Add count-based checkpoint retention policies

.github/workflows/tests.yaml

@@ -5,6 +5,14 @@ on: [push, pull_request]
 jobs:
   build-and-deploy:
     runs-on: ubuntu-latest
+    services:
+      registry:
+        image: registry:2
+        ports:
+          - 5000:5000
+    env:
+      CHECKPOINT_DIR: /var/lib/kubelet/checkpoints
+      REGISTRY_PORT: 5000
     steps:
       - name: Checkout code
         uses: actions/checkout@v3
@@ -14,8 +22,20 @@ jobs:
         with:
           go-version-file: 'go.mod'
 
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+        with:
+          driver-opts: network=host
+
       - name: Install dependencies
         run: |
+          # Install bats
+          sudo apt-get update
+          sudo apt-get install -y bats
+
           # Install kind
           curl -sLo kind "$(curl -sL https://api.github.com/repos/kubernetes-sigs/kind/releases/latest | jq -r '[.assets[] | select(.name == "kind-linux-amd64")] | first | .browser_download_url')"
           chmod +x kind
@@ -26,23 +46,52 @@ jobs:
           chmod +x kubectl
           sudo mv kubectl /bin/
 
+      - name: Prepare Checkpoints Directory in Host
+        run: |
+          sudo mkdir -p ${{ env.CHECKPOINT_DIR }}
+          sudo chmod 700 ${{ env.CHECKPOINT_DIR }}
+
+      - name: Generate Checkpoint Tar Files
+        run: ./test/generate_checkpoint_tar.sh
+
       - name: Create Kubernetes cluster
         run: |
-          kind create cluster
+          kind create cluster --config=./test/kind-config.yaml
           kind export kubeconfig
 
-      - name: Load Docker image into kind cluster
+      - name: Connect the registry to the Kind network
+        id: connect-registry
         run: |
-          make docker-build IMG=checkpoint-restore-operator:ci
-          kind load docker-image checkpoint-restore-operator:ci
+          REGISTRY_ID=$(docker ps -qf "ancestor=registry:2")
+          REGISTRY_NAME=$(docker inspect -f '{{.Name}}' $REGISTRY_ID | sed 's/\///')
+          echo "REGISTRY_NAME=$REGISTRY_NAME" >> $GITHUB_ENV
+          ./test/configure_local_registry.sh
+        env:
+          REGISTRY_PORT: ${{ env.REGISTRY_PORT }}
+          REGISTRY_NAME: ${{ env.REGISTRY_NAME }}
+
+      - name: Build and push to local registry
+        uses: docker/build-push-action@v6
+        with:
+          push: true
+          tags: localhost:5000/checkpoint-restore-operator:ci
 
       - name: Deploy to Kubernetes
         run: |
           make install
-          make deploy IMG=checkpoint-restore-operator:ci
+          make deploy IMG=localhost:5000/checkpoint-restore-operator:ci
 
       - name: Wait for deployments to be ready
         run: ./test/wait_for_deployment.sh checkpoint-restore-operator-controller-manager
 
       - name: Check resources
         run: kubectl get all -n checkpoint-restore-operator-system
+
+      - name: Test Garbage Collection
+        run: sudo -E bats -f "test_garbage_collection" ./test/run_tests.bats
+
+      - name: Test Max Checkpoints Set to 0
+        run: sudo -E bats -f "test_max_checkpoints_set_to_0" ./test/run_tests.bats
+
+      - name: Test Max Checkpoints Set to 1
+        run: sudo -E bats -f "test_max_checkpoints_set_to_1" ./test/run_tests.bats

api/v1/checkpointrestoreoperator_types.go

@@ -25,8 +25,36 @@ import (
 // CheckpointRestoreOperatorSpec defines the desired state of CheckpointRestoreOperator
 type CheckpointRestoreOperatorSpec struct {
 	// Important: Run "make" to regenerate code after modifying this file
-	CheckpointDirectory        string `json:"checkpointDirectory,omitempty"`
-	MaxCheckpointsPerContainer *int   `json:"maxCheckpointsPerContainer,omitempty"`
+	CheckpointDirectory      string                `json:"checkpointDirectory,omitempty"`
+	ApplyPoliciesImmediately bool                  `json:"applyPoliciesImmediately,omitempty"`
+	GlobalPolicies           GlobalPolicySpec      `json:"globalPolicy,omitempty"`
+	ContainerPolicies        []ContainerPolicySpec `json:"containerPolicies,omitempty"`
+	PodPolicies              []PodPolicySpec       `json:"podPolicies,omitempty"`
+	NamespacePolicies        []NamespacePolicySpec `json:"namespacePolicies,omitempty"`
+}
+
+type GlobalPolicySpec struct {
+	MaxCheckpointsPerNamespaces *int `json:"maxCheckpointsPerNamespace,omitempty"`
+	MaxCheckpointsPerPod        *int `json:"maxCheckpointsPerPod,omitempty"`
+	MaxCheckpointsPerContainer  *int `json:"maxCheckpointsPerContainer,omitempty"`
+}
+
+type ContainerPolicySpec struct {
+	Namespace      string `json:"namespace,omitempty"`
+	Pod            string `json:"pod,omitempty"`
+	Container      string `json:"container,omitempty"`
+	MaxCheckpoints *int64 `json:"maxCheckpoints,omitempty"`
+}
+
+type PodPolicySpec struct {
+	Namespace      string `json:"namespace,omitempty"`
+	Pod            string `json:"pod,omitempty"`
+	MaxCheckpoints *int64 `json:"maxCheckpoints,omitempty"`
+}
+
+type NamespacePolicySpec struct {
+	Namespace      string `json:"namespace,omitempty"`
+	MaxCheckpoints *int64 `json:"maxCheckpoints,omitempty"`
 }
 
 // CheckpointRestoreOperatorStatus defines the observed state of CheckpointRestoreOperator

config/crd/bases/criu.org_checkpointrestoreoperators.yaml

@@ -41,12 +41,57 @@ spec:
             description: CheckpointRestoreOperatorSpec defines the desired state of
               CheckpointRestoreOperator
             properties:
+              applyPoliciesImmediately:
+                type: boolean
               checkpointDirectory:
                 description: 'Important: Run "make" to regenerate code after modifying
                   this file'
                 type: string
-              maxCheckpointsPerContainer:
-                type: integer
+              containerPolicies:
+                items:
+                  properties:
+                    container:
+                      type: string
+                    maxCheckpoints:
+                      format: int64
+                      type: integer
+                    namespace:
+                      type: string
+                    pod:
+                      type: string
+                  type: object
+                type: array
+              globalPolicy:
+                properties:
+                  maxCheckpointsPerContainer:
+                    type: integer
+                  maxCheckpointsPerNamespace:
+                    type: integer
+                  maxCheckpointsPerPod:
+                    type: integer
+                type: object
+              namespacePolicies:
+                items:
+                  properties:
+                    maxCheckpoints:
+                      format: int64
+                      type: integer
+                    namespace:
+                      type: string
+                  type: object
+                type: array
+              podPolicies:
+                items:
+                  properties:
+                    maxCheckpoints:
+                      format: int64
+                      type: integer
+                    namespace:
+                      type: string
+                    pod:
+                      type: string
+                  type: object
+                type: array
             type: object
           status:
             description: CheckpointRestoreOperatorStatus defines the observed state

config/samples/_v1_checkpointrestoreoperator.yaml

@@ -9,5 +9,21 @@ metadata:
     app.kubernetes.io/created-by: checkpoint-restore-operator
   name: checkpointrestoreoperator-sample
 spec:
-  maxCheckpointsPerContainer: 5
   checkpointDirectory: /var/lib/kubelet/checkpoints
+  applyPoliciesImmediately: false
+  globalPolicy:
+    maxCheckpointsPerNamespace: 50
+    maxCheckpointsPerPod: 30
+    maxCheckpointsPerContainer: 10
+  # containerPolicies:
+  #   - namespace: <namespace>
+  #     pod: <pod_name>
+  #     container: <container_name>
+  #     maxCheckpoints: 5
+  # podPolicies:
+  #   - namespace: <namespace>
+  #     pod: <pod_name>
+  #     maxCheckpoints: 10
+  # namespacePolicies:
+  #   - namespace: <namespace>
+  #     maxCheckpoints: 15