OCPBUGS-52556: Remove SVM console instance if v1alpha1 ConsolePlugin version in present in the CRDs status

Author: jhadvig

manifests/03-rbac-role-cluster.yaml

@@ -135,6 +135,15 @@ rules:
       - get
       - list
       - watch
+  - apiGroups:
+      - migration.k8s.io
+    resources:
+      - storageversionmigrations
+    verbs:
+      - get
+      - list
+      - watch
+      - delete
 ---
 kind: ClusterRole
 apiVersion: rbac.authorization.k8s.io/v1
@@ -151,10 +160,12 @@ rules:
       - apiextensions.k8s.io
     resources:
       - customresourcedefinitions
+      - customresourcedefinitions/status
     verbs:
       - get
       - list
       - watch
+      - patch
   - apiGroups:
       - admissionregistration.k8s.io
     resources:

pkg/console/controllers/storageversionmigration/controller.go

@@ -0,0 +1,213 @@
+package storageversionmigration
+
+import (
+	"context"
+	"time"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/apimachinery/pkg/util/json"
+	"k8s.io/client-go/dynamic"
+	"k8s.io/client-go/dynamic/dynamicinformer"
+	"k8s.io/klog/v2"
+
+	"github.com/openshift/console-operator/pkg/console/status"
+	"github.com/openshift/library-go/pkg/controller/factory"
+	"github.com/openshift/library-go/pkg/operator/events"
+	"github.com/openshift/library-go/pkg/operator/v1helpers"
+)
+
+const (
+	storageVersionMigrationName = "console-plugin-storage-version-migration"
+	consolePluginCRDName        = "consoleplugins.console.openshift.io"
+	maxRetries                  = 5
+	retryDelay                  = 2 * time.Second
+)
+
+var (
+	storageVersionMigrationGVR = schema.GroupVersionResource{
+		Group:    "migration.k8s.io",
+		Version:  "v1alpha1",
+		Resource: "storageversionmigrations",
+	}
+	crdGVR = schema.GroupVersionResource{
+		Group:    "apiextensions.k8s.io",
+		Version:  "v1",
+		Resource: "customresourcedefinitions",
+	}
+)
+
+type StorageVersionMigrationController struct {
+	dynamicClient  dynamic.Interface
+	operatorClient v1helpers.OperatorClient
+}
+
+func NewStorageVersionMigrationController(
+	operatorClient v1helpers.OperatorClient,
+	dynamicClient dynamic.Interface,
+	dynamicInformers dynamicinformer.DynamicSharedInformerFactory,
+	recorder events.Recorder,
+) factory.Controller {
+	c := &StorageVersionMigrationController{
+		dynamicClient:  dynamicClient,
+		operatorClient: operatorClient,
+	}
+
+	return factory.New().
+		WithInformers(
+			dynamicInformers.ForResource(storageVersionMigrationGVR).Informer(),
+		).
+		WithSync(c.sync).
+		ToController("StorageVersionMigrationController", recorder)
+}
+
+func (c *StorageVersionMigrationController) sync(ctx context.Context, syncContext factory.SyncContext) error {
+	statusHandler := status.NewStatusHandler(c.operatorClient)
+
+	// Get the StorageVersionMigration instance
+	svm, err := c.dynamicClient.Resource(storageVersionMigrationGVR).Get(ctx, storageVersionMigrationName, metav1.GetOptions{})
+	if err != nil {
+		klog.Errorf("Failed to get StorageVersionMigration: %v", err)
+		return statusHandler.FlushAndReturn(err)
+	}
+
+	// Check if the migration has succeeded with retry logic
+	succeeded, err := c.hasSucceededConditionWithRetry(svm)
+	if err != nil {
+		klog.Errorf("Failed to check conditions for StorageVersionMigration after %d retries: %v", maxRetries, err)
+		return statusHandler.FlushAndReturn(err)
+	}
+
+	if !succeeded {
+		klog.V(4).Infof("StorageVersionMigration has not succeeded yet")
+		// Delete the StorageVersionMigration if it has not succeeded yet
+		if err := c.deleteStorageVersionMigration(ctx); err != nil {
+			statusHandler.AddCondition(status.HandleDegraded("StorageVersionMigration", "FailedDeleteSVM", err))
+			return statusHandler.FlushAndReturn(err)
+		}
+		return statusHandler.FlushAndReturn(nil)
+	}
+
+	klog.Infof("StorageVersionMigration has succeeded, setting ConsolePlugin CRD storedVersions to v1")
+
+	// Set CRD storedVersions to v1
+	if err := c.removeV1Alpha1FromCRD(ctx); err != nil {
+		statusHandler.AddCondition(status.HandleDegraded("StorageVersionMigration", "FailedPatchCRD", err))
+		return statusHandler.FlushAndReturn(err)
+	}
+
+	return statusHandler.FlushAndReturn(nil)
+}
+
+// hasSucceededConditionWithRetry checks if the StorageVersionMigration has succeeded with retry logic
+func (c *StorageVersionMigrationController) hasSucceededConditionWithRetry(svm *unstructured.Unstructured) (bool, error) {
+	var lastErr error
+	for attempt := 1; attempt <= maxRetries; attempt++ {
+		succeeded, err := c.hasSucceededCondition(svm)
+		if err == nil {
+			return succeeded, nil
+		}
+
+		lastErr = err
+		klog.Warningf("Attempt %d/%d failed to check StorageVersionMigration conditions: %v", attempt, maxRetries, err)
+
+		if attempt < maxRetries {
+			klog.V(4).Infof("Retrying in %v...", retryDelay)
+			time.Sleep(retryDelay)
+		}
+	}
+
+	return false, lastErr
+}
+
+// hasSucceededCondition checks if the StorageVersionMigration has a 'Succeeded' condition with status 'True'
+func (c *StorageVersionMigrationController) hasSucceededCondition(svm *unstructured.Unstructured) (bool, error) {
+	conditions, found, err := unstructured.NestedSlice(svm.Object, "status", "conditions")
+	if err != nil {
+		return false, err
+	}
+	if !found {
+		return false, nil
+	}
+
+	for _, condition := range conditions {
+		conditionMap, ok := condition.(map[string]interface{})
+		if !ok {
+			continue
+		}
+
+		conditionType, typeFound := conditionMap["type"].(string)
+		conditionStatus, statusFound := conditionMap["status"].(string)
+
+		if typeFound && statusFound && conditionType == "Succeeded" && conditionStatus == "True" {
+			return true, nil
+		}
+	}
+
+	return false, nil
+}
+
+// removeV1Alpha1FromCRD removes 'v1alpha1' from the ConsolePlugin CRD's storedVersions
+func (c *StorageVersionMigrationController) removeV1Alpha1FromCRD(ctx context.Context) error {
+	// Get the ConsolePlugin CRD
+	crd, err := c.dynamicClient.Resource(crdGVR).Get(ctx, consolePluginCRDName, metav1.GetOptions{})
+	if err != nil {
+		klog.Errorf("Failed to get ConsolePlugin CRD: %v", err)
+		return err
+	}
+
+	// Verify CRD exists and has status
+	_, found, err := unstructured.NestedMap(crd.Object, "status")
+	if err != nil {
+		klog.Errorf("Failed to get status for ConsolePlugin CRD: %v", err)
+		return err
+	}
+	if !found {
+		klog.Info("No status found for ConsolePlugin CRD")
+		return nil
+	}
+
+	// Set storedVersions to only contain v1
+	newStoredVersions := []string{"v1"}
+
+	// Create and apply patch
+	return c.patchCRDStoredVersions(ctx, newStoredVersions)
+}
+
+// patchCRDStoredVersions applies a patch to update the CRD's storedVersions
+func (c *StorageVersionMigrationController) patchCRDStoredVersions(ctx context.Context, newStoredVersions []string) error {
+	patch := map[string]interface{}{
+		"status": map[string]interface{}{
+			"storedVersions": newStoredVersions,
+		},
+	}
+
+	patchBytes, err := json.Marshal(patch)
+	if err != nil {
+		klog.Errorf("Failed to marshal patch for ConsolePlugin CRD: %v", err)
+		return err
+	}
+
+	// Apply the patch
+	_, err = c.dynamicClient.Resource(crdGVR).Patch(ctx, consolePluginCRDName, types.MergePatchType, patchBytes, metav1.PatchOptions{})
+	if err != nil {
+		klog.Errorf("Failed to patch ConsolePlugin CRD: %v", err)
+		return err
+	}
+
+	klog.Infof("Successfully set ConsolePlugin CRD storedVersions to v1")
+	return nil
+}
+
+// deleteStorageVersionMigration deletes the StorageVersionMigration resource
+func (c *StorageVersionMigrationController) deleteStorageVersionMigration(ctx context.Context) error {
+	klog.Infof("Deleting StorageVersionMigration")
+	err := c.dynamicClient.Resource(storageVersionMigrationGVR).Delete(ctx, storageVersionMigrationName, metav1.DeleteOptions{})
+	if err != nil {
+		klog.Errorf("Failed to delete StorageVersionMigration: %v", err)
+		return err
+	}
+	return nil
+}

pkg/console/starter/starter.go

@@ -36,6 +36,7 @@ import (
 	pdb "github.com/openshift/console-operator/pkg/console/controllers/poddisruptionbudget"
 	"github.com/openshift/console-operator/pkg/console/controllers/route"
 	"github.com/openshift/console-operator/pkg/console/controllers/service"
+	"github.com/openshift/console-operator/pkg/console/controllers/storageversionmigration"
 	upgradenotification "github.com/openshift/console-operator/pkg/console/controllers/upgradenotification"
 	"github.com/openshift/console-operator/pkg/console/controllers/util"
 	"github.com/openshift/library-go/pkg/controller/controllercmd"
@@ -555,6 +556,14 @@ func RunOperator(ctx context.Context, controllerContext *controllercmd.Controlle
 		recorder,
 	)
 
+	// Create the StorageVersionMigration controller
+	storageversionmigrationController := storageversionmigration.NewStorageVersionMigrationController(
+		operatorClient,
+		dynamicClient,
+		dynamicInformers,
+		recorder,
+	)
+
 	configUpgradeableController := unsupportedconfigoverridescontroller.NewUnsupportedConfigOverridesController("UnsupportedConfigOverridesController", operatorClient, controllerContext.EventRecorder)
 	logLevelController := loglevel.NewClusterOperatorLoggingController(operatorClient, controllerContext.EventRecorder)
 	managementStateController := managementstatecontroller.NewOperatorManagementStateController(api.ClusterOperatorName, operatorClient, controllerContext.EventRecorder)
@@ -603,6 +612,7 @@ func RunOperator(ctx context.Context, controllerContext *controllercmd.Controlle
 		cliOIDCClientStatusController,
 		upgradeNotificationController,
 		staleConditionsController,
+		storageversionmigrationController,
 	} {
 		go controller.Run(ctx, 1)
 	}