatingnodetaints.SetDefaults_RemovePodsViolatingNodeTaintsArgs, pluginregistry.PluginRegistry)
+	pluginregistry.Register(nodeutilization.LowNodeUtilizationPluginName, nodeutilization.NewLowNodeUtilization, &nodeutilization.LowNodeUtilization{}, &nodeutilization.LowNodeUtilizationArgs{}, nodeutilization.ValidateLowNodeUtilizationArgs, nodeutilization.SetDefaults_LowNodeUtilizationArgs, pluginregistry.PluginRegistry)
+}
+
+func removePodsViolatingNodeTaintsPolicy() *api.DeschedulerPolicy {
+	return &api.DeschedulerPolicy{
+		Profiles: []api.DeschedulerProfile{
+			{
+				Name: "Profile",
+				PluginConfigs: []api.PluginConfig{
+					{
+						Name: "RemovePodsViolatingNodeTaints",
+						Args: &removepodsviolatingnodetaints.RemovePodsViolatingNodeTaintsArgs{},
+					},
+					{
+						Name: "DefaultEvictor",
+						Args: &defaultevictor.DefaultEvictorArgs{},
+					},
+				},
+				Plugins: api.Plugins{
+					Filter: api.PluginSet{
+						Enabled: []string{
+							"DefaultEvictor",
+						},
+					},
+					Deschedule: api.PluginSet{
+						Enabled: []string{
+							"RemovePodsViolatingNodeTaints",
+						},
+					},
+				},
 			},
 		},
 	}
+}
+
+func removeDuplicatesPolicy() *api.DeschedulerPolicy {
+	return &api.DeschedulerPolicy{
+		Profiles: []api.DeschedulerProfile{
+			{
+				Name: "Profile",
+				PluginConfigs: []api.PluginConfig{
+					{
+						Name: "RemoveDuplicates",
+						Args: &removeduplicates.RemoveDuplicatesArgs{},
+					},
+					{
+						Name: "DefaultEvictor",
+						Args: &defaultevictor.DefaultEvictorArgs{},
+					},
+				},
+				Plugins: api.Plugins{
+					Filter: api.PluginSet{
+						Enabled: []string{
+							"DefaultEvictor",
+						},
+					},
+					Balance: api.PluginSet{
+						Enabled: []string{
+							"RemoveDuplicates",
+						},
+					},
+				},
+			},
+		},
+	}
+}
+
+func lowNodeUtilizationPolicy(thresholds, targetThresholds api.ResourceThresholds, metricsEnabled bool) *api.DeschedulerPolicy {
+	return &api.DeschedulerPolicy{
+		Profiles: []api.DeschedulerProfile{
+			{
+				Name: "Profile",
+				PluginConfigs: []api.PluginConfig{
+					{
+						Name: nodeutilization.LowNodeUtilizationPluginName,
+						Args: &nodeutilization.LowNodeUtilizationArgs{
+							Thresholds:       thresholds,
+							TargetThresholds: targetThresholds,
+							MetricsUtilization: nodeutilization.MetricsUtilization{
+								MetricsServer: metricsEnabled,
+							},
+						},
+					},
+					{
+						Name: defaultevictor.PluginName,
+						Args: &defaultevictor.DefaultEvictorArgs{},
+					},
+				},
+				Plugins: api.Plugins{
+					Filter: api.PluginSet{
+						Enabled: []string{
+							defaultevictor.PluginName,
+						},
+					},
+					Balance: api.PluginSet{
+						Enabled: []string{
+							nodeutilization.LowNodeUtilizationPluginName,
+						},
+					},
+				},
+			},
+		},
+	}
+}
+
+func initDescheduler(t *testing.T, ctx context.Context, featureGates featuregate.FeatureGate, internalDeschedulerPolicy *api.DeschedulerPolicy, metricsClient metricsclient.Interface, objects ...runtime.Object) (*options.DeschedulerServer, *descheduler, *fakeclientset.Clientset) {
+	client := fakeclientset.NewSimpleClientset(objects...)
+	eventClient := fakeclientset.NewSimpleClientset(objects...)
 
 	rs, err := options.NewDeschedulerServer()
 	if err != nil {
@@ -49,6 +183,44 @@ func TestTaintsUpdated(t *testing.T) {
 	}
 	rs.Client = client
 	rs.EventClient = eventClient
+	rs.DefaultFeatureGates = featureGates
+	rs.MetricsClient = metricsClient
+
+	sharedInformerFactory := informers.NewSharedInformerFactoryWithOptions(rs.Client, 0, informers.WithTransform(trimManagedFields))
+	eventBroadcaster, eventRecorder := utils.GetRecorderAndBroadcaster(ctx, client)
+
+	descheduler, err := newDescheduler(ctx, rs, internalDeschedulerPolicy, "v1", eventRecorder, sharedInformerFactory)
+	if err != nil {
+		eventBroadcaster.Shutdown()
+		t.Fatalf("Unable to create a descheduler instance: %v", err)
+	}
+
+	sharedInformerFactory.Start(ctx.Done())
+	sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+	return rs, descheduler, client
+}
+
+func TestTaintsUpdated(t *testing.T) {
+	initPluginRegistry()
+
+	ctx := context.Background()
+	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+
+	p1 := test.BuildTestPod(fmt.Sprintf("pod_1_%s", n1.Name), 200, 0, n1.Name, nil)
+	p1.ObjectMeta.OwnerReferences = test.GetReplicaSetOwnerRefList()
+
+	client := fakeclientset.NewSimpleClientset(n1, n2, p1)
+	eventClient := fakeclientset.NewSimpleClientset(n1, n2, p1)
+
+	rs, err := options.NewDeschedulerServer()
+	if err != nil {
+		t.Fatalf("Unable to initialize server: %v", err)
+	}
+	rs.Client = client
+	rs.EventClient = eventClient
+	rs.DefaultFeatureGates = initFeatureGates()
 
 	pods, err := client.CoreV1().Pods(p1.Namespace).List(ctx, metav1.ListOptions{})
 	if err != nil {
@@ -72,14 +244,9 @@ func TestTaintsUpdated(t *testing.T) {
 	}
 
 	var evictedPods []string
-	client.PrependReactor("create", "pods", podEvictionReactionFuc(&evictedPods))
+	client.PrependReactor("create", "pods", podEvictionReactionTestingFnc(&evictedPods, nil, nil))
 
-	internalDeschedulerPolicy, err := V1alpha1ToInternal(client, dp, pluginbuilder.PluginRegistry)
+	if err := RunDeschedulerStrategies(ctx, rs, removePodsViolatingNodeTaintsPolicy(), "v1"); err != nil {
-	if err != nil {
-		t.Fatalf("Unable to convert v1alpha1 to v1alpha2: %v", err)
-	}
-
-	if err := RunDeschedulerStrategies(ctx, rs, internalDeschedulerPolicy, "v1"); err != nil {
 		t.Fatalf("Unable to run descheduler strategies: %v", err)
 	}
 
@@ -89,8 +256,8 @@ func TestTaintsUpdated(t *testing.T) {
 }
 
 func TestDuplicate(t *testing.T) {
-	pluginbuilder.PluginRegistry = pluginbuilder.NewRegistry()
+	initPluginRegistry()
-	pluginbuilder.Register(removeduplicates.PluginName, removeduplicates.New, &removeduplicates.RemoveDuplicatesArgs{}, pluginbuilder.PluginRegistry)
+
 	ctx := context.Background()
 	node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
 	node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
@@ -109,13 +276,6 @@ func TestDuplicate(t *testing.T) {
 
 	client := fakeclientset.NewSimpleClientset(node1, node2, p1, p2, p3)
 	eventClient := fakeclientset.NewSimpleClientset(node1, node2, p1, p2, p3)
-	dp := &v1alpha1.DeschedulerPolicy{
-		Strategies: v1alpha1.StrategyList{
-			"RemoveDuplicates": v1alpha1.DeschedulerStrategy{
-				Enabled: true,
-			},
-		},
-	}
 
 	rs, err := options.NewDeschedulerServer()
 	if err != nil {
@@ -123,6 +283,7 @@ func TestDuplicate(t *testing.T) {
 	}
 	rs.Client = client
 	rs.EventClient = eventClient
+	rs.DefaultFeatureGates = initFeatureGates()
 
 	pods, err := client.CoreV1().Pods(p1.Namespace).List(ctx, metav1.ListOptions{})
 	if err != nil {
@@ -134,18 +295,14 @@ func TestDuplicate(t *testing.T) {
 	}
 
 	var evictedPods []string
-	client.PrependReactor("create", "pods", podEvictionReactionFuc(&evictedPods))
+	client.PrependReactor("create", "pods", podEvictionReactionTestingFnc(&evictedPods, nil, nil))
 
-	internalDeschedulerPolicy, err := V1alpha1ToInternal(client, dp, pluginbuilder.PluginRegistry)
+	if err := RunDeschedulerStrategies(ctx, rs, removeDuplicatesPolicy(), "v1"); err != nil {
-	if err != nil {
-		t.Fatalf("Unable to convert v1alpha1 to v1alpha2: %v", err)
-	}
-	if err := RunDeschedulerStrategies(ctx, rs, internalDeschedulerPolicy, "v1"); err != nil {
 		t.Fatalf("Unable to run descheduler strategies: %v", err)
 	}
 
 	if len(evictedPods) == 0 {
-		t.Fatalf("Unable to evict pod, node taint did not get propagated to descheduler strategies %v\n", err)
+		t.Fatalf("Unable to evict pods\n")
 	}
 }
 
@@ -156,7 +313,7 @@ func TestRootCancel(t *testing.T) {
 	client := fakeclientset.NewSimpleClientset(n1, n2)
 	eventClient := fakeclientset.NewSimpleClientset(n1, n2)
 	dp := &api.DeschedulerPolicy{
-		Profiles: []api.Profile{}, // no strategies needed for this test
+		Profiles: []api.DeschedulerProfile{}, // no strategies needed for this test
 	}
 
 	rs, err := options.NewDeschedulerServer()
@@ -166,6 +323,7 @@ func TestRootCancel(t *testing.T) {
 	rs.Client = client
 	rs.EventClient = eventClient
 	rs.DeschedulingInterval = 100 * time.Millisecond
+	rs.DefaultFeatureGates = initFeatureGates()
 	errChan := make(chan error, 1)
 	defer close(errChan)
 
@@ -191,7 +349,7 @@ func TestRootCancelWithNoInterval(t *testing.T) {
 	client := fakeclientset.NewSimpleClientset(n1, n2)
 	eventClient := fakeclientset.NewSimpleClientset(n1, n2)
 	dp := &api.DeschedulerPolicy{
-		Profiles: []api.Profile{}, // no strategies needed for this test
+		Profiles: []api.DeschedulerProfile{}, // no strategies needed for this test
 	}
 
 	rs, err := options.NewDeschedulerServer()
@@ -201,6 +359,7 @@ func TestRootCancelWithNoInterval(t *testing.T) {
 	rs.Client = client
 	rs.EventClient = eventClient
 	rs.DeschedulingInterval = 0
+	rs.DefaultFeatureGates = initFeatureGates()
 	errChan := make(chan error, 1)
 	defer close(errChan)
 
@@ -219,7 +378,67 @@ func TestRootCancelWithNoInterval(t *testing.T) {
 	}
 }
 
-func podEvictionReactionFuc(evictedPods *[]string) func(action core.Action) (bool, runtime.Object, error) {
+func TestValidateVersionCompatibility(t *testing.T) {
+	type testCase struct {
+		name               string
+		deschedulerVersion deschedulerversion.Info
+		serverVersion      string
+		expectError        bool
+	}
+	testCases := []testCase{
+		{
+			name:               "no error when descheduler minor equals to server minor",
+			deschedulerVersion: deschedulerversion.Info{Major: "0", Minor: "26"},
+			serverVersion:      "v1.26.1",
+			expectError:        false,
+		},
+		{
+			name:               "no error when descheduler minor is 3 behind server minor",
+			deschedulerVersion: deschedulerversion.Info{Major: "0", Minor: "23"},
+			serverVersion:      "v1.26.1",
+			expectError:        false,
+		},
+		{
+			name:               "no error when descheduler minor is 3 ahead of server minor",
+			deschedulerVersion: deschedulerversion.Info{Major: "0", Minor: "26"},
+			serverVersion:      "v1.26.1",
+			expectError:        false,
+		},
+		{
+			name:               "error when descheduler minor is 4 behind server minor",
+			deschedulerVersion: deschedulerversion.Info{Major: "0", Minor: "22"},
+			serverVersion:      "v1.26.1",
+			expectError:        true,
+		},
+		{
+			name:               "error when descheduler minor is 4 ahead of server minor",
+			deschedulerVersion: deschedulerversion.Info{Major: "0", Minor: "27"},
+			serverVersion:      "v1.23.1",
+			expectError:        true,
+		},
+		{
+			name:               "no error when using managed provider version",
+			deschedulerVersion: deschedulerversion.Info{Major: "0", Minor: "25"},
+			serverVersion:      "v1.25.12-eks-2d98532",
+			expectError:        false,
+		},
+	}
+	client := fakeclientset.NewSimpleClientset()
+	fakeDiscovery, _ := client.Discovery().(*fakediscovery.FakeDiscovery)
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			fakeDiscovery.FakedServerVersion = &apiversion.Info{GitVersion: tc.serverVersion}
+			err := validateVersionCompatibility(fakeDiscovery, tc.deschedulerVersion)
+
+			hasError := err != nil
+			if tc.expectError != hasError {
+				t.Error("unexpected version compatibility behavior")
+			}
+		})
+	}
+}
+
+func podEvictionReactionTestingFnc(evictedPods *[]string, isEvictionsInBackground func(podName string) bool, evictionErr error) func(action core.Action) (bool, runtime.Object, error) {
 	return func(action core.Action) (bool, runtime.Object, error) {
 		if action.GetSubresource() == "eviction" {
 			createAct, matched := action.(core.CreateActionImpl)
@@ -227,9 +446,420 @@ func podEvictionReactionFuc(evictedPods *[]string) func(action core.Action) (boo
 				return false, nil, fmt.Errorf("unable to convert action to core.CreateActionImpl")
 			}
 			if eviction, matched := createAct.Object.(*policy.Eviction); matched {
+				if isEvictionsInBackground != nil && isEvictionsInBackground(eviction.GetName()) {
+					return true, nil, tooManyRequestsError
+				}
+				if evictionErr != nil {
+					return true, nil, evictionErr
+				}
 				*evictedPods = append(*evictedPods, eviction.GetName())
+				return true, nil, nil
 			}
 		}
 		return false, nil, nil // fallback to the default reactor
 	}
 }
+
+func taintNodeNoSchedule(node *v1.Node) {
+	node.Spec.Taints = []v1.Taint{
+		{
+			Key:    "key",
+			Value:  "value",
+			Effect: v1.TaintEffectNoSchedule,
+		},
+	}
+}
+
+func TestPodEvictorReset(t *testing.T) {
+	initPluginRegistry()
+
+	ctx := context.Background()
+	node1 := test.BuildTestNode("n1", 2000, 3000, 10, taintNodeNoSchedule)
+	node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	nodes := []*v1.Node{node1, node2}
+
+	ownerRef1 := test.GetReplicaSetOwnerRefList()
+	updatePod := func(pod *v1.Pod) {
+		pod.Namespace = "dev"
+		pod.ObjectMeta.OwnerReferences = ownerRef1
+	}
+
+	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, updatePod)
+	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, updatePod)
+
+	internalDeschedulerPolicy := removePodsViolatingNodeTaintsPolicy()
+	ctxCancel, cancel := context.WithCancel(ctx)
+	rs, descheduler, client := initDescheduler(t, ctxCancel, initFeatureGates(), internalDeschedulerPolicy, nil, node1, node2, p1, p2)
+	defer cancel()
+
+	var evictedPods []string
+	client.PrependReactor("create", "pods", podEvictionReactionTestingFnc(&evictedPods, nil, nil))
+
+	var fakeEvictedPods []string
+	descheduler.podEvictionReactionFnc = func(*fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error) {
+		return podEvictionReactionTestingFnc(&fakeEvictedPods, nil, nil)
+	}
+
+	// a single pod eviction expected
+	klog.Infof("2 pod eviction expected per a descheduling cycle, 2 real evictions in total")
+	if err := descheduler.runDeschedulerLoop(ctx, nodes); err != nil {
+		t.Fatalf("Unable to run a descheduling loop: %v", err)
+	}
+	if descheduler.podEvictor.TotalEvicted() != 2 || len(evictedPods) != 2 || len(fakeEvictedPods) != 0 {
+		t.Fatalf("Expected (2,2,0) pods evicted, got (%v, %v, %v) instead", descheduler.podEvictor.TotalEvicted(), len(evictedPods), len(fakeEvictedPods))
+	}
+
+	// a single pod eviction expected
+	klog.Infof("2 pod eviction expected per a descheduling cycle, 4 real evictions in total")
+	if err := descheduler.runDeschedulerLoop(ctx, nodes); err != nil {
+		t.Fatalf("Unable to run a descheduling loop: %v", err)
+	}
+	if descheduler.podEvictor.TotalEvicted() != 2 || len(evictedPods) != 4 || len(fakeEvictedPods) != 0 {
+		t.Fatalf("Expected (2,4,0) pods evicted, got (%v, %v, %v) instead", descheduler.podEvictor.TotalEvicted(), len(evictedPods), len(fakeEvictedPods))
+	}
+
+	// check the fake client syncing and the right pods evicted
+	klog.Infof("Enabling the dry run mode")
+	rs.DryRun = true
+	evictedPods = []string{}
+
+	klog.Infof("2 pod eviction expected per a descheduling cycle, 2 fake evictions in total")
+	if err := descheduler.runDeschedulerLoop(ctx, nodes); err != nil {
+		t.Fatalf("Unable to run a descheduling loop: %v", err)
+	}
+	if descheduler.podEvictor.TotalEvicted() != 2 || len(evictedPods) != 0 || len(fakeEvictedPods) != 2 {
+		t.Fatalf("Expected (2,0,2) pods evicted, got (%v, %v, %v) instead", descheduler.podEvictor.TotalEvicted(), len(evictedPods), len(fakeEvictedPods))
+	}
+
+	klog.Infof("2 pod eviction expected per a descheduling cycle, 4 fake evictions in total")
+	if err := descheduler.runDeschedulerLoop(ctx, nodes); err != nil {
+		t.Fatalf("Unable to run a descheduling loop: %v", err)
+	}
+	if descheduler.podEvictor.TotalEvicted() != 2 || len(evictedPods) != 0 || len(fakeEvictedPods) != 4 {
+		t.Fatalf("Expected (2,0,4) pods evicted, got (%v, %v, %v) instead", descheduler.podEvictor.TotalEvicted(), len(evictedPods), len(fakeEvictedPods))
+	}
+}
+
+func checkTotals(t *testing.T, ctx context.Context, descheduler *descheduler, totalEvictionRequests, totalEvicted uint) {
+	if total := descheduler.podEvictor.TotalEvictionRequests(); total != totalEvictionRequests {
+		t.Fatalf("Expected %v total eviction requests, got %v instead", totalEvictionRequests, total)
+	}
+	if total := descheduler.podEvictor.TotalEvicted(); total != totalEvicted {
+		t.Fatalf("Expected %v total evictions, got %v instead", totalEvicted, total)
+	}
+	t.Logf("Total evictions: %v, total eviction requests: %v, total evictions and eviction requests: %v", totalEvicted, totalEvictionRequests, totalEvicted+totalEvictionRequests)
+}
+
+func runDeschedulingCycleAndCheckTotals(t *testing.T, ctx context.Context, nodes []*v1.Node, descheduler *descheduler, totalEvictionRequests, totalEvicted uint) {
+	err := descheduler.runDeschedulerLoop(ctx, nodes)
+	if err != nil {
+		t.Fatalf("Unable to run a descheduling loop: %v", err)
+	}
+	checkTotals(t, ctx, descheduler, totalEvictionRequests, totalEvicted)
+}
+
+func TestEvictionRequestsCache(t *testing.T) {
+	initPluginRegistry()
+
+	ctx := context.Background()
+	node1 := test.BuildTestNode("n1", 2000, 3000, 10, taintNodeNoSchedule)
+	node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	nodes := []*v1.Node{node1, node2}
+
+	ownerRef1 := test.GetReplicaSetOwnerRefList()
+	updatePod := func(pod *v1.Pod) {
+		pod.Namespace = "dev"
+		pod.ObjectMeta.OwnerReferences = ownerRef1
+		pod.Status.Phase = v1.PodRunning
+	}
+	updatePodWithEvictionInBackground := func(pod *v1.Pod) {
+		updatePod(pod)
+		pod.Annotations = map[string]string{
+			evictions.EvictionRequestAnnotationKey: "",
+		}
+	}
+
+	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, updatePodWithEvictionInBackground)
+	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, updatePodWithEvictionInBackground)
+	p3 := test.BuildTestPod("p3", 100, 0, node1.Name, updatePod)
+	p4 := test.BuildTestPod("p4", 100, 0, node1.Name, updatePod)
+	p5 := test.BuildTestPod("p5", 100, 0, node1.Name, updatePod)
+
+	internalDeschedulerPolicy := removePodsViolatingNodeTaintsPolicy()
+	ctxCancel, cancel := context.WithCancel(ctx)
+	featureGates := featuregate.NewFeatureGate()
+	featureGates.Add(map[featuregate.Feature]featuregate.FeatureSpec{
+		features.EvictionsInBackground: {Default: true, PreRelease: featuregate.Alpha},
+	})
+	_, descheduler, client := initDescheduler(t, ctxCancel, featureGates, internalDeschedulerPolicy, nil, node1, node2, p1, p2, p3, p4)
+	defer cancel()
+
+	var fakeEvictedPods []string
+	descheduler.podEvictionReactionFnc = func(*fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error) {
+		return podEvictionReactionTestingFnc(&fakeEvictedPods, nil, podEvictionError)
+	}
+
+	var evictedPods []string
+	client.PrependReactor("create", "pods", podEvictionReactionTestingFnc(&evictedPods, func(name string) bool { return name == "p1" || name == "p2" }, nil))
+
+	klog.Infof("2 evictions in background expected, 2 normal evictions")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 2, 2)
+
+	klog.Infof("Repeat the same as previously to confirm no more evictions in background are requested")
+	// No evicted pod is actually deleted on purpose so the test can run the descheduling cycle repeatedly
+	// without recreating the pods.
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 2, 2)
+
+	klog.Infof("Scenario: Eviction in background got initiated")
+	p2.Annotations[evictions.EvictionInProgressAnnotationKey] = ""
+	if _, err := client.CoreV1().Pods(p2.Namespace).Update(context.TODO(), p2, metav1.UpdateOptions{}); err != nil {
+		t.Fatalf("unable to update a pod: %v", err)
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	klog.Infof("Repeat the same as previously to confirm no more evictions in background are requested")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 2, 2)
+
+	klog.Infof("Scenario: Another eviction in background got initiated")
+	p1.Annotations[evictions.EvictionInProgressAnnotationKey] = ""
+	if _, err := client.CoreV1().Pods(p1.Namespace).Update(context.TODO(), p1, metav1.UpdateOptions{}); err != nil {
+		t.Fatalf("unable to update a pod: %v", err)
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	klog.Infof("Repeat the same as previously to confirm no more evictions in background are requested")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 2, 2)
+
+	klog.Infof("Scenario: Eviction in background completed")
+	if err := client.CoreV1().Pods(p1.Namespace).Delete(context.TODO(), p1.Name, metav1.DeleteOptions{}); err != nil {
+		t.Fatalf("unable to delete a pod: %v", err)
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	klog.Infof("Check the number of evictions in background decreased")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 1, 2)
+
+	klog.Infof("Scenario: A new pod without eviction in background added")
+	if _, err := client.CoreV1().Pods(p5.Namespace).Create(context.TODO(), p5, metav1.CreateOptions{}); err != nil {
+		t.Fatalf("unable to create a pod: %v", err)
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	klog.Infof("Check the number of evictions increased after running a descheduling cycle")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 1, 3)
+
+	klog.Infof("Scenario: Eviction in background canceled => eviction in progress annotation removed")
+	delete(p2.Annotations, evictions.EvictionInProgressAnnotationKey)
+	if _, err := client.CoreV1().Pods(p2.Namespace).Update(context.TODO(), p2, metav1.UpdateOptions{}); err != nil {
+		t.Fatalf("unable to update a pod: %v", err)
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	klog.Infof("Check the number of evictions in background decreased")
+	checkTotals(t, ctx, descheduler, 0, 3)
+
+	klog.Infof("Scenario: Re-run the descheduling cycle to re-request eviction in background")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 1, 3)
+
+	klog.Infof("Scenario: Eviction in background completed with a pod in completed state")
+	p2.Status.Phase = v1.PodSucceeded
+	if _, err := client.CoreV1().Pods(p2.Namespace).Update(context.TODO(), p2, metav1.UpdateOptions{}); err != nil {
+		t.Fatalf("unable to delete a pod: %v", err)
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	klog.Infof("Check the number of evictions in background decreased")
+	runDeschedulingCycleAndCheckTotals(t, ctx, nodes, descheduler, 0, 3)
+}
+
+func TestDeschedulingLimits(t *testing.T) {
+	initPluginRegistry()
+
+	tests := []struct {
+		description string
+		policy      *api.DeschedulerPolicy
+		limit       uint
+	}{
+		{
+			description: "limits per node",
+			policy: func() *api.DeschedulerPolicy {
+				policy := removePodsViolatingNodeTaintsPolicy()
+				policy.MaxNoOfPodsToEvictPerNode = utilptr.To[uint](4)
+				return policy
+			}(),
+			limit: uint(4),
+		},
+		{
+			description: "limits per namespace",
+			policy: func() *api.DeschedulerPolicy {
+				policy := removePodsViolatingNodeTaintsPolicy()
+				policy.MaxNoOfPodsToEvictPerNamespace = utilptr.To[uint](4)
+				return policy
+			}(),
+			limit: uint(4),
+		},
+		{
+			description: "limits per cycle",
+			policy: func() *api.DeschedulerPolicy {
+				policy := removePodsViolatingNodeTaintsPolicy()
+				policy.MaxNoOfPodsToEvictTotal = utilptr.To[uint](4)
+				return policy
+			}(),
+			limit: uint(4),
+		},
+	}
+
+	ownerRef1 := test.GetReplicaSetOwnerRefList()
+	updatePod := func(pod *v1.Pod) {
+		pod.Namespace = "dev"
+		pod.ObjectMeta.OwnerReferences = ownerRef1
+	}
+
+	updatePodWithEvictionInBackground := func(pod *v1.Pod) {
+		updatePod(pod)
+		pod.Annotations = map[string]string{
+			evictions.EvictionRequestAnnotationKey: "",
+		}
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
+			ctx := context.Background()
+			node1 := test.BuildTestNode("n1", 2000, 3000, 10, taintNodeNoSchedule)
+			node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+			nodes := []*v1.Node{node1, node2}
+			ctxCancel, cancel := context.WithCancel(ctx)
+			featureGates := featuregate.NewFeatureGate()
+			featureGates.Add(map[featuregate.Feature]featuregate.FeatureSpec{
+				features.EvictionsInBackground: {Default: true, PreRelease: featuregate.Alpha},
+			})
+			_, descheduler, client := initDescheduler(t, ctxCancel, featureGates, tc.policy, nil, node1, node2)
+			defer cancel()
+
+			var fakeEvictedPods []string
+			descheduler.podEvictionReactionFnc = func(*fakeclientset.Clientset) func(action core.Action) (bool, runtime.Object, error) {
+				return podEvictionReactionTestingFnc(&fakeEvictedPods, nil, podEvictionError)
+			}
+
+			var evictedPods []string
+			client.PrependReactor("create", "pods", podEvictionReactionTestingFnc(&evictedPods, func(name string) bool { return name == "p1" || name == "p2" }, nil))
+
+			rand.Seed(time.Now().UnixNano())
+			pods := []*v1.Pod{
+				test.BuildTestPod("p1", 100, 0, node1.Name, updatePodWithEvictionInBackground),
+				test.BuildTestPod("p2", 100, 0, node1.Name, updatePodWithEvictionInBackground),
+				test.BuildTestPod("p3", 100, 0, node1.Name, updatePod),
+				test.BuildTestPod("p4", 100, 0, node1.Name, updatePod),
+				test.BuildTestPod("p5", 100, 0, node1.Name, updatePod),
+			}
+
+			for i := 0; i < 10; i++ {
+				rand.Shuffle(len(pods), func(i, j int) { pods[i], pods[j] = pods[j], pods[i] })
+				func() {
+					for j := 0; j < 5; j++ {
+						idx := j
+						if _, err := client.CoreV1().Pods(pods[idx].Namespace).Create(context.TODO(), pods[idx], metav1.CreateOptions{}); err != nil {
+							t.Fatalf("unable to create a pod: %v", err)
+						}
+						defer func() {
+							if err := client.CoreV1().Pods(pods[idx].Namespace).Delete(context.TODO(), pods[idx].Name, metav1.DeleteOptions{}); err != nil {
+								t.Fatalf("unable to delete a pod: %v", err)
+							}
+						}()
+					}
+					time.Sleep(100 * time.Millisecond)
+
+					klog.Infof("2 evictions in background expected, 2 normal evictions")
+					err := descheduler.runDeschedulerLoop(ctx, nodes)
+					if err != nil {
+						t.Fatalf("Unable to run a descheduling loop: %v", err)
+					}
+					totalERs := descheduler.podEvictor.TotalEvictionRequests()
+					totalEs := descheduler.podEvictor.TotalEvicted()
+					if totalERs+totalEs > tc.limit {
+						t.Fatalf("Expected %v evictions and eviction requests in total, got %v instead", tc.limit, totalERs+totalEs)
+					}
+					t.Logf("Total evictions and eviction requests: %v (er=%v, e=%v)", totalERs+totalEs, totalERs, totalEs)
+				}()
+			}
+		})
+	}
+}
+
+func TestLoadAwareDescheduling(t *testing.T) {
+	initPluginRegistry()
+
+	ownerRef1 := test.GetReplicaSetOwnerRefList()
+	updatePod := func(pod *v1.Pod) {
+		pod.ObjectMeta.OwnerReferences = ownerRef1
+	}
+
+	ctx := context.Background()
+	node1 := test.BuildTestNode("n1", 2000, 3000, 10, taintNodeNoSchedule)
+	node2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	nodes := []*v1.Node{node1, node2}
+
+	p1 := test.BuildTestPod("p1", 300, 0, node1.Name, updatePod)
+	p2 := test.BuildTestPod("p2", 300, 0, node1.Name, updatePod)
+	p3 := test.BuildTestPod("p3", 300, 0, node1.Name, updatePod)
+	p4 := test.BuildTestPod("p4", 300, 0, node1.Name, updatePod)
+	p5 := test.BuildTestPod("p5", 300, 0, node1.Name, updatePod)
+
+	nodemetricses := []*v1beta1.NodeMetrics{
+		test.BuildNodeMetrics("n1", 2400, 3000),
+		test.BuildNodeMetrics("n2", 400, 0),
+	}
+
+	podmetricses := []*v1beta1.PodMetrics{
+		test.BuildPodMetrics("p1", 400, 0),
+		test.BuildPodMetrics("p2", 400, 0),
+		test.BuildPodMetrics("p3", 400, 0),
+		test.BuildPodMetrics("p4", 400, 0),
+		test.BuildPodMetrics("p5", 400, 0),
+	}
+
+	metricsClientset := fakemetricsclient.NewSimpleClientset()
+	for _, nodemetrics := range nodemetricses {
+		metricsClientset.Tracker().Create(nodesgvr, nodemetrics, "")
+	}
+	for _, podmetrics := range podmetricses {
+		metricsClientset.Tracker().Create(podsgvr, podmetrics, podmetrics.Namespace)
+	}
+
+	policy := lowNodeUtilizationPolicy(
+		api.ResourceThresholds{
+			v1.ResourceCPU:  30,
+			v1.ResourcePods: 30,
+		},
+		api.ResourceThresholds{
+			v1.ResourceCPU:  50,
+			v1.ResourcePods: 50,
+		},
+		true, // enabled metrics utilization
+	)
+	policy.MetricsCollector.Enabled = true
+
+	ctxCancel, cancel := context.WithCancel(ctx)
+	_, descheduler, _ := initDescheduler(
+		t,
+		ctxCancel,
+		initFeatureGates(),
+		policy,
+		metricsClientset,
+		node1, node2, p1, p2, p3, p4, p5)
+	defer cancel()
+
+	// This needs to be run since the metrics collector is started
+	// after newDescheduler in RunDeschedulerStrategies.
+	descheduler.metricsCollector.Collect(ctx)
+
+	err := descheduler.runDeschedulerLoop(ctx, nodes)
+	if err != nil {
+		t.Fatalf("Unable to run a descheduling loop: %v", err)
+	}
+	totalEs := descheduler.podEvictor.TotalEvicted()
+	if totalEs != 2 {
+		t.Fatalf("Expected %v evictions in total, got %v instead", 2, totalEs)
+	}
+	t.Logf("Total evictions: %v", totalEs)
+}

pkg/descheduler/evictions/errors.go

@@ -0,0 +1,45 @@
+package evictions
+
+type EvictionNodeLimitError struct {
+	node string
+}
+
+func (e EvictionNodeLimitError) Error() string {
+	return "maximum number of evicted pods per node reached"
+}
+
+func NewEvictionNodeLimitError(node string) *EvictionNodeLimitError {
+	return &EvictionNodeLimitError{
+		node: node,
+	}
+}
+
+var _ error = &EvictionNodeLimitError{}
+
+type EvictionNamespaceLimitError struct {
+	namespace string
+}
+
+func (e EvictionNamespaceLimitError) Error() string {
+	return "maximum number of evicted pods per namespace reached"
+}
+
+func NewEvictionNamespaceLimitError(namespace string) *EvictionNamespaceLimitError {
+	return &EvictionNamespaceLimitError{
+		namespace: namespace,
+	}
+}
+
+var _ error = &EvictionNamespaceLimitError{}
+
+type EvictionTotalLimitError struct{}
+
+func (e EvictionTotalLimitError) Error() string {
+	return "maximum number of evicted pods per a descheduling cycle reached"
+}
+
+func NewEvictionTotalLimitError() *EvictionTotalLimitError {
+	return &EvictionTotalLimitError{}
+}
+
+var _ error = &EvictionTotalLimitError{}

pkg/descheduler/evictions/evictions.go

@@ -19,17 +19,184 @@ package evictions
 import (
 	"context"
 	"fmt"
+	"strings"
+	"sync"
+	"time"
 
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/trace"
 	v1 "k8s.io/api/core/v1"
 	policy "k8s.io/api/policy/v1"
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/wait"
 	clientset "k8s.io/client-go/kubernetes"
+	"k8s.io/client-go/tools/cache"
 	"k8s.io/client-go/tools/events"
+	"k8s.io/component-base/featuregate"
 	"k8s.io/klog/v2"
-	"sigs.k8s.io/descheduler/metrics"
 
+	"sigs.k8s.io/descheduler/metrics"
 	eutils "sigs.k8s.io/descheduler/pkg/descheduler/evictions/utils"
+	"sigs.k8s.io/descheduler/pkg/features"
+	"sigs.k8s.io/descheduler/pkg/tracing"
+)
+
+var (
+	assumedEvictionRequestTimeoutSeconds uint          = 10 * 60 // 10 minutes
+	evictionRequestsCacheResyncPeriod    time.Duration = 10 * time.Minute
+	// syncedPollPeriod controls how often you look at the status of your sync funcs
+	syncedPollPeriod = 100 * time.Millisecond
+)
+
+type evictionRequestItem struct {
+	podName, podNamespace, podNodeName string
+	evictionAssumed                    bool
+	assumedTimestamp                   metav1.Time
+}
+
+type evictionRequestsCache struct {
+	mu                           sync.RWMutex
+	requests                     map[string]evictionRequestItem
+	requestsPerNode              map[string]uint
+	requestsPerNamespace         map[string]uint
+	requestsTotal                uint
+	assumedRequestTimeoutSeconds uint
+}
+
+func newEvictionRequestsCache(assumedRequestTimeoutSeconds uint) *evictionRequestsCache {
+	return &evictionRequestsCache{
+		requests:                     make(map[string]evictionRequestItem),
+		requestsPerNode:              make(map[string]uint),
+		requestsPerNamespace:         make(map[string]uint),
+		assumedRequestTimeoutSeconds: assumedRequestTimeoutSeconds,
+	}
+}
+
+func (erc *evictionRequestsCache) run(ctx context.Context) {
+	wait.UntilWithContext(ctx, erc.cleanCache, evictionRequestsCacheResyncPeriod)
+}
+
+// cleanCache removes all assumed entries that has not been confirmed
+// for more than a specified timeout
+func (erc *evictionRequestsCache) cleanCache(ctx context.Context) {
+	erc.mu.Lock()
+	defer erc.mu.Unlock()
+	klog.V(4).Infof("Cleaning cache of assumed eviction requests in background")
+	for uid, item := range erc.requests {
+		if item.evictionAssumed {
+			requestAgeSeconds := uint(metav1.Now().Sub(item.assumedTimestamp.Local()).Seconds())
+			if requestAgeSeconds > erc.assumedRequestTimeoutSeconds {
+				klog.V(4).InfoS("Assumed eviction request in background timed out, deleting", "timeout", erc.assumedRequestTimeoutSeconds, "podNamespace", item.podNamespace, "podName", item.podName)
+				erc.deleteItem(uid)
+			}
+		}
+	}
+}
+
+func (erc *evictionRequestsCache) evictionRequestsPerNode(nodeName string) uint {
+	erc.mu.RLock()
+	defer erc.mu.RUnlock()
+	return erc.requestsPerNode[nodeName]
+}
+
+func (erc *evictionRequestsCache) evictionRequestsPerNamespace(ns string) uint {
+	erc.mu.RLock()
+	defer erc.mu.RUnlock()
+	return erc.requestsPerNamespace[ns]
+}
+
+func (erc *evictionRequestsCache) evictionRequestsTotal() uint {
+	erc.mu.RLock()
+	defer erc.mu.RUnlock()
+	return erc.requestsTotal
+}
+
+func (erc *evictionRequestsCache) TotalEvictionRequests() uint {
+	erc.mu.RLock()
+	defer erc.mu.RUnlock()
+	return uint(len(erc.requests))
+}
+
+// getPodKey returns the string key of a pod.
+func getPodKey(pod *v1.Pod) string {
+	uid := string(pod.UID)
+	// Every pod is expected to have the UID set.
+	// When the descheduling framework is used for simulation
+	// user created workload may forget to set the UID.
+	if len(uid) == 0 {
+		panic(fmt.Errorf("cannot get cache key for %v/%v pod with empty UID", pod.Namespace, pod.Name))
+	}
+	return uid
+}
+
+func (erc *evictionRequestsCache) addPod(pod *v1.Pod) {
+	erc.mu.Lock()
+	defer erc.mu.Unlock()
+	uid := getPodKey(pod)
+	if _, exists := erc.requests[uid]; exists {
+		return
+	}
+	erc.requests[uid] = evictionRequestItem{podNamespace: pod.Namespace, podName: pod.Name, podNodeName: pod.Spec.NodeName}
+	erc.requestsPerNode[pod.Spec.NodeName]++
+	erc.requestsPerNamespace[pod.Namespace]++
+	erc.requestsTotal++
+}
+
+func (erc *evictionRequestsCache) assumePod(pod *v1.Pod) {
+	erc.mu.Lock()
+	defer erc.mu.Unlock()
+	uid := getPodKey(pod)
+	if _, exists := erc.requests[uid]; exists {
+		return
+	}
+	erc.requests[uid] = evictionRequestItem{
+		podNamespace:     pod.Namespace,
+		podName:          pod.Name,
+		podNodeName:      pod.Spec.NodeName,
+		evictionAssumed:  true,
+		assumedTimestamp: metav1.NewTime(time.Now()),
+	}
+	erc.requestsPerNode[pod.Spec.NodeName]++
+	erc.requestsPerNamespace[pod.Namespace]++
+	erc.requestsTotal++
+}
+
+// no locking, expected to be invoked from protected methods only
+func (erc *evictionRequestsCache) deleteItem(uid string) {
+	erc.requestsPerNode[erc.requests[uid].podNodeName]--
+	if erc.requestsPerNode[erc.requests[uid].podNodeName] == 0 {
+		delete(erc.requestsPerNode, erc.requests[uid].podNodeName)
+	}
+	erc.requestsPerNamespace[erc.requests[uid].podNamespace]--
+	if erc.requestsPerNamespace[erc.requests[uid].podNamespace] == 0 {
+		delete(erc.requestsPerNamespace, erc.requests[uid].podNamespace)
+	}
+	erc.requestsTotal--
+	delete(erc.requests, uid)
+}
+
+func (erc *evictionRequestsCache) deletePod(pod *v1.Pod) {
+	erc.mu.Lock()
+	defer erc.mu.Unlock()
+	uid := getPodKey(pod)
+	if _, exists := erc.requests[uid]; exists {
+		erc.deleteItem(uid)
+	}
+}
+
+func (erc *evictionRequestsCache) hasPod(pod *v1.Pod) bool {
+	erc.mu.RLock()
+	defer erc.mu.RUnlock()
+	uid := getPodKey(pod)
+	_, exists := erc.requests[uid]
+	return exists
+}
+
+var (
+	EvictionRequestAnnotationKey    = "descheduler.alpha.kubernetes.io/request-evict-only"
+	EvictionInProgressAnnotationKey = "descheduler.alpha.kubernetes.io/eviction-in-progress"
+	EvictionInBackgroundErrorText   = "Eviction triggered evacuation"
 )
 
 // nodePodEvictedCount keeps count of pods evicted on node
@@ -39,145 +206,368 @@ type (
 )
 
 type PodEvictor struct {
-	client                     clientset.Interface
+	mu                               sync.RWMutex
-	nodes                      []*v1.Node
+	client                           clientset.Interface
-	policyGroupVersion         string
+	policyGroupVersion               string
-	dryRun                     bool
+	dryRun                           bool
-	maxPodsToEvictPerNode      *uint
+	evictionFailureEventNotification bool
-	maxPodsToEvictPerNamespace *uint
+	maxPodsToEvictPerNode            *uint
-	nodepodCount               nodePodEvictedCount
+	maxPodsToEvictPerNamespace       *uint
-	namespacePodCount          namespacePodEvictCount
+	maxPodsToEvictTotal              *uint
-	metricsEnabled             bool
+	nodePodCount                     nodePodEvictedCount
-	eventRecorder              events.EventRecorder
+	namespacePodCount                namespacePodEvictCount
+	totalPodCount                    uint
+	metricsEnabled                   bool
+	eventRecorder                    events.EventRecorder
+	erCache                          *evictionRequestsCache
+	featureGates                     featuregate.FeatureGate
+
+	// registeredHandlers contains the registrations of all handlers. It's used to check if all handlers have finished syncing before the scheduling cycles start.
+	registeredHandlers []cache.ResourceEventHandlerRegistration
 }
 
 func NewPodEvictor(
+	ctx context.Context,
 	client clientset.Interface,
-	policyGroupVersion string,
-	dryRun bool,
-	maxPodsToEvictPerNode *uint,
-	maxPodsToEvictPerNamespace *uint,
-	nodes []*v1.Node,
-	metricsEnabled bool,
 	eventRecorder events.EventRecorder,
-) *PodEvictor {
+	podInformer cache.SharedIndexInformer,
-	nodePodCount := make(nodePodEvictedCount)
+	featureGates featuregate.FeatureGate,
-	namespacePodCount := make(namespacePodEvictCount)
+	options *Options,
-	for _, node := range nodes {
+) (*PodEvictor, error) {
-		// Initialize podsEvicted till now with 0.
+	if options == nil {
-		nodePodCount[node.Name] = 0
+		options = NewOptions()
 	}
 
-	return &PodEvictor{
+	podEvictor := &PodEvictor{
-		client:                     client,
+		client:                           client,
-		nodes:                      nodes,
+		eventRecorder:                    eventRecorder,
-		policyGroupVersion:         policyGroupVersion,
+		policyGroupVersion:               options.policyGroupVersion,
-		dryRun:                     dryRun,
+		dryRun:                           options.dryRun,
-		maxPodsToEvictPerNode:      maxPodsToEvictPerNode,
+		evictionFailureEventNotification: options.evictionFailureEventNotification,
-		maxPodsToEvictPerNamespace: maxPodsToEvictPerNamespace,
+		maxPodsToEvictPerNode:            options.maxPodsToEvictPerNode,
-		nodepodCount:               nodePodCount,
+		maxPodsToEvictPerNamespace:       options.maxPodsToEvictPerNamespace,
-		namespacePodCount:          namespacePodCount,
+		maxPodsToEvictTotal:              options.maxPodsToEvictTotal,
-		metricsEnabled:             metricsEnabled,
+		metricsEnabled:                   options.metricsEnabled,
-		eventRecorder:              eventRecorder,
+		nodePodCount:                     make(nodePodEvictedCount),
+		namespacePodCount:                make(namespacePodEvictCount),
+		featureGates:                     featureGates,
 	}
+
+	if featureGates.Enabled(features.EvictionsInBackground) {
+		erCache := newEvictionRequestsCache(assumedEvictionRequestTimeoutSeconds)
+
+		handlerRegistration, err := podInformer.AddEventHandler(
+			cache.ResourceEventHandlerFuncs{
+				AddFunc: func(obj interface{}) {
+					pod, ok := obj.(*v1.Pod)
+					if !ok {
+						klog.ErrorS(nil, "Cannot convert to *v1.Pod", "obj", obj)
+						return
+					}
+					if _, exists := pod.Annotations[EvictionRequestAnnotationKey]; exists {
+						if _, exists := pod.Annotations[EvictionInProgressAnnotationKey]; exists {
+							// Ignore completed/suceeeded or failed pods
+							if pod.Status.Phase != v1.PodSucceeded && pod.Status.Phase != v1.PodFailed {
+								klog.V(3).InfoS("Eviction in background detected. Adding pod to the cache.", "pod", klog.KObj(pod))
+								erCache.addPod(pod)
+							}
+						}
+					}
+				},
+				UpdateFunc: func(oldObj, newObj interface{}) {
+					oldPod, ok := oldObj.(*v1.Pod)
+					if !ok {
+						klog.ErrorS(nil, "Cannot convert oldObj to *v1.Pod", "oldObj", oldObj)
+						return
+					}
+					newPod, ok := newObj.(*v1.Pod)
+					if !ok {
+						klog.ErrorS(nil, "Cannot convert newObj to *v1.Pod", "newObj", newObj)
+						return
+					}
+					// Ignore pod's that are not subject to an eviction in background
+					if _, exists := newPod.Annotations[EvictionRequestAnnotationKey]; !exists {
+						if erCache.hasPod(newPod) {
+							klog.V(3).InfoS("Pod with eviction in background lost annotation. Removing pod from the cache.", "pod", klog.KObj(newPod))
+						}
+						erCache.deletePod(newPod)
+						return
+					}
+					// Remove completed/suceeeded or failed pods from the cache
+					if newPod.Status.Phase == v1.PodSucceeded || newPod.Status.Phase == v1.PodFailed {
+						klog.V(3).InfoS("Pod with eviction in background completed. Removing pod from the cache.", "pod", klog.KObj(newPod))
+						erCache.deletePod(newPod)
+						return
+					}
+					// Ignore any pod that does not have eviction in progress
+					if _, exists := newPod.Annotations[EvictionInProgressAnnotationKey]; !exists {
+						// In case EvictionInProgressAnnotationKey annotation is not present/removed
+						// it's unclear whether the eviction was restarted or terminated.
+						// If the eviction gets restarted the pod needs to be removed from the cache
+						// to allow re-triggering the eviction.
+						if _, exists := oldPod.Annotations[EvictionInProgressAnnotationKey]; !exists {
+							return
+						}
+						// the annotation was removed -> remove the pod from the cache to allow to
+						// request for eviction again. In case the eviction got restarted requesting
+						// the eviction again is expected to be a no-op. In case the eviction
+						// got terminated with no-retry, requesting a new eviction is a normal
+						// operation.
+						klog.V(3).InfoS("Eviction in background canceled (annotation removed). Removing pod from the cache.", "annotation", EvictionInProgressAnnotationKey, "pod", klog.KObj(newPod))
+						erCache.deletePod(newPod)
+						return
+					}
+					// Pick up the eviction in progress
+					if !erCache.hasPod(newPod) {
+						klog.V(3).InfoS("Eviction in background detected. Updating the cache.", "pod", klog.KObj(newPod))
+					}
+					erCache.addPod(newPod)
+				},
+				DeleteFunc: func(obj interface{}) {
+					var pod *v1.Pod
+					switch t := obj.(type) {
+					case *v1.Pod:
+						pod = t
+					case cache.DeletedFinalStateUnknown:
+						var ok bool
+						pod, ok = t.Obj.(*v1.Pod)
+						if !ok {
+							klog.ErrorS(nil, "Cannot convert to *v1.Pod", "obj", t.Obj)
+							return
+						}
+					default:
+						klog.ErrorS(nil, "Cannot convert to *v1.Pod", "obj", t)
+						return
+					}
+					if erCache.hasPod(pod) {
+						klog.V(3).InfoS("Pod with eviction in background deleted/evicted. Removing pod from the cache.", "pod", klog.KObj(pod))
+					}
+					erCache.deletePod(pod)
+				},
+			},
+		)
+		if err != nil {
+			return nil, fmt.Errorf("unable to register event handler for pod evictor: %v", err)
+		}
+
+		podEvictor.registeredHandlers = append(podEvictor.registeredHandlers, handlerRegistration)
+
+		go erCache.run(ctx)
+
+		podEvictor.erCache = erCache
+	}
+
+	return podEvictor, nil
+}
+
+// WaitForEventHandlersSync waits for EventHandlers to sync.
+// It returns true if it was successful, false if the controller should shut down
+func (pe *PodEvictor) WaitForEventHandlersSync(ctx context.Context) error {
+	return wait.PollUntilContextCancel(ctx, syncedPollPeriod, true, func(ctx context.Context) (done bool, err error) {
+		for _, handler := range pe.registeredHandlers {
+			if !handler.HasSynced() {
+				return false, nil
+			}
+		}
+		return true, nil
+	})
 }
 
 // NodeEvicted gives a number of pods evicted for node
 func (pe *PodEvictor) NodeEvicted(node *v1.Node) uint {
-	return pe.nodepodCount[node.Name]
+	pe.mu.RLock()
+	defer pe.mu.RUnlock()
+	return pe.nodePodCount[node.Name]
 }
 
 // TotalEvicted gives a number of pods evicted through all nodes
 func (pe *PodEvictor) TotalEvicted() uint {
-	var total uint
+	pe.mu.RLock()
-	for _, count := range pe.nodepodCount {
+	defer pe.mu.RUnlock()
-		total += count
+	return pe.totalPodCount
-	}
-	return total
 }
 
-// NodeLimitExceeded checks if the number of evictions for a node was exceeded
+func (pe *PodEvictor) ResetCounters() {
-func (pe *PodEvictor) NodeLimitExceeded(node *v1.Node) bool {
+	pe.mu.Lock()
-	if pe.maxPodsToEvictPerNode != nil {
+	defer pe.mu.Unlock()
-		return pe.nodepodCount[node.Name] == *pe.maxPodsToEvictPerNode
+	pe.nodePodCount = make(nodePodEvictedCount)
+	pe.namespacePodCount = make(namespacePodEvictCount)
+	pe.totalPodCount = 0
+}
+
+func (pe *PodEvictor) SetClient(client clientset.Interface) {
+	pe.mu.Lock()
+	defer pe.mu.Unlock()
+	pe.client = client
+}
+
+func (pe *PodEvictor) evictionRequestsTotal() uint {
+	if pe.featureGates.Enabled(features.EvictionsInBackground) {
+		return pe.erCache.evictionRequestsTotal()
+	} else {
+		return 0
+	}
+}
+
+func (pe *PodEvictor) evictionRequestsPerNode(node string) uint {
+	if pe.featureGates.Enabled(features.EvictionsInBackground) {
+		return pe.erCache.evictionRequestsPerNode(node)
+	} else {
+		return 0
+	}
+}
+
+func (pe *PodEvictor) evictionRequestsPerNamespace(ns string) uint {
+	if pe.featureGates.Enabled(features.EvictionsInBackground) {
+		return pe.erCache.evictionRequestsPerNamespace(ns)
+	} else {
+		return 0
+	}
+}
+
+func (pe *PodEvictor) EvictionRequests(node *v1.Node) uint {
+	pe.mu.RLock()
+	defer pe.mu.RUnlock()
+	return pe.evictionRequestsTotal()
+}
+
+func (pe *PodEvictor) TotalEvictionRequests() uint {
+	pe.mu.RLock()
+	defer pe.mu.RUnlock()
+	if pe.featureGates.Enabled(features.EvictionsInBackground) {
+		return pe.erCache.TotalEvictionRequests()
+	} else {
+		return 0
 	}
-	return false
 }
 
 // EvictOptions provides a handle for passing additional info to EvictPod
 type EvictOptions struct {
 	// Reason allows for passing details about the specific eviction for logging.
 	Reason string
+	// ProfileName allows for passing details about profile for observability.
+	ProfileName string
+	// StrategyName allows for passing details about strategy for observability.
+	StrategyName string
 }
 
 // EvictPod evicts a pod while exercising eviction limits.
 // Returns true when the pod is evicted on the server side.
-func (pe *PodEvictor) EvictPod(ctx context.Context, pod *v1.Pod, opts EvictOptions) bool {
+func (pe *PodEvictor) EvictPod(ctx context.Context, pod *v1.Pod, opts EvictOptions) error {
-	// TODO: Replace context-propagated Strategy name with a defined framework handle for accessing Strategy info
+	if len(pod.UID) == 0 {
-	strategy := ""
+		klog.InfoS("Ignoring pod eviction due to missing UID", "pod", pod)
-	if ctx.Value("strategyName") != nil {
+		return fmt.Errorf("Pod %v is missing UID", klog.KObj(pod))
-		strategy = ctx.Value("strategyName").(string)
+	}
+
+	if pe.featureGates.Enabled(features.EvictionsInBackground) {
+		// eviction in background requested
+		if _, exists := pod.Annotations[EvictionRequestAnnotationKey]; exists {
+			if pe.erCache.hasPod(pod) {
+				klog.V(3).InfoS("Eviction in background already requested (ignoring)", "pod", klog.KObj(pod))
+				return nil
+			}
+		}
+	}
+
+	pe.mu.Lock()
+	defer pe.mu.Unlock()
+
+	var span trace.Span
+	ctx, span = tracing.Tracer().Start(ctx, "EvictPod", trace.WithAttributes(attribute.String("podName", pod.Name), attribute.String("podNamespace", pod.Namespace), attribute.String("reason", opts.Reason), attribute.String("operation", tracing.EvictOperation)))
+	defer span.End()
+
+	if pe.maxPodsToEvictTotal != nil && pe.totalPodCount+pe.evictionRequestsTotal()+1 > *pe.maxPodsToEvictTotal {
+		err := NewEvictionTotalLimitError()
+		if pe.metricsEnabled {
+			metrics.PodsEvicted.With(map[string]string{"result": err.Error(), "strategy": opts.StrategyName, "namespace": pod.Namespace, "node": pod.Spec.NodeName, "profile": opts.ProfileName}).Inc()
+		}
+		span.AddEvent("Eviction Failed", trace.WithAttributes(attribute.String("node", pod.Spec.NodeName), attribute.String("err", err.Error())))
+		klog.ErrorS(err, "Error evicting pod", "limit", *pe.maxPodsToEvictTotal)
+		if pe.evictionFailureEventNotification {
+			pe.eventRecorder.Eventf(pod, nil, v1.EventTypeWarning, "EvictionFailed", "Descheduled", "pod eviction from %v node by sigs.k8s.io/descheduler failed: total eviction limit exceeded (%v)", pod.Spec.NodeName, *pe.maxPodsToEvictTotal)
+		}
+		return err
 	}
 
 	if pod.Spec.NodeName != "" {
-		if pe.maxPodsToEvictPerNode != nil && pe.nodepodCount[pod.Spec.NodeName]+1 > *pe.maxPodsToEvictPerNode {
+		if pe.maxPodsToEvictPerNode != nil && pe.nodePodCount[pod.Spec.NodeName]+pe.evictionRequestsPerNode(pod.Spec.NodeName)+1 > *pe.maxPodsToEvictPerNode {
+			err := NewEvictionNodeLimitError(pod.Spec.NodeName)
 			if pe.metricsEnabled {
-				metrics.PodsEvicted.With(map[string]string{"result": "maximum number of pods per node reached", "strategy": strategy, "namespace": pod.Namespace, "node": pod.Spec.NodeName}).Inc()
+				metrics.PodsEvicted.With(map[string]string{"result": err.Error(), "strategy": opts.StrategyName, "namespace": pod.Namespace, "node": pod.Spec.NodeName, "profile": opts.ProfileName}).Inc()
 			}
-			klog.ErrorS(fmt.Errorf("Maximum number of evicted pods per node reached"), "limit", *pe.maxPodsToEvictPerNode, "node", pod.Spec.NodeName)
+			span.AddEvent("Eviction Failed", trace.WithAttributes(attribute.String("node", pod.Spec.NodeName), attribute.String("err", err.Error())))
-			return false
+			klog.ErrorS(err, "Error evicting pod", "limit", *pe.maxPodsToEvictPerNode, "node", pod.Spec.NodeName)
+			if pe.evictionFailureEventNotification {
+				pe.eventRecorder.Eventf(pod, nil, v1.EventTypeWarning, "EvictionFailed", "Descheduled", "pod eviction from %v node by sigs.k8s.io/descheduler failed: node eviction limit exceeded (%v)", pod.Spec.NodeName, *pe.maxPodsToEvictPerNode)
+			}
+			return err
 		}
 	}
 
-	if pe.maxPodsToEvictPerNamespace != nil && pe.namespacePodCount[pod.Namespace]+1 > *pe.maxPodsToEvictPerNamespace {
+	if pe.maxPodsToEvictPerNamespace != nil && pe.namespacePodCount[pod.Namespace]+pe.evictionRequestsPerNamespace(pod.Namespace)+1 > *pe.maxPodsToEvictPerNamespace {
+		err := NewEvictionNamespaceLimitError(pod.Namespace)
 		if pe.metricsEnabled {
-			metrics.PodsEvicted.With(map[string]string{"result": "maximum number of pods per namespace reached", "strategy": strategy, "namespace": pod.Namespace, "node": pod.Spec.NodeName}).Inc()
+			metrics.PodsEvicted.With(map[string]string{"result": err.Error(), "strategy": opts.StrategyName, "namespace": pod.Namespace, "node": pod.Spec.NodeName, "profile": opts.ProfileName}).Inc()
 		}
-		klog.ErrorS(fmt.Errorf("Maximum number of evicted pods per namespace reached"), "limit", *pe.maxPodsToEvictPerNamespace, "namespace", pod.Namespace)
+		span.AddEvent("Eviction Failed", trace.WithAttributes(attribute.String("node", pod.Spec.NodeName), attribute.String("err", err.Error())))
-		return false
+		klog.ErrorS(err, "Error evicting pod", "limit", *pe.maxPodsToEvictPerNamespace, "namespace", pod.Namespace, "pod", klog.KObj(pod))
+		if pe.evictionFailureEventNotification {
+			pe.eventRecorder.Eventf(pod, nil, v1.EventTypeWarning, "EvictionFailed", "Descheduled", "pod eviction from %v node by sigs.k8s.io/descheduler failed: namespace eviction limit exceeded (%v)", pod.Spec.NodeName, *pe.maxPodsToEvictPerNamespace)
+		}
+		return err
 	}
 
-	err := evictPod(ctx, pe.client, pod, pe.policyGroupVersion)
+	ignore, err := pe.evictPod(ctx, pod)
 	if err != nil {
 		// err is used only for logging purposes
+		span.AddEvent("Eviction Failed", trace.WithAttributes(attribute.String("node", pod.Spec.NodeName), attribute.String("err", err.Error())))
 		klog.ErrorS(err, "Error evicting pod", "pod", klog.KObj(pod), "reason", opts.Reason)
 		if pe.metricsEnabled {
-			metrics.PodsEvicted.With(map[string]string{"result": "error", "strategy": strategy, "namespace": pod.Namespace, "node": pod.Spec.NodeName}).Inc()
+			metrics.PodsEvicted.With(map[string]string{"result": "error", "strategy": opts.StrategyName, "namespace": pod.Namespace, "node": pod.Spec.NodeName, "profile": opts.ProfileName}).Inc()
 		}
-		return false
+		if pe.evictionFailureEventNotification {
+			pe.eventRecorder.Eventf(pod, nil, v1.EventTypeWarning, "EvictionFailed", "Descheduled", "pod eviction from %v node by sigs.k8s.io/descheduler failed: %v", pod.Spec.NodeName, err.Error())
+		}
+		return err
+	}
+
+	if ignore {
+		return nil
 	}
 
 	if pod.Spec.NodeName != "" {
-		pe.nodepodCount[pod.Spec.NodeName]++
+		pe.nodePodCount[pod.Spec.NodeName]++
 	}
 	pe.namespacePodCount[pod.Namespace]++
+	pe.totalPodCount++
 
 	if pe.metricsEnabled {
-		metrics.PodsEvicted.With(map[string]string{"result": "success", "strategy": strategy, "namespace": pod.Namespace, "node": pod.Spec.NodeName}).Inc()
+		metrics.PodsEvicted.With(map[string]string{"result": "success", "strategy": opts.StrategyName, "namespace": pod.Namespace, "node": pod.Spec.NodeName, "profile": opts.ProfileName}).Inc()
 	}
 
 	if pe.dryRun {
-		klog.V(1).InfoS("Evicted pod in dry run mode", "pod", klog.KObj(pod), "reason", opts.Reason, "strategy", strategy, "node", pod.Spec.NodeName)
+		klog.V(1).InfoS("Evicted pod in dry run mode", "pod", klog.KObj(pod), "reason", opts.Reason, "strategy", opts.StrategyName, "node", pod.Spec.NodeName, "profile", opts.ProfileName)
 	} else {
-		klog.V(1).InfoS("Evicted pod", "pod", klog.KObj(pod), "reason", opts.Reason, "strategy", strategy, "node", pod.Spec.NodeName)
+		klog.V(1).InfoS("Evicted pod", "pod", klog.KObj(pod), "reason", opts.Reason, "strategy", opts.StrategyName, "node", pod.Spec.NodeName, "profile", opts.ProfileName)
 		reason := opts.Reason
 		if len(reason) == 0 {
-			reason = strategy
+			reason = opts.StrategyName
 			if len(reason) == 0 {
 				reason = "NotSet"
 			}
 		}
-		pe.eventRecorder.Eventf(pod, nil, v1.EventTypeNormal, reason, "Descheduled", "pod evicted from %v node by sigs.k8s.io/descheduler", pod.Spec.NodeName)
+		pe.eventRecorder.Eventf(pod, nil, v1.EventTypeNormal, reason, "Descheduled", "pod eviction from %v node by sigs.k8s.io/descheduler", pod.Spec.NodeName)
 	}
-	return true
+	return nil
 }
 
-func evictPod(ctx context.Context, client clientset.Interface, pod *v1.Pod, policyGroupVersion string) error {
+// return (ignore, err)
+func (pe *PodEvictor) evictPod(ctx context.Context, pod *v1.Pod) (bool, error) {
 	deleteOptions := &metav1.DeleteOptions{}
 	// GracePeriodSeconds ?
 	eviction := &policy.Eviction{
 		TypeMeta: metav1.TypeMeta{
-			APIVersion: policyGroupVersion,
+			APIVersion: pe.policyGroupVersion,
 			Kind:       eutils.EvictionKind,
 		},
 		ObjectMeta: metav1.ObjectMeta{
@@ -186,13 +576,36 @@ func evictPod(ctx context.Context, client clientset.Interface, pod *v1.Pod, poli
 		},
 		DeleteOptions: deleteOptions,
 	}
-	err := client.PolicyV1().Evictions(eviction.Namespace).Evict(ctx, eviction)
+	err := pe.client.PolicyV1().Evictions(eviction.Namespace).Evict(ctx, eviction)
+	if err == nil {
+		return false, nil
+	}
+	if pe.featureGates.Enabled(features.EvictionsInBackground) {
+		// eviction in background requested
+		if _, exists := pod.Annotations[EvictionRequestAnnotationKey]; exists {
+			// Simulating https://github.com/kubevirt/kubevirt/pull/11532/files#diff-059cc1fc09e8b469143348cc3aa80b40de987670e008fa18a6fe010061f973c9R77
+			if apierrors.IsTooManyRequests(err) && strings.Contains(err.Error(), EvictionInBackgroundErrorText) {
+				// Ignore eviction of any pod that's failed or completed.
+				// It can happen an eviction in background ends up with the pod stuck in the completed state.
+				// Normally, any request eviction is expected to end with the pod deletion.
+				// However, some custom eviction policies may end up with completed pods around.
+				// Which leads to all the completed pods to be considered still as unfinished evictions in background.
+				if pod.Status.Phase == v1.PodSucceeded || pod.Status.Phase == v1.PodFailed {
+					klog.V(3).InfoS("Ignoring eviction of a completed/failed pod", "pod", klog.KObj(pod))
+					return true, nil
+				}
+				klog.V(3).InfoS("Eviction in background assumed", "pod", klog.KObj(pod))
+				pe.erCache.assumePod(pod)
+				return true, nil
+			}
+		}
+	}
 
 	if apierrors.IsTooManyRequests(err) {
-		return fmt.Errorf("error when evicting pod (ignoring) %q: %v", pod.Name, err)
+		return false, fmt.Errorf("error when evicting pod (ignoring) %q: %v", pod.Name, err)
 	}
 	if apierrors.IsNotFound(err) {
-		return fmt.Errorf("pod not found when evicting %q: %v", pod.Name, err)
+		return false, fmt.Errorf("pod not found when evicting %q: %v", pod.Name, err)
 	}
-	return err
+	return false, err
 }

pkg/descheduler/evictions/evictions_test.go

@@ -18,54 +18,107 @@ package evictions
 
 import (
 	"context"
+	"fmt"
+	"reflect"
 	"testing"
+	"time"
 
 	v1 "k8s.io/api/core/v1"
+	policy "k8s.io/api/policy/v1"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/util/wait"
+	"k8s.io/client-go/informers"
 	"k8s.io/client-go/kubernetes/fake"
+	fakeclientset "k8s.io/client-go/kubernetes/fake"
 	core "k8s.io/client-go/testing"
+	"k8s.io/client-go/tools/events"
+	"k8s.io/component-base/featuregate"
+	"k8s.io/klog/v2"
+	utilptr "k8s.io/utils/ptr"
+
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
+	"sigs.k8s.io/descheduler/pkg/features"
 	"sigs.k8s.io/descheduler/pkg/utils"
 	"sigs.k8s.io/descheduler/test"
 )
 
+const (
+	notFoundText    = "pod not found when evicting \"%s\": pods \"%s\" not found"
+	tooManyRequests = "error when evicting pod (ignoring) \"%s\": Too many requests: too many requests"
+)
+
+func initFeatureGates() featuregate.FeatureGate {
+	featureGates := featuregate.NewFeatureGate()
+	featureGates.Add(map[featuregate.Feature]featuregate.FeatureSpec{
+		features.EvictionsInBackground: {Default: true, PreRelease: featuregate.Alpha},
+	})
+	return featureGates
+}
+
 func TestEvictPod(t *testing.T) {
-	ctx := context.Background()
 	node1 := test.BuildTestNode("node1", 1000, 2000, 9, nil)
 	pod1 := test.BuildTestPod("p1", 400, 0, "node1", nil)
 	tests := []struct {
 		description string
 		node        *v1.Node
-		pod         *v1.Pod
+		evictedPod  *v1.Pod
-		pods        []v1.Pod
+		pods        []runtime.Object
-		want        error
+		wantErr     error
 	}{
 		{
 			description: "test pod eviction - pod present",
 			node:        node1,
-			pod:         pod1,
+			evictedPod:  pod1,
-			pods:        []v1.Pod{*pod1},
+			pods:        []runtime.Object{pod1},
-			want:        nil,
 		},
 		{
-			description: "test pod eviction - pod absent",
+			description: "test pod eviction - pod absent (not found error)",
 			node:        node1,
-			pod:         pod1,
+			evictedPod:  pod1,
-			pods:        []v1.Pod{*test.BuildTestPod("p2", 400, 0, "node1", nil), *test.BuildTestPod("p3", 450, 0, "node1", nil)},
+			pods:        []runtime.Object{test.BuildTestPod("p2", 400, 0, "node1", nil), test.BuildTestPod("p3", 450, 0, "node1", nil)},
-			want:        nil,
+			wantErr:     fmt.Errorf(notFoundText, pod1.Name, pod1.Name),
+		},
+		{
+			description: "test pod eviction - pod absent (too many requests error)",
+			node:        node1,
+			evictedPod:  pod1,
+			pods:        []runtime.Object{test.BuildTestPod("p2", 400, 0, "node1", nil), test.BuildTestPod("p3", 450, 0, "node1", nil)},
+			wantErr:     fmt.Errorf(tooManyRequests, pod1.Name),
 		},
 	}
 
 	for _, test := range tests {
-		fakeClient := &fake.Clientset{}
+		t.Run(test.description, func(t *testing.T) {
-		fakeClient.Fake.AddReactor("list", "pods", func(action core.Action) (bool, runtime.Object, error) {
+			ctx := context.Background()
-			return true, &v1.PodList{Items: test.pods}, nil
+			fakeClient := fake.NewClientset(test.pods...)
+			fakeClient.PrependReactor("create", "pods/eviction", func(action core.Action) (handled bool, ret runtime.Object, err error) {
+				return true, nil, test.wantErr
+			})
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			eventRecorder := &events.FakeRecorder{}
+			podEvictor, err := NewPodEvictor(
+				ctx,
+				fakeClient,
+				eventRecorder,
+				sharedInformerFactory.Core().V1().Pods().Informer(),
+				initFeatureGates(),
+				NewOptions(),
+			)
+			if err != nil {
+				t.Fatalf("Unexpected error when creating a pod evictor: %v", err)
+			}
+
+			_, got := podEvictor.evictPod(ctx, test.evictedPod)
+			if got != test.wantErr {
+				t.Errorf("Test error for Desc: %s. Expected %v pod eviction to be %v, got %v", test.description, test.evictedPod.Name, test.wantErr, got)
+			}
 		})
-		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)
-		}
 	}
 }
 
@@ -113,3 +166,319 @@ func TestPodTypes(t *testing.T) {
 		t.Errorf("Expected p1 to be a normal pod.")
 	}
 }
+
+func TestNewPodEvictor(t *testing.T) {
+	ctx := context.Background()
+
+	pod1 := test.BuildTestPod("pod", 400, 0, "node", nil)
+	type podEvictorTest struct {
+		description                      string
+		pod                              *v1.Pod
+		dryRun                           bool
+		evictionFailureEventNotification *bool
+		maxPodsToEvictTotal              *uint
+		maxPodsToEvictPerNode            *uint
+		maxPodsToEvictPerNamespace       *uint
+		expectedNodeEvictions            uint
+		expectedTotalEvictions           uint
+		expectedError                    error
+		// expectedEvent is a slice of strings representing expected events.
+		// Each string in the slice should follow the format: "EventType Reason Message".
+		// - "Warning Failed processing failed"
+		events []string
+	}
+	tests := []podEvictorTest{
+		{
+			description:                      "one eviction expected with eviction failure event notification",
+			pod:                              pod1,
+			evictionFailureEventNotification: utilptr.To[bool](true),
+			maxPodsToEvictTotal:              utilptr.To[uint](1),
+			maxPodsToEvictPerNode:            utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace:       utilptr.To[uint](1),
+			expectedNodeEvictions:            1,
+			expectedTotalEvictions:           1,
+			expectedError:                    nil,
+			events:                           []string{"Normal NotSet pod eviction from node node by sigs.k8s.io/descheduler"},
+		},
+		{
+			description:                      "eviction limit exceeded on total with eviction failure event notification",
+			pod:                              pod1,
+			evictionFailureEventNotification: utilptr.To[bool](true),
+			maxPodsToEvictTotal:              utilptr.To[uint](0),
+			maxPodsToEvictPerNode:            utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace:       utilptr.To[uint](1),
+			expectedNodeEvictions:            0,
+			expectedTotalEvictions:           0,
+			expectedError:                    NewEvictionTotalLimitError(),
+			events:                           []string{"Warning EvictionFailed pod eviction from node node by sigs.k8s.io/descheduler failed: total eviction limit exceeded (0)"},
+		},
+		{
+			description:                      "eviction limit exceeded on node with eviction failure event notification",
+			pod:                              pod1,
+			evictionFailureEventNotification: utilptr.To[bool](true),
+			maxPodsToEvictTotal:              utilptr.To[uint](1),
+			maxPodsToEvictPerNode:            utilptr.To[uint](0),
+			maxPodsToEvictPerNamespace:       utilptr.To[uint](1),
+			expectedNodeEvictions:            0,
+			expectedTotalEvictions:           0,
+			expectedError:                    NewEvictionNodeLimitError("node"),
+			events:                           []string{"Warning EvictionFailed pod eviction from node node by sigs.k8s.io/descheduler failed: node eviction limit exceeded (0)"},
+		},
+		{
+			description:                      "eviction limit exceeded on node with eviction failure event notification",
+			pod:                              pod1,
+			evictionFailureEventNotification: utilptr.To[bool](true),
+			maxPodsToEvictTotal:              utilptr.To[uint](1),
+			maxPodsToEvictPerNode:            utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace:       utilptr.To[uint](0),
+			expectedNodeEvictions:            0,
+			expectedTotalEvictions:           0,
+			expectedError:                    NewEvictionNamespaceLimitError("default"),
+			events:                           []string{"Warning EvictionFailed pod eviction from node node by sigs.k8s.io/descheduler failed: namespace eviction limit exceeded (0)"},
+		},
+		{
+			description:                      "eviction error with eviction failure event notification",
+			pod:                              pod1,
+			evictionFailureEventNotification: utilptr.To[bool](true),
+			maxPodsToEvictTotal:              utilptr.To[uint](1),
+			maxPodsToEvictPerNode:            utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace:       utilptr.To[uint](1),
+			expectedNodeEvictions:            0,
+			expectedTotalEvictions:           0,
+			expectedError:                    fmt.Errorf("eviction error"),
+			events:                           []string{"Warning EvictionFailed pod eviction from node node by sigs.k8s.io/descheduler failed: eviction error"},
+		},
+		{
+			description:                      "eviction with dryRun with eviction failure event notification",
+			pod:                              pod1,
+			dryRun:                           true,
+			evictionFailureEventNotification: utilptr.To[bool](true),
+			maxPodsToEvictTotal:              utilptr.To[uint](1),
+			maxPodsToEvictPerNode:            utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace:       utilptr.To[uint](1),
+			expectedNodeEvictions:            1,
+			expectedTotalEvictions:           1,
+			expectedError:                    nil,
+		},
+		{
+			description:                "one eviction expected without eviction failure event notification",
+			pod:                        pod1,
+			maxPodsToEvictTotal:        utilptr.To[uint](1),
+			maxPodsToEvictPerNode:      utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace: utilptr.To[uint](1),
+			expectedNodeEvictions:      1,
+			expectedTotalEvictions:     1,
+			expectedError:              nil,
+			events:                     []string{"Normal NotSet pod eviction from node node by sigs.k8s.io/descheduler"},
+		},
+		{
+			description:                "eviction limit exceeded on total without eviction failure event notification",
+			pod:                        pod1,
+			maxPodsToEvictTotal:        utilptr.To[uint](0),
+			maxPodsToEvictPerNode:      utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace: utilptr.To[uint](1),
+			expectedNodeEvictions:      0,
+			expectedTotalEvictions:     0,
+			expectedError:              NewEvictionTotalLimitError(),
+		},
+		{
+			description:                "eviction limit exceeded on node without eviction failure event notification",
+			pod:                        pod1,
+			maxPodsToEvictTotal:        utilptr.To[uint](1),
+			maxPodsToEvictPerNode:      utilptr.To[uint](0),
+			maxPodsToEvictPerNamespace: utilptr.To[uint](1),
+			expectedNodeEvictions:      0,
+			expectedTotalEvictions:     0,
+			expectedError:              NewEvictionNodeLimitError("node"),
+		},
+		{
+			description:                "eviction limit exceeded on node without eviction failure event notification",
+			pod:                        pod1,
+			maxPodsToEvictTotal:        utilptr.To[uint](1),
+			maxPodsToEvictPerNode:      utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace: utilptr.To[uint](0),
+			expectedNodeEvictions:      0,
+			expectedTotalEvictions:     0,
+			expectedError:              NewEvictionNamespaceLimitError("default"),
+		},
+		{
+			description:                "eviction error without eviction failure event notification",
+			pod:                        pod1,
+			maxPodsToEvictTotal:        utilptr.To[uint](1),
+			maxPodsToEvictPerNode:      utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace: utilptr.To[uint](1),
+			expectedNodeEvictions:      0,
+			expectedTotalEvictions:     0,
+			expectedError:              fmt.Errorf("eviction error"),
+		},
+		{
+			description:                "eviction without dryRun with eviction failure event notification",
+			pod:                        pod1,
+			dryRun:                     true,
+			maxPodsToEvictTotal:        utilptr.To[uint](1),
+			maxPodsToEvictPerNode:      utilptr.To[uint](1),
+			maxPodsToEvictPerNamespace: utilptr.To[uint](1),
+			expectedNodeEvictions:      1,
+			expectedTotalEvictions:     1,
+			expectedError:              nil,
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.description, func(t *testing.T) {
+			fakeClient := fake.NewSimpleClientset(pod1)
+			fakeClient.PrependReactor("create", "pods/eviction", func(action core.Action) (handled bool, ret runtime.Object, err error) {
+				return true, nil, test.expectedError
+			})
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			sharedInformerFactory.Start(ctx.Done())
+			sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+			eventRecorder := events.NewFakeRecorder(100)
+
+			podEvictor, err := NewPodEvictor(
+				ctx,
+				fakeClient,
+				eventRecorder,
+				sharedInformerFactory.Core().V1().Pods().Informer(),
+				initFeatureGates(),
+				NewOptions().
+					WithDryRun(test.dryRun).
+					WithMaxPodsToEvictTotal(test.maxPodsToEvictTotal).
+					WithMaxPodsToEvictPerNode(test.maxPodsToEvictPerNode).
+					WithEvictionFailureEventNotification(test.evictionFailureEventNotification).
+					WithMaxPodsToEvictPerNamespace(test.maxPodsToEvictPerNamespace),
+			)
+			if err != nil {
+				t.Fatalf("Unexpected error when creating a pod evictor: %v", err)
+			}
+
+			stubNode := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "node"}}
+
+			if actualErr := podEvictor.EvictPod(ctx, test.pod, EvictOptions{}); actualErr != nil && actualErr.Error() != test.expectedError.Error() {
+				t.Errorf("Expected error: %v, got: %v", test.expectedError, actualErr)
+			}
+
+			if evictions := podEvictor.NodeEvicted(stubNode); evictions != test.expectedNodeEvictions {
+				t.Errorf("Expected %d node evictions, got %d instead", test.expectedNodeEvictions, evictions)
+			}
+
+			if evictions := podEvictor.TotalEvicted(); evictions != test.expectedTotalEvictions {
+				t.Errorf("Expected %d total evictions, got %d instead", test.expectedTotalEvictions, evictions)
+			}
+
+			// Assert that the events are correct.
+			assertEqualEvents(t, test.events, eventRecorder.Events)
+		})
+	}
+}
+
+func TestEvictionRequestsCacheCleanup(t *testing.T) {
+	ctx := context.Background()
+	node1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+
+	ownerRef1 := test.GetReplicaSetOwnerRefList()
+	updatePod := func(pod *v1.Pod) {
+		pod.Namespace = "dev"
+		pod.ObjectMeta.OwnerReferences = ownerRef1
+	}
+	updatePodWithEvictionInBackground := func(pod *v1.Pod) {
+		updatePod(pod)
+		pod.Annotations = map[string]string{
+			EvictionRequestAnnotationKey: "",
+		}
+	}
+
+	p1 := test.BuildTestPod("p1", 100, 0, node1.Name, updatePodWithEvictionInBackground)
+	p2 := test.BuildTestPod("p2", 100, 0, node1.Name, updatePodWithEvictionInBackground)
+	p3 := test.BuildTestPod("p3", 100, 0, node1.Name, updatePod)
+	p4 := test.BuildTestPod("p4", 100, 0, node1.Name, updatePod)
+
+	client := fakeclientset.NewSimpleClientset(node1, p1, p2, p3, p4)
+	sharedInformerFactory := informers.NewSharedInformerFactory(client, 0)
+	_, eventRecorder := utils.GetRecorderAndBroadcaster(ctx, client)
+
+	podEvictor, err := NewPodEvictor(
+		ctx,
+		client,
+		eventRecorder,
+		sharedInformerFactory.Core().V1().Pods().Informer(),
+		initFeatureGates(),
+		nil,
+	)
+	if err != nil {
+		t.Fatalf("Unexpected error when creating a pod evictor: %v", err)
+	}
+
+	client.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")
+			}
+			if eviction, matched := createAct.Object.(*policy.Eviction); matched {
+				podName := eviction.GetName()
+				if podName == "p1" || podName == "p2" {
+					return true, nil, &apierrors.StatusError{
+						ErrStatus: metav1.Status{
+							Reason:  metav1.StatusReasonTooManyRequests,
+							Message: "Eviction triggered evacuation",
+						},
+					}
+				}
+				return true, nil, nil
+			}
+		}
+		return false, nil, nil
+	})
+
+	sharedInformerFactory.Start(ctx.Done())
+	sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+	podEvictor.EvictPod(ctx, p1, EvictOptions{})
+	podEvictor.EvictPod(ctx, p2, EvictOptions{})
+	podEvictor.EvictPod(ctx, p3, EvictOptions{})
+	podEvictor.EvictPod(ctx, p4, EvictOptions{})
+
+	klog.Infof("2 evictions in background expected, 2 normal evictions")
+	if total := podEvictor.TotalEvictionRequests(); total != 2 {
+		t.Fatalf("Expected %v total eviction requests, got %v instead", 2, total)
+	}
+	if total := podEvictor.TotalEvicted(); total != 2 {
+		t.Fatalf("Expected %v total evictions, got %v instead", 2, total)
+	}
+
+	klog.Infof("2 evictions in background assumed. Wait for few seconds and check the assumed requests timed out")
+	time.Sleep(2 * time.Second)
+	klog.Infof("Checking the assumed requests timed out and were deleted")
+	// Set the timeout to 1s so the cleaning can be tested
+	podEvictor.erCache.assumedRequestTimeoutSeconds = 1
+	podEvictor.erCache.cleanCache(ctx)
+	if totalERs := podEvictor.TotalEvictionRequests(); totalERs > 0 {
+		t.Fatalf("Expected 0 eviction requests, got %v instead", totalERs)
+	}
+}
+
+func assertEqualEvents(t *testing.T, expected []string, actual <-chan string) {
+	t.Logf("Assert for events: %v", expected)
+	c := time.After(wait.ForeverTestTimeout)
+	for _, e := range expected {
+		select {
+		case a := <-actual:
+			if !reflect.DeepEqual(a, e) {
+				t.Errorf("Expected event %q, got %q instead", e, a)
+			}
+		case <-c:
+			t.Errorf("Expected event %q, got nothing", e)
+			// continue iterating to print all expected events
+		}
+	}
+	for {
+		select {
+		case a := <-actual:
+			t.Errorf("Unexpected event: %q", a)
+		default:
+			return // No more events, as expected.
+		}
+	}
+}

pkg/descheduler/evictions/options.go

@@ -0,0 +1,59 @@
+package evictions
+
+import (
+	policy "k8s.io/api/policy/v1"
+)
+
+type Options struct {
+	policyGroupVersion               string
+	dryRun                           bool
+	maxPodsToEvictPerNode            *uint
+	maxPodsToEvictPerNamespace       *uint
+	maxPodsToEvictTotal              *uint
+	evictionFailureEventNotification bool
+	metricsEnabled                   bool
+}
+
+// NewOptions returns an Options with default values.
+func NewOptions() *Options {
+	return &Options{
+		policyGroupVersion: policy.SchemeGroupVersion.String(),
+	}
+}
+
+func (o *Options) WithPolicyGroupVersion(policyGroupVersion string) *Options {
+	o.policyGroupVersion = policyGroupVersion
+	return o
+}
+
+func (o *Options) WithDryRun(dryRun bool) *Options {
+	o.dryRun = dryRun
+	return o
+}
+
+func (o *Options) WithMaxPodsToEvictPerNode(maxPodsToEvictPerNode *uint) *Options {
+	o.maxPodsToEvictPerNode = maxPodsToEvictPerNode
+	return o
+}
+
+func (o *Options) WithMaxPodsToEvictPerNamespace(maxPodsToEvictPerNamespace *uint) *Options {
+	o.maxPodsToEvictPerNamespace = maxPodsToEvictPerNamespace
+	return o
+}
+
+func (o *Options) WithMaxPodsToEvictTotal(maxPodsToEvictTotal *uint) *Options {
+	o.maxPodsToEvictTotal = maxPodsToEvictTotal
+	return o
+}
+
+func (o *Options) WithMetricsEnabled(metricsEnabled bool) *Options {
+	o.metricsEnabled = metricsEnabled
+	return o
+}
+
+func (o *Options) WithEvictionFailureEventNotification(evictionFailureEventNotification *bool) *Options {
+	if evictionFailureEventNotification != nil {
+		o.evictionFailureEventNotification = *evictionFailureEventNotification
+	}
+	return o
+}

pkg/descheduler/metricscollector/metricscollector.go

@@ -0,0 +1,151 @@
+/*
+Copyright 2024 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 metricscollector
+
+import (
+	"context"
+	"fmt"
+	"math"
+	"sync"
+	"time"
+
+	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"
+	listercorev1 "k8s.io/client-go/listers/core/v1"
+	"k8s.io/klog/v2"
+	metricsclient "k8s.io/metrics/pkg/client/clientset/versioned"
+	utilptr "k8s.io/utils/ptr"
+)
+
+const (
+	beta float64 = 0.9
+)
+
+type MetricsCollector struct {
+	nodeLister       listercorev1.NodeLister
+	metricsClientset metricsclient.Interface
+	nodeSelector     labels.Selector
+
+	nodes map[string]map[v1.ResourceName]*resource.Quantity
+
+	mu sync.RWMutex
+	// hasSynced signals at least one sync succeeded
+	hasSynced bool
+}
+
+func NewMetricsCollector(nodeLister listercorev1.NodeLister, metricsClientset metricsclient.Interface, nodeSelector labels.Selector) *MetricsCollector {
+	return &MetricsCollector{
+		nodeLister:       nodeLister,
+		metricsClientset: metricsClientset,
+		nodeSelector:     nodeSelector,
+		nodes:            make(map[string]map[v1.ResourceName]*resource.Quantity),
+	}
+}
+
+func (mc *MetricsCollector) Run(ctx context.Context) {
+	wait.NonSlidingUntil(func() {
+		mc.Collect(ctx)
+	}, 5*time.Second, ctx.Done())
+}
+
+// During experiments rounding to int error causes weightedAverage to never
+// reach value even when weightedAverage is repeated many times in a row.
+// The difference between the limit and computed average stops within 5 units.
+// Nevertheless, the value is expected to change in time. So the weighted
+// average nevers gets a chance to converge. Which makes the computed
+// error negligible.
+// The speed of convergence depends on how often the metrics collector
+// syncs with the current value. Currently, the interval is set to 5s.
+func weightedAverage(prevValue, value int64) int64 {
+	return int64(math.Round(beta*float64(prevValue) + (1-beta)*float64(value)))
+}
+
+func (mc *MetricsCollector) AllNodesUsage() (map[string]map[v1.ResourceName]*resource.Quantity, error) {
+	mc.mu.RLock()
+	defer mc.mu.RUnlock()
+
+	allNodesUsage := make(map[string]map[v1.ResourceName]*resource.Quantity)
+	for nodeName := range mc.nodes {
+		allNodesUsage[nodeName] = map[v1.ResourceName]*resource.Quantity{
+			v1.ResourceCPU:    utilptr.To[resource.Quantity](mc.nodes[nodeName][v1.ResourceCPU].DeepCopy()),
+			v1.ResourceMemory: utilptr.To[resource.Quantity](mc.nodes[nodeName][v1.ResourceMemory].DeepCopy()),
+		}
+	}
+
+	return allNodesUsage, nil
+}
+
+func (mc *MetricsCollector) NodeUsage(node *v1.Node) (map[v1.ResourceName]*resource.Quantity, error) {
+	mc.mu.RLock()
+	defer mc.mu.RUnlock()
+
+	if _, exists := mc.nodes[node.Name]; !exists {
+		klog.V(4).InfoS("unable to find node in the collected metrics", "node", klog.KObj(node))
+		return nil, fmt.Errorf("unable to find node %q in the collected metrics", node.Name)
+	}
+	return map[v1.ResourceName]*resource.Quantity{
+		v1.ResourceCPU:    utilptr.To[resource.Quantity](mc.nodes[node.Name][v1.ResourceCPU].DeepCopy()),
+		v1.ResourceMemory: utilptr.To[resource.Quantity](mc.nodes[node.Name][v1.ResourceMemory].DeepCopy()),
+	}, nil
+}
+
+func (mc *MetricsCollector) HasSynced() bool {
+	return mc.hasSynced
+}
+
+func (mc *MetricsCollector) MetricsClient() metricsclient.Interface {
+	return mc.metricsClientset
+}
+
+func (mc *MetricsCollector) Collect(ctx context.Context) error {
+	mc.mu.Lock()
+	defer mc.mu.Unlock()
+	nodes, err := mc.nodeLister.List(mc.nodeSelector)
+	if err != nil {
+		return fmt.Errorf("unable to list nodes: %v", err)
+	}
+
+	for _, node := range nodes {
+		metrics, err := mc.metricsClientset.MetricsV1beta1().NodeMetricses().Get(ctx, node.Name, metav1.GetOptions{})
+		if err != nil {
+			klog.ErrorS(err, "Error fetching metrics", "node", node.Name)
+			// No entry -> duplicate the previous value -> do nothing as beta*PV + (1-beta)*PV = PV
+			continue
+		}
+
+		if _, exists := mc.nodes[node.Name]; !exists {
+			mc.nodes[node.Name] = map[v1.ResourceName]*resource.Quantity{
+				v1.ResourceCPU:    utilptr.To[resource.Quantity](metrics.Usage.Cpu().DeepCopy()),
+				v1.ResourceMemory: utilptr.To[resource.Quantity](metrics.Usage.Memory().DeepCopy()),
+			}
+		} else {
+			// get MilliValue to reduce loss of precision
+			mc.nodes[node.Name][v1.ResourceCPU].SetMilli(
+				weightedAverage(mc.nodes[node.Name][v1.ResourceCPU].MilliValue(), metrics.Usage.Cpu().MilliValue()),
+			)
+			mc.nodes[node.Name][v1.ResourceMemory].Set(
+				weightedAverage(mc.nodes[node.Name][v1.ResourceMemory].Value(), metrics.Usage.Memory().Value()),
+			)
+		}
+	}
+
+	mc.hasSynced = true
+	return nil
+}

pkg/descheduler/metricscollector/metricscollector_test.go

@@ -0,0 +1,141 @@
+/*
+Copyright 2024 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 metricscollector
+
+import (
+	"context"
+	"math"
+	"testing"
+
+	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	"k8s.io/client-go/informers"
+	fakeclientset "k8s.io/client-go/kubernetes/fake"
+	fakemetricsclient "k8s.io/metrics/pkg/client/clientset/versioned/fake"
+
+	"sigs.k8s.io/descheduler/test"
+)
+
+func checkCpuNodeUsage(t *testing.T, usage map[v1.ResourceName]*resource.Quantity, millicpu int64) {
+	t.Logf("current node cpu usage: %v\n", usage[v1.ResourceCPU].MilliValue())
+	if usage[v1.ResourceCPU].MilliValue() != millicpu {
+		t.Fatalf("cpu node usage expected to be %v, got %v instead", millicpu, usage[v1.ResourceCPU].MilliValue())
+	}
+}
+
+func TestMetricsCollector(t *testing.T) {
+	gvr := schema.GroupVersionResource{Group: "metrics.k8s.io", Version: "v1beta1", Resource: "nodes"}
+
+	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	n3 := test.BuildTestNode("n3", 2000, 3000, 10, nil)
+
+	n1metrics := test.BuildNodeMetrics("n1", 400, 1714978816)
+	n2metrics := test.BuildNodeMetrics("n2", 1400, 1714978816)
+	n3metrics := test.BuildNodeMetrics("n3", 300, 1714978816)
+
+	clientset := fakeclientset.NewSimpleClientset(n1, n2, n3)
+	metricsClientset := fakemetricsclient.NewSimpleClientset()
+	metricsClientset.Tracker().Create(gvr, n1metrics, "")
+	metricsClientset.Tracker().Create(gvr, n2metrics, "")
+	metricsClientset.Tracker().Create(gvr, n3metrics, "")
+
+	ctx := context.TODO()
+	sharedInformerFactory := informers.NewSharedInformerFactory(clientset, 0)
+	nodeLister := sharedInformerFactory.Core().V1().Nodes().Lister()
+	sharedInformerFactory.Start(ctx.Done())
+	sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+	t.Logf("Set initial node cpu usage to 1400")
+	collector := NewMetricsCollector(nodeLister, metricsClientset, labels.Everything())
+	collector.Collect(context.TODO())
+	nodesUsage, _ := collector.NodeUsage(n2)
+	checkCpuNodeUsage(t, nodesUsage, 1400)
+	allnodesUsage, _ := collector.AllNodesUsage()
+	checkCpuNodeUsage(t, allnodesUsage[n2.Name], 1400)
+
+	t.Logf("Set current node cpu usage to 500")
+	n2metrics.Usage[v1.ResourceCPU] = *resource.NewMilliQuantity(500, resource.DecimalSI)
+	metricsClientset.Tracker().Update(gvr, n2metrics, "")
+	collector.Collect(context.TODO())
+	nodesUsage, _ = collector.NodeUsage(n2)
+	checkCpuNodeUsage(t, nodesUsage, 1310)
+	allnodesUsage, _ = collector.AllNodesUsage()
+	checkCpuNodeUsage(t, allnodesUsage[n2.Name], 1310)
+
+	t.Logf("Set current node cpu usage to 900")
+	n2metrics.Usage[v1.ResourceCPU] = *resource.NewMilliQuantity(900, resource.DecimalSI)
+	metricsClientset.Tracker().Update(gvr, n2metrics, "")
+	collector.Collect(context.TODO())
+	nodesUsage, _ = collector.NodeUsage(n2)
+	checkCpuNodeUsage(t, nodesUsage, 1269)
+	allnodesUsage, _ = collector.AllNodesUsage()
+	checkCpuNodeUsage(t, allnodesUsage[n2.Name], 1269)
+}
+
+func TestMetricsCollectorConvergence(t *testing.T) {
+	gvr := schema.GroupVersionResource{Group: "metrics.k8s.io", Version: "v1beta1", Resource: "nodes"}
+
+	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
+	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
+	n3 := test.BuildTestNode("n3", 2000, 3000, 10, nil)
+
+	n1metrics := test.BuildNodeMetrics("n1", 400, 1714978816)
+	n2metrics := test.BuildNodeMetrics("n2", 1400, 1714978816)
+	n3metrics := test.BuildNodeMetrics("n3", 300, 1714978816)
+
+	clientset := fakeclientset.NewSimpleClientset(n1, n2, n3)
+	metricsClientset := fakemetricsclient.NewSimpleClientset()
+	metricsClientset.Tracker().Create(gvr, n1metrics, "")
+	metricsClientset.Tracker().Create(gvr, n2metrics, "")
+	metricsClientset.Tracker().Create(gvr, n3metrics, "")
+
+	ctx := context.TODO()
+	sharedInformerFactory := informers.NewSharedInformerFactory(clientset, 0)
+	nodeLister := sharedInformerFactory.Core().V1().Nodes().Lister()
+	sharedInformerFactory.Start(ctx.Done())
+	sharedInformerFactory.WaitForCacheSync(ctx.Done())
+
+	t.Logf("Set initial node cpu usage to 1400")
+	collector := NewMetricsCollector(nodeLister, metricsClientset, labels.Everything())
+	collector.Collect(context.TODO())
+	nodesUsage, _ := collector.NodeUsage(n2)
+	checkCpuNodeUsage(t, nodesUsage, 1400)
+	allnodesUsage, _ := collector.AllNodesUsage()
+	checkCpuNodeUsage(t, allnodesUsage[n2.Name], 1400)
+
+	t.Logf("Set current node cpu/memory usage to 900/1614978816 and wait until it converges to it")
+	n2metrics.Usage[v1.ResourceCPU] = *resource.NewMilliQuantity(900, resource.DecimalSI)
+	n2metrics.Usage[v1.ResourceMemory] = *resource.NewQuantity(1614978816, resource.BinarySI)
+	metricsClientset.Tracker().Update(gvr, n2metrics, "")
+	converged := false
+	for i := 0; i < 300; i++ {
+		collector.Collect(context.TODO())
+		nodesUsage, _ = collector.NodeUsage(n2)
+		if math.Abs(float64(900-nodesUsage[v1.ResourceCPU].MilliValue())) < 6 && math.Abs(float64(1614978816-nodesUsage[v1.ResourceMemory].Value())) < 6 {
+			t.Logf("Node cpu/memory usage converged to 900+-5/1614978816+-5")
+			converged = true
+			break
+		}
+		t.Logf("The current node usage: cpu=%v, memory=%v", nodesUsage[v1.ResourceCPU].MilliValue(), nodesUsage[v1.ResourceMemory].Value())
+	}
+	if !converged {
+		t.Fatalf("The node usage did not converged to 900+-1")
+	}
+}

pkg/descheduler/node/node.go

@@ -18,7 +18,9 @@ package node
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"sync/atomic"
 
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
@@ -26,11 +28,14 @@ import (
 	"k8s.io/apimachinery/pkg/labels"
 	clientset "k8s.io/client-go/kubernetes"
 	listersv1 "k8s.io/client-go/listers/core/v1"
+	"k8s.io/client-go/util/workqueue"
 	"k8s.io/klog/v2"
 	podutil "sigs.k8s.io/descheduler/pkg/descheduler/pod"
 	"sigs.k8s.io/descheduler/pkg/utils"
 )
 
+const workersCount = 100
+
 // ReadyNodes returns ready nodes irrespective of whether they are
 // schedulable or not.
 func ReadyNodes(ctx context.Context, client clientset.Interface, nodeLister listersv1.NodeLister, nodeSelector string) ([]*v1.Node, error) {
@@ -103,90 +108,97 @@ func IsReady(node *v1.Node) bool {
 // This function is used when the NodeFit pod filtering feature of the Descheduler is enabled.
 // This function currently considers a subset of the Kubernetes Scheduler's predicates when
 // deciding if a pod would fit on a node, but more predicates may be added in the future.
-func NodeFit(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) []error {
+// There should be no methods to modify nodes or pods in this method.
+func NodeFit(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) error {
 	// Check node selector and required affinity
-	var errors []error
 	if ok, err := utils.PodMatchNodeSelector(pod, node); err != nil {
-		errors = append(errors, err)
+		return err
 	} else if !ok {
-		errors = append(errors, fmt.Errorf("pod node selector does not match the node label"))
+		return errors.New("pod node selector does not match the node label")
 	}
+
 	// Check taints (we only care about NoSchedule and NoExecute taints)
 	ok := utils.TolerationsTolerateTaintsWithFilter(pod.Spec.Tolerations, node.Spec.Taints, func(taint *v1.Taint) bool {
 		return taint.Effect == v1.TaintEffectNoSchedule || taint.Effect == v1.TaintEffectNoExecute
 	})
 	if !ok {
-		errors = append(errors, fmt.Errorf("pod does not tolerate taints on the node"))
+		return errors.New("pod does not tolerate taints on the node")
-	}
-	// Check if the pod can fit on a node based off it's requests
-	if pod.Spec.NodeName == "" || pod.Spec.NodeName != node.Name {
-		ok, reqErrors := fitsRequest(nodeIndexer, pod, node)
-		if !ok {
-			errors = append(errors, reqErrors...)
-		}
-	}
-	// Check if node is schedulable
-	if IsNodeUnschedulable(node) {
-		errors = append(errors, fmt.Errorf("node is not schedulable"))
 	}
 
-	return errors
+	// Check if the pod can fit on a node based off it's requests
+	if pod.Spec.NodeName == "" || pod.Spec.NodeName != node.Name {
+		if ok, reqError := fitsRequest(nodeIndexer, pod, node); !ok {
+			return reqError
+		}
+	}
+
+	// Check if node is schedulable
+	if IsNodeUnschedulable(node) {
+		return errors.New("node is not schedulable")
+	}
+
+	// Check if pod matches inter-pod anti-affinity rule of pod on node
+	if match, err := podMatchesInterPodAntiAffinity(nodeIndexer, pod, node); err != nil {
+		return err
+	} else if match {
+		return errors.New("pod matches inter-pod anti-affinity rule of other pod on node")
+	}
+
+	return nil
+}
+
+func podFitsNodes(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, nodes []*v1.Node, excludeFilter func(pod *v1.Pod, node *v1.Node) bool) bool {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	var filteredLen int32
+	checkNode := func(i int) {
+		node := nodes[i]
+		if excludeFilter != nil && excludeFilter(pod, node) {
+			return
+		}
+		err := NodeFit(nodeIndexer, pod, node)
+		if err == nil {
+			klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
+			atomic.AddInt32(&filteredLen, 1)
+			cancel()
+		} else {
+			klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node), "err", err.Error())
+		}
+	}
+
+	// Stops searching for more nodes once a node are found.
+	workqueue.ParallelizeUntil(ctx, workersCount, len(nodes), checkNode)
+
+	return filteredLen > 0
 }
 
 // PodFitsAnyOtherNode checks if the given pod will fit any of the given nodes, besides the node
 // the pod is already running on. The predicates used to determine if the pod will fit can be found in the NodeFit function.
 func PodFitsAnyOtherNode(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, nodes []*v1.Node) bool {
-	for _, node := range nodes {
+	return podFitsNodes(nodeIndexer, pod, nodes, func(pod *v1.Pod, node *v1.Node) bool {
-		// Skip node pod is already on
+		return pod.Spec.NodeName == node.Name
-		if node.Name == pod.Spec.NodeName {
+	})
-			continue
-		}
-
-		errors := NodeFit(nodeIndexer, pod, node)
-		if len(errors) == 0 {
-			klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-			return true
-		} else {
-			klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-			for _, err := range errors {
-				klog.V(4).InfoS(err.Error())
-			}
-		}
-	}
-	return false
 }
 
 // PodFitsAnyNode checks if the given pod will fit any of the given nodes. The predicates used
 // to determine if the pod will fit can be found in the NodeFit function.
 func PodFitsAnyNode(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, nodes []*v1.Node) bool {
-	for _, node := range nodes {
+	return podFitsNodes(nodeIndexer, pod, nodes, nil)
-		errors := NodeFit(nodeIndexer, pod, node)
-		if len(errors) == 0 {
-			klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-			return true
-		} else {
-			klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-			for _, err := range errors {
-				klog.V(4).InfoS(err.Error())
-			}
-		}
-	}
-	return false
 }
 
 // PodFitsCurrentNode checks if the given pod will fit onto the given node. The predicates used
 // to determine if the pod will fit can be found in the NodeFit function.
 func PodFitsCurrentNode(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) bool {
-	errors := NodeFit(nodeIndexer, pod, node)
+	err := NodeFit(nodeIndexer, pod, node)
-	if len(errors) == 0 {
+	if err == nil {
 		klog.V(4).InfoS("Pod fits on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
 		return true
-	} else {
-		klog.V(4).InfoS("Pod does not fit on node", "pod", klog.KObj(pod), "node", klog.KObj(node))
-		for _, err := range errors {
-			klog.V(4).InfoS(err.Error())
-		}
 	}
+
+	klog.V(4).InfoS("Pod does not fit on current node",
+		"pod", klog.KObj(pod), "node", klog.KObj(node), "error", err)
+
 	return false
 }
 
@@ -198,9 +210,7 @@ func IsNodeUnschedulable(node *v1.Node) bool {
 
 // fitsRequest determines if a pod can fit on a node based on its resource requests. It returns true if
 // the pod will fit.
-func fitsRequest(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) (bool, []error) {
+func fitsRequest(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) (bool, error) {
-	var insufficientResources []error
-
 	// Get pod requests
 	podRequests, _ := utils.PodRequestsAndLimits(pod)
 	resourceNames := make([]v1.ResourceName, 0, len(podRequests))
@@ -208,30 +218,41 @@ func fitsRequest(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, nod
 		resourceNames = append(resourceNames, name)
 	}
 
-	availableResources, err := nodeAvailableResources(nodeIndexer, node, resourceNames)
+	availableResources, err := nodeAvailableResources(nodeIndexer, node, resourceNames,
+		func(pod *v1.Pod) (v1.ResourceList, error) {
+			req, _ := utils.PodRequestsAndLimits(pod)
+			return req, nil
+		},
+	)
 	if err != nil {
-		return false, []error{err}
+		return false, err
 	}
 
-	podFitsOnNode := true
 	for _, resource := range resourceNames {
 		podResourceRequest := podRequests[resource]
 		availableResource, ok := availableResources[resource]
 		if !ok || podResourceRequest.MilliValue() > availableResource.MilliValue() {
-			insufficientResources = append(insufficientResources, fmt.Errorf("insufficient %v", resource))
+			return false, fmt.Errorf("insufficient %v", resource)
-			podFitsOnNode = false
 		}
 	}
-	return podFitsOnNode, insufficientResources
+	// check pod num, at least one pod number is avaibalbe
+	if availableResources[v1.ResourcePods].MilliValue() <= 0 {
+		return false, fmt.Errorf("insufficient %v", v1.ResourcePods)
+	}
+
+	return true, nil
 }
 
 // nodeAvailableResources returns resources mapped to the quanitity available on the node.
-func nodeAvailableResources(nodeIndexer podutil.GetPodsAssignedToNodeFunc, node *v1.Node, resourceNames []v1.ResourceName) (map[v1.ResourceName]*resource.Quantity, error) {
+func nodeAvailableResources(nodeIndexer podutil.GetPodsAssignedToNodeFunc, node *v1.Node, resourceNames []v1.ResourceName, podUtilization podutil.PodUtilizationFnc) (map[v1.ResourceName]*resource.Quantity, error) {
 	podsOnNode, err := podutil.ListPodsOnANode(node.Name, nodeIndexer, nil)
 	if err != nil {
 		return nil, err
 	}
-	nodeUtilization := NodeUtilization(podsOnNode, resourceNames)
+	nodeUtilization, err := NodeUtilization(podsOnNode, resourceNames, podUtilization)
+	if err != nil {
+		return nil, err
+	}
 	remainingResources := map[v1.ResourceName]*resource.Quantity{
 		v1.ResourceCPU:    resource.NewMilliQuantity(node.Status.Allocatable.Cpu().MilliValue()-nodeUtilization[v1.ResourceCPU].MilliValue(), resource.DecimalSI),
 		v1.ResourceMemory: resource.NewQuantity(node.Status.Allocatable.Memory().Value()-nodeUtilization[v1.ResourceMemory].Value(), resource.BinarySI),
@@ -252,31 +273,34 @@ func nodeAvailableResources(nodeIndexer podutil.GetPodsAssignedToNodeFunc, node
 }
 
 // NodeUtilization returns the resources requested by the given pods. Only resources supplied in the resourceNames parameter are calculated.
-func NodeUtilization(pods []*v1.Pod, resourceNames []v1.ResourceName) map[v1.ResourceName]*resource.Quantity {
+func NodeUtilization(pods []*v1.Pod, resourceNames []v1.ResourceName, podUtilization podutil.PodUtilizationFnc) (map[v1.ResourceName]*resource.Quantity, error) {
-	totalReqs := map[v1.ResourceName]*resource.Quantity{
+	totalUtilization := 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)
+			totalUtilization[name] = resource.NewQuantity(0, resource.DecimalSI)
 		}
 	}
 
 	for _, pod := range pods {
-		req, _ := utils.PodRequestsAndLimits(pod)
+		podUtil, err := podUtilization(pod)
+		if err != nil {
+			return nil, err
+		}
 		for _, name := range resourceNames {
-			quantity, ok := req[name]
+			quantity, ok := podUtil[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)
+				totalUtilization[name].Add(quantity)
 			}
 		}
 	}
 
-	return totalReqs
+	return totalUtilization, nil
 }
 
 // IsBasicResource checks if resource is basic native.
@@ -288,3 +312,77 @@ func IsBasicResource(name v1.ResourceName) bool {
 		return false
 	}
 }
+
+// GetNodeWeightGivenPodPreferredAffinity returns the weight
+// that the pod gives to a node by analyzing the soft node affinity of that pod
+// (nodeAffinity.preferredDuringSchedulingIgnoredDuringExecution)
+func GetNodeWeightGivenPodPreferredAffinity(pod *v1.Pod, node *v1.Node) int32 {
+	totalWeight, err := utils.GetNodeWeightGivenPodPreferredAffinity(pod, node)
+	if err != nil {
+		return 0
+	}
+	return totalWeight
+}
+
+// GetBestNodeWeightGivenPodPreferredAffinity returns the best weight
+// (maximum one) that the pod gives to the best node by analyzing the soft node affinity
+// of that pod (nodeAffinity.preferredDuringSchedulingIgnoredDuringExecution)
+func GetBestNodeWeightGivenPodPreferredAffinity(pod *v1.Pod, nodes []*v1.Node) int32 {
+	var bestWeight int32 = 0
+	for _, node := range nodes {
+		weight := GetNodeWeightGivenPodPreferredAffinity(pod, node)
+		if weight > bestWeight {
+			bestWeight = weight
+		}
+	}
+	return bestWeight
+}
+
+// PodMatchNodeSelector checks if a pod node selector matches the node label.
+func PodMatchNodeSelector(pod *v1.Pod, node *v1.Node) bool {
+	matches, err := utils.PodMatchNodeSelector(pod, node)
+	if err != nil {
+		return false
+	}
+	return matches
+}
+
+// podMatchesInterPodAntiAffinity checks if the pod matches the anti-affinity rule
+// of another pod that is already on the given node.
+// If a match is found, it returns true.
+func podMatchesInterPodAntiAffinity(nodeIndexer podutil.GetPodsAssignedToNodeFunc, pod *v1.Pod, node *v1.Node) (bool, error) {
+	if pod.Spec.Affinity == nil || pod.Spec.Affinity.PodAntiAffinity == nil {
+		return false, nil
+	}
+
+	podsOnNode, err := podutil.ListPodsOnANode(node.Name, nodeIndexer, nil)
+	if err != nil {
+		return false, fmt.Errorf("error listing all pods: %v", err)
+	}
+	assignedPodsInNamespace := podutil.GroupByNamespace(podsOnNode)
+
+	for _, term := range utils.GetPodAntiAffinityTerms(pod.Spec.Affinity.PodAntiAffinity) {
+		namespaces := utils.GetNamespacesFromPodAffinityTerm(pod, &term)
+		selector, err := metav1.LabelSelectorAsSelector(term.LabelSelector)
+		if err != nil {
+			klog.ErrorS(err, "Unable to convert LabelSelector into Selector")
+			return false, err
+		}
+
+		for namespace := range namespaces {
+			for _, assignedPod := range assignedPodsInNamespace[namespace] {
+				if assignedPod.Name == pod.Name || !utils.PodMatchesTermsNamespaceAndSelector(assignedPod, namespaces, selector) {
+					klog.V(4).InfoS("Pod doesn't match inter-pod anti-affinity rule of assigned pod on node", "candidatePod", klog.KObj(pod), "assignedPod", klog.KObj(assignedPod))
+					continue
+				}
+
+				if _, ok := node.Labels[term.TopologyKey]; ok {
+					klog.V(1).InfoS("Pod matches inter-pod anti-affinity rule of assigned pod on node", "candidatePod", klog.KObj(pod), "assignedPod", klog.KObj(assignedPod))
+					return true, nil
+				}
+			}
+		}
+	}
+
+	return false, nil
+}

pkg/descheduler/node/node_test.go

@@ -18,6 +18,8 @@ package node
 
 import (
 	"context"
+	"errors"
+	"sync"
 	"testing"
 
 	v1 "k8s.io/api/core/v1"
@@ -229,7 +231,7 @@ func TestPodFitsAnyOtherNode(t *testing.T) {
 	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"}
+	nodeNames := []string{"node1", "node2", "stagingNode", "node4"}
 
 	tests := []struct {
 		description string
@@ -715,6 +717,151 @@ func TestPodFitsAnyOtherNode(t *testing.T) {
 			},
 			success: false,
 		},
+		{
+			description: "There are four nodes. One node has a taint, and the other three nodes do not meet the resource requirements, should fail",
+			pod: test.BuildTestPod("p1", 1000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode(nodeNames[1], 3000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 3000, 8*1000*1000*1000, 12, nil),
+				test.BuildTestNode(nodeNames[3], 0, 0, 0, nil),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("3-core-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+					pod.Spec.Overhead = createResourceList(1000, 1000*1000*1000, 1000*1000*1000)
+				}),
+			},
+			success: false,
+		},
+		{
+			description: "There are four nodes. First node has a taint, second node has no label, third node do not meet the resource requirements, just fourth node meets the requirements, should success",
+			pod: test.BuildTestPod("p1", 1000, 2*1000*1000*1000, nodeNames[2], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode(nodeNames[1], 8000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 1000, 8*1000*1000*1000, 12, nil),
+				test.BuildTestNode(nodeNames[3], 8000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("3-core-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+					pod.Spec.Overhead = createResourceList(1000, 1000*1000*1000, 1000*1000*1000)
+				}),
+			},
+			success: true,
+		},
+		{
+			description: "There are four nodes. First node has a taint, second node has no label, third node do not meet the resource requirements, fourth node is the one where the pod is located, should fail",
+			pod: test.BuildTestPod("p1", 1000, 2*1000*1000*1000, nodeNames[3], func(pod *v1.Pod) {
+				pod.Spec.NodeSelector = map[string]string{
+					nodeLabelKey: nodeLabelValue,
+				}
+				pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+			}),
+			nodes: []*v1.Node{
+				test.BuildTestNode(nodeNames[0], 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+					node.Spec.Taints = []v1.Taint{
+						{
+							Key:    nodeTaintKey,
+							Value:  nodeTaintValue,
+							Effect: v1.TaintEffectNoSchedule,
+						},
+					}
+				}),
+				test.BuildTestNode(nodeNames[1], 8000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(200*1000*1000*1000, resource.DecimalSI)
+				}),
+				test.BuildTestNode(nodeNames[2], 1000, 8*1000*1000*1000, 12, nil),
+				test.BuildTestNode(nodeNames[3], 8000, 8*1000*1000*1000, 12, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						nodeLabelKey: nodeLabelValue,
+					}
+
+					node.Status.Allocatable[v1.ResourceEphemeralStorage] = *resource.NewQuantity(1000*1000*1000*1000, resource.DecimalSI)
+				}),
+			},
+			podsOnNodes: []*v1.Pod{
+				test.BuildTestPod("3-core-pod", 2000, 4*1000*1000*1000, nodeNames[1], func(pod *v1.Pod) {
+					pod.ObjectMeta = metav1.ObjectMeta{
+						Namespace: "test",
+						Labels: map[string]string{
+							"test": "true",
+						},
+					}
+					pod.Spec.Containers[0].Resources.Requests[v1.ResourceEphemeralStorage] = *resource.NewQuantity(10*1000*1000*1000, resource.DecimalSI)
+					pod.Spec.Overhead = createResourceList(1000, 1000*1000*1000, 1000*1000*1000)
+				}),
+			},
+			success: false,
+		},
 	}
 
 	for _, tc := range tests {
@@ -752,6 +899,230 @@ func TestPodFitsAnyOtherNode(t *testing.T) {
 	}
 }
 
+func TestPodFitsNodes(t *testing.T) {
+	nodeNames := []string{"node1", "node2", "node3", "node4"}
+	pod := test.BuildTestPod("p1", 950, 2*1000*1000*1000, nodeNames[0], nil)
+	nodes := []*v1.Node{
+		test.BuildTestNode(nodeNames[0], 1000, 8*1000*1000*1000, 12, nil),
+		test.BuildTestNode(nodeNames[1], 200, 8*1000*1000*1000, 12, nil),
+		test.BuildTestNode(nodeNames[2], 300, 8*1000*1000*1000, 12, nil),
+		test.BuildTestNode(nodeNames[3], 400, 8*1000*1000*1000, 12, nil),
+	}
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	var objs []runtime.Object
+	for _, node := range nodes {
+		objs = append(objs, node)
+	}
+	objs = append(objs, pod)
+
+	fakeClient := fake.NewSimpleClientset(objs...)
+
+	sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+	podInformer := sharedInformerFactory.Core().V1().Pods().Informer()
+
+	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())
+
+	var nodesTraversed sync.Map
+	podFitsNodes(getPodsAssignedToNode, pod, nodes, func(pod *v1.Pod, node *v1.Node) bool {
+		nodesTraversed.Store(node.Name, node)
+		return true
+	})
+
+	for _, node := range nodes {
+		if _, exists := nodesTraversed.Load(node.Name); !exists {
+			t.Errorf("Node %v was not proccesed", node.Name)
+		}
+	}
+}
+
+func TestNodeFit(t *testing.T) {
+	node := test.BuildTestNode("node", 64000, 128*1000*1000*1000, 2, func(node *v1.Node) {
+		node.ObjectMeta.Labels = map[string]string{
+			"region": "main-region",
+		}
+	})
+
+	nodeNolabel := test.BuildTestNode("node", 64000, 128*1000*1000*1000, 2, nil)
+
+	tests := []struct {
+		description string
+		pod         *v1.Pod
+		node        *v1.Node
+		podsOnNode  []*v1.Pod
+		err         error
+	}{
+		{
+			description: "Insufficient cpu",
+			pod:         test.BuildTestPod("p1", 10000, 2*1000*1000*1000, "", nil),
+			node:        node,
+			podsOnNode: []*v1.Pod{
+				test.BuildTestPod("p2", 60000, 60*1000*1000*1000, "node", nil),
+			},
+			err: errors.New("insufficient cpu"),
+		},
+		{
+			description: "Insufficient pod num",
+			pod:         test.BuildTestPod("p1", 1000, 2*1000*1000*1000, "", nil),
+			node:        node,
+			podsOnNode: []*v1.Pod{
+				test.BuildTestPod("p2", 1000, 2*1000*1000*1000, "node", nil),
+				test.BuildTestPod("p3", 1000, 2*1000*1000*1000, "node", nil),
+			},
+			err: errors.New("insufficient pods"),
+		},
+		{
+			description: "Pod matches inter-pod anti-affinity rule of other pod on node",
+			pod:         test.PodWithPodAntiAffinity(test.BuildTestPod("p1", 1000, 1000, node.Name, nil), "foo", "bar"),
+			node:        node,
+			podsOnNode: []*v1.Pod{
+				test.PodWithPodAntiAffinity(test.BuildTestPod("p2", 1000, 1000, node.Name, nil), "foo", "bar"),
+			},
+			err: errors.New("pod matches inter-pod anti-affinity rule of other pod on node"),
+		},
+		{
+			description: "Pod doesn't match inter-pod anti-affinity rule of other pod on node, because pod and other pod is not same namespace",
+			pod:         test.PodWithPodAntiAffinity(test.BuildTestPod("p1", 1000, 1000, node.Name, nil), "foo", "bar"),
+			node:        node,
+			podsOnNode: []*v1.Pod{
+				test.PodWithPodAntiAffinity(test.BuildTestPod("p2", 1000, 1000, node.Name, func(pod *v1.Pod) {
+					pod.Namespace = "test"
+				}), "foo", "bar"),
+			},
+		},
+		{
+			description: "Pod doesn't match inter-pod anti-affinity rule of other pod on node, because other pod not match labels of pod",
+			pod:         test.PodWithPodAntiAffinity(test.BuildTestPod("p1", 1000, 1000, node.Name, nil), "foo", "bar"),
+			node:        node,
+			podsOnNode: []*v1.Pod{
+				test.PodWithPodAntiAffinity(test.BuildTestPod("p2", 1000, 1000, node.Name, nil), "foo1", "bar1"),
+			},
+		},
+		{
+			description: "Pod doesn't match inter-pod anti-affinity rule of other pod on node, because node have no topologyKey",
+			pod:         test.PodWithPodAntiAffinity(test.BuildTestPod("p1", 1000, 1000, "node1", nil), "foo", "bar"),
+			node:        nodeNolabel,
+			podsOnNode: []*v1.Pod{
+				test.PodWithPodAntiAffinity(test.BuildTestPod("p2", 1000, 1000, node.Name, nil), "foo", "bar"),
+			},
+		},
+		{
+			description: "Pod fits on node",
+			pod:         test.BuildTestPod("p1", 1000, 1000, "", func(pod *v1.Pod) {}),
+			node:        node,
+			podsOnNode:  []*v1.Pod{},
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+			objs := []runtime.Object{tc.node, tc.pod}
+			for _, pod := range tc.podsOnNode {
+				objs = append(objs, pod)
+			}
+
+			fakeClient := fake.NewSimpleClientset(objs...)
+
+			sharedInformerFactory := informers.NewSharedInformerFactory(fakeClient, 0)
+			podInformer := sharedInformerFactory.Core().V1().Pods().Informer()
+
+			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())
+			err = NodeFit(getPodsAssignedToNode, tc.pod, tc.node)
+			if (err == nil && tc.err != nil) || (err != nil && err.Error() != tc.err.Error()) {
+				t.Errorf("Test %#v failed, got %v, expect %v", tc.description, err, tc.err)
+			}
+		})
+	}
+}
+
+func TestBestPodNodeAffinityWeight(t *testing.T) {
+	defaultPod := test.BuildTestPod("p1", 0, 0, "node1", func(p *v1.Pod) {
+		p.Spec.Affinity = &v1.Affinity{
+			NodeAffinity: &v1.NodeAffinity{
+				PreferredDuringSchedulingIgnoredDuringExecution: []v1.PreferredSchedulingTerm{
+					{
+						Weight: 10,
+						Preference: v1.NodeSelectorTerm{
+							MatchExpressions: []v1.NodeSelectorRequirement{
+								{
+									Key:      "key1",
+									Operator: "In",
+									Values:   []string{"value1"},
+								},
+							},
+						},
+					},
+				},
+			},
+		}
+	})
+	tests := []struct {
+		description    string
+		pod            *v1.Pod
+		nodes          []*v1.Node
+		expectedWeight int32
+	}{
+		{
+			description: "No node matches the preferred affinity",
+			pod:         defaultPod,
+			nodes: []*v1.Node{
+				test.BuildTestNode("node2", 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						"key2": "value2",
+					}
+				}),
+				test.BuildTestNode("node3", 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						"key3": "value3",
+					}
+				}),
+			},
+			expectedWeight: 0,
+		},
+		{
+			description: "A single node matches the preferred affinity",
+			pod:         defaultPod,
+			nodes: []*v1.Node{
+				test.BuildTestNode("node1", 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						"key1": "value1",
+					}
+				}),
+				test.BuildTestNode("node2", 64000, 128*1000*1000*1000, 200, func(node *v1.Node) {
+					node.ObjectMeta.Labels = map[string]string{
+						"key2": "value2",
+					}
+				}),
+			},
+			expectedWeight: 10,
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.description, func(t *testing.T) {
+			bestWeight := GetBestNodeWeightGivenPodPreferredAffinity(tc.pod, tc.nodes)
+			if bestWeight != tc.expectedWeight {
+				t.Errorf("Test %#v failed", tc.description)
+			}
+		})
+	}
+}
+
 // createResourceList builds a small resource list of core resources
 func createResourceList(cpu, memory, ephemeralStorage int64) v1.ResourceList {
 	resourceList := make(map[v1.ResourceName]resource.Quantity)

pkg/descheduler/pod/pods.go

@@ -39,6 +39,9 @@ type FilterFunc func(*v1.Pod) bool
 // as input and returns the pods that assigned to the node.
 type GetPodsAssignedToNodeFunc func(string, FilterFunc) ([]*v1.Pod, error)
 
+// PodUtilizationFnc is a function for getting pod's utilization. E.g. requested resources of utilization from metrics.
+type PodUtilizationFnc func(pod *v1.Pod) (v1.ResourceList, error)
+
 // WrapFilterFuncs wraps a set of FilterFunc in one.
 func WrapFilterFuncs(filters ...FilterFunc) FilterFunc {
 	return func(pod *v1.Pod) bool {
@@ -53,8 +56,8 @@ func WrapFilterFuncs(filters ...FilterFunc) FilterFunc {
 
 type Options struct {
 	filter             FilterFunc
-	includedNamespaces sets.String
+	includedNamespaces sets.Set[string]
-	excludedNamespaces sets.String
+	excludedNamespaces sets.Set[string]
 	labelSelector      *metav1.LabelSelector
 }
 
@@ -71,13 +74,13 @@ func (o *Options) WithFilter(filter FilterFunc) *Options {
 }
 
 // WithNamespaces sets included namespaces
-func (o *Options) WithNamespaces(namespaces sets.String) *Options {
+func (o *Options) WithNamespaces(namespaces sets.Set[string]) *Options {
 	o.includedNamespaces = namespaces
 	return o
 }
 
 // WithoutNamespaces sets excluded namespaces
-func (o *Options) WithoutNamespaces(namespaces sets.String) *Options {
+func (o *Options) WithoutNamespaces(namespaces sets.Set[string]) *Options {
 	o.excludedNamespaces = namespaces
 	return o
 }
@@ -99,9 +102,6 @@ func (o *Options) BuildFilterFunc() (FilterFunc, error) {
 		}
 	}
 	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
 		}
@@ -111,6 +111,9 @@ func (o *Options) BuildFilterFunc() (FilterFunc, error) {
 		if s != nil && !s.Matches(labels.Set(pod.GetLabels())) {
 			return false
 		}
+		if o.filter != nil && !o.filter(pod) {
+			return false
+		}
 		return true
 	}, nil
 }
@@ -142,21 +145,39 @@ func BuildGetPodsAssignedToNodeFunc(podInformer cache.SharedIndexInformer) (GetP
 		if err != nil {
 			return nil, err
 		}
-		pods := make([]*v1.Pod, 0, len(objs))
+		return ConvertToPods(objs, filter), nil
-		for _, obj := range objs {
-			pod, ok := obj.(*v1.Pod)
-			if !ok {
-				continue
-			}
-			if filter(pod) {
-				pods = append(pods, pod)
-			}
-		}
-		return pods, nil
 	}
 	return getPodsAssignedToNode, nil
 }
 
+func ConvertToPods(objs []interface{}, filter FilterFunc) []*v1.Pod {
+	pods := make([]*v1.Pod, 0, len(objs))
+	for _, obj := range objs {
+		pod, ok := obj.(*v1.Pod)
+		if !ok {
+			continue
+		}
+		if filter == nil || filter(pod) {
+			pods = append(pods, pod)
+		}
+	}
+	return pods
+}
+
+// ListPodsOnNodes returns all pods on given nodes.
+func ListPodsOnNodes(nodes []*v1.Node, getPodsAssignedToNode GetPodsAssignedToNodeFunc, filter FilterFunc) ([]*v1.Pod, error) {
+	pods := make([]*v1.Pod, 0)
+	for _, node := range nodes {
+		podsOnNode, err := ListPodsOnANode(node.Name, getPodsAssignedToNode, filter)
+		if err != nil {
+			return nil, err
+		}
+
+		pods = append(pods, podsOnNode...)
+	}
+	return pods, nil
+}
+
 // 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
@@ -187,11 +208,28 @@ func ListAllPodsOnANode(
 	return pods, nil
 }
 
+func GroupByNamespace(pods []*v1.Pod) map[string][]*v1.Pod {
+	m := make(map[string][]*v1.Pod)
+	for i := 0; i < len(pods); i++ {
+		pod := pods[i]
+		m[pod.Namespace] = append(m[pod.Namespace], pod)
+	}
+	return m
+}
+
 // OwnerRef returns the ownerRefList for the pod.
 func OwnerRef(pod *v1.Pod) []metav1.OwnerReference {
 	return pod.ObjectMeta.GetOwnerReferences()
 }
 
+func OwnerRefUIDs(pod *v1.Pod) []string {
+	var ownerRefUIDs []string
+	for _, ownerRef := range OwnerRef(pod) {
+		ownerRefUIDs = append(ownerRefUIDs, string(ownerRef.UID))
+	}
+	return ownerRefUIDs
+}
+
 func IsBestEffortPod(pod *v1.Pod) bool {
 	return utils.GetPodQOS(pod) == v1.PodQOSBestEffort
 }
@@ -234,3 +272,12 @@ func SortPodsBasedOnAge(pods []*v1.Pod) {
 		return pods[i].CreationTimestamp.Before(&pods[j].CreationTimestamp)
 	})
 }
+
+func GroupByNodeName(pods []*v1.Pod) map[string][]*v1.Pod {
+	m := make(map[string][]*v1.Pod)
+	for i := 0; i < len(pods); i++ {
+		pod := pods[i]
+		m[pod.Spec.NodeName] = append(m[pod.Spec.NodeName], pod)
+	}
+	return m
+}

pkg/descheduler/pod/pods_test.go

@@ -176,3 +176,67 @@ func TestSortPodsBasedOnAge(t *testing.T) {
 		}
 	}
 }
+
+func TestGroupByNodeName(t *testing.T) {
+	tests := []struct {
+		name   string
+		pods   []*v1.Pod
+		expMap map[string][]*v1.Pod
+	}{
+		{
+			name:   "list of pods is empty",
+			pods:   []*v1.Pod{},
+			expMap: map[string][]*v1.Pod{},
+		},
+		{
+			name: "pods are on same node",
+			pods: []*v1.Pod{
+				{Spec: v1.PodSpec{
+					NodeName: "node1",
+				}},
+				{Spec: v1.PodSpec{
+					NodeName: "node1",
+				}},
+			},
+			expMap: map[string][]*v1.Pod{"node1": {
+				{Spec: v1.PodSpec{
+					NodeName: "node1",
+				}},
+				{Spec: v1.PodSpec{
+					NodeName: "node1",
+				}},
+			}},
+		},
+		{
+			name: "pods are on different nodes",
+			pods: []*v1.Pod{
+				{Spec: v1.PodSpec{
+					NodeName: "node1",
+				}},
+				{Spec: v1.PodSpec{
+					NodeName: "node2",
+				}},
+			},
+			expMap: map[string][]*v1.Pod{
+				"node1": {
+					{Spec: v1.PodSpec{
+						NodeName: "node1",
+					}},
+				},
+				"node2": {
+					{Spec: v1.PodSpec{
+						NodeName: "node2",
+					}},
+				},
+			},
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			resultMap := GroupByNodeName(test.pods)
+			if !reflect.DeepEqual(resultMap, test.expMap) {
+				t.Errorf("Expected %v node map, got %v", test.expMap, resultMap)
+			}
+		})
+	}
+}