fixup storage tracking

Author: lmiccini

apis/bases/rabbitmq.openstack.org_rabbitmqs.yaml

@@ -1969,6 +1969,14 @@ spec:
                   the opentack-operator in the top-level CR (e.g. the ContainerImage)
                 format: int64
                 type: integer
+              pvsBeingDeleted:
+                description: |-
+                  PVsBeingDeleted - list of PV names that are expected to be deleted during storage wipe
+                  Tracked to avoid scanning all PVs in the cluster during cleanup verification
+                items:
+                  type: string
+                type: array
+                x-kubernetes-list-type: set
               queueType:
                 description: QueueType - store whether default ha-all policy is present
                   or not

apis/rabbitmq/v1beta1/rabbitmq_types.go

@@ -151,6 +151,11 @@ type RabbitMqStatus struct {
 	// StorageWipeStartedAt - timestamp when storage wipe process started
 	// Used to implement timeout protection against stuck PV/PVC deletions
 	StorageWipeStartedAt *metav1.Time `json:"storageWipeStartedAt,omitempty"`
+
+	// PVsBeingDeleted - list of PV names that are expected to be deleted during storage wipe
+	// Tracked to avoid scanning all PVs in the cluster during cleanup verification
+	// +listType=set
+	PVsBeingDeleted []string `json:"pvsBeingDeleted,omitempty"`
 }
 
 //+kubebuilder:object:root=true

apis/rabbitmq/v1beta1/zz_generated.deepcopy.go

@@ -1969,6 +1969,14 @@ spec:
                   the opentack-operator in the top-level CR (e.g. the ContainerImage)
                 format: int64
                 type: integer
+              pvsBeingDeleted:
+                description: |-
+                  PVsBeingDeleted - list of PV names that are expected to be deleted during storage wipe
+                  Tracked to avoid scanning all PVs in the cluster during cleanup verification
+                items:
+                  type: string
+                type: array
+                x-kubernetes-list-type: set
               queueType:
                 description: QueueType - store whether default ha-all policy is present
                   or not

config/crd/bases/rabbitmq.openstack.org_rabbitmqs.yaml

@@ -23,11 +23,21 @@ rules:
   - ""
   resources:
   - persistentvolumeclaims
+  verbs:
+  - delete
+  - get
+  - list
+  - update
+  - watch
+- apiGroups:
+  - ""
+  resources:
   - persistentvolumes
   verbs:
   - delete
   - get
   - list
+  - patch
   - update
   - watch
 - apiGroups:

config/rbac/role.yaml

@@ -153,7 +153,7 @@ type Reconciler struct {
 // +kubebuilder:rbac:groups=core,resources=persistentvolumeclaims,verbs=get;list;watch;update;delete
 
 // Required to manage PersistentVolumes for version upgrades (ensuring clean storage)
-// +kubebuilder:rbac:groups=core,resources=persistentvolumes,verbs=get;list;watch;update;delete
+// +kubebuilder:rbac:groups=core,resources=persistentvolumes,verbs=get;list;watch;update;patch;delete
 
 // Required to manage PodDisruptionBudgets for multi-replica deployments
 // +kubebuilder:rbac:groups=policy,resources=poddisruptionbudgets,verbs=get;list;watch;create;update;patch;delete
@@ -560,6 +560,29 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (result ct
 			instance.Status.StorageWipeStartedAt = &now
 			Log.Info("Starting storage wipe", "reason", wipeReason, "phase", "DeletingStorage")
 
+			// CRITICAL: Capture PV names NOW, before any deletion happens
+			// This ensures we track the correct PVs from the current cluster
+			// Always reset tracking when starting a new wipe to clear stale data
+			pvcList, err := r.listPVCsForInstance(ctx, instance.Namespace, instance.Name)
+			if err != nil {
+				Log.Error(err, "Failed to list PVCs for PV tracking during wipe start")
+				return ctrl.Result{}, err
+			}
+
+			// Always update pvsBeingDeleted, even if empty - this clears stale data
+			pvNames := make([]string, 0, len(pvcList.Items))
+			for i := range pvcList.Items {
+				if pvcList.Items[i].Spec.VolumeName != "" {
+					pvNames = append(pvNames, pvcList.Items[i].Spec.VolumeName)
+				}
+			}
+			instance.Status.PVsBeingDeleted = pvNames
+			if len(pvNames) > 0 {
+				Log.Info("Captured PV names for tracking before wipe", "pvCount", len(pvNames), "pvNames", pvNames)
+			} else {
+				Log.Info("No PVCs found to track - clearing stale PV tracking data")
+			}
+
 			// Emit event for observability
 			if r.Recorder != nil {
 				if wipeReason == "version upgrade" && instance.Annotations != nil {
@@ -574,7 +597,7 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (result ct
 				}
 			}
 
-			// Persist the phase update and timestamp
+			// Persist the phase update, timestamp, and PV tracking
 			if err := helper.PatchInstance(ctx, instance); err != nil {
 				return ctrl.Result{}, err
 			}
@@ -598,6 +621,7 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (result ct
 			CurrentQueueType:     string(instance.Status.QueueType),
 			Reason:               wipeReason,
 			StorageWipeStartedAt: storageWipeStartedAt,
+			PVsBeingDeleted:      instance.Status.PVsBeingDeleted,
 			DeleteCluster: func(ctx context.Context) error {
 				err := rmqCluster.Delete(ctx, helper)
 				if err != nil && !k8s_errors.IsNotFound(err) {
@@ -612,12 +636,23 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (result ct
 
 		// Perform the storage wipe (handles all steps: delete cluster, wait for pods,
 		// patch PV reclaim policy, delete PVCs, verify cleanup)
-		result, err := r.performStorageWipe(ctx, wipeParams, Log)
+		result, err := r.performStorageWipe(ctx, &wipeParams, Log)
 		if err != nil {
 			Log.Error(err, "Storage wipe failed")
 			return result, err
 		}
 
+		// Update status with PV tracking information (if changed during this reconcile)
+		if len(wipeParams.PVsBeingDeleted) > 0 && len(instance.Status.PVsBeingDeleted) == 0 {
+			instance.Status.PVsBeingDeleted = wipeParams.PVsBeingDeleted
+			// Persist status immediately so PV tracking survives across reconciles
+			if err := helper.PatchInstance(ctx, instance); err != nil {
+				Log.Error(err, "Failed to update status with PV tracking")
+				return ctrl.Result{}, err
+			}
+			Log.Info("Updated status with PV tracking", "pvCount", len(wipeParams.PVsBeingDeleted))
+		}
+
 		// If result has Requeue set, we're still in progress
 		if result.Requeue || result.RequeueAfter > 0 {
 			return result, nil
@@ -628,9 +663,10 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (result ct
 			if instance.Annotations != nil {
 				if targetVersion, hasTarget := instance.Annotations[rabbitmqv1beta1.AnnotationTargetVersion]; hasTarget && targetVersion != "" {
 					instance.Status.CurrentVersion = targetVersion
-					// Clear the upgrade phase and timestamp
+					// Clear the upgrade phase, timestamp, and PV tracking
 					instance.Status.UpgradePhase = ""
 					instance.Status.StorageWipeStartedAt = nil
+					instance.Status.PVsBeingDeleted = nil
 
 					// If queue type changed during upgrade, update Status.QueueType to prevent
 					// triggering another wipe for "queue type migration"
@@ -654,6 +690,7 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (result ct
 			// Queue migration complete - update Status.QueueType
 			instance.Status.UpgradePhase = ""
 			instance.Status.StorageWipeStartedAt = nil
+			instance.Status.PVsBeingDeleted = nil
 			if instance.Spec.QueueType != nil {
 				switch *instance.Spec.QueueType {
 				case rabbitmqv1beta1.QueueTypeQuorum: