bug: stop treating a deleted cluster as a deploy failure

controllers/classifier_controller.go

@@ -383,6 +383,8 @@ func (r *ClassifierReconciler) SetupWithManager(ctx context.Context,
 		go removeStaleClassifierResources(ctx, logger)
 	}
 
+	go removeStaleClassifierReports(ctx, logger)
+
 	return c, nil
 }
 

controllers/classifier_deployer.go

@@ -155,6 +155,19 @@ func (r *ClassifierReconciler) deployClassifier(ctx context.Context, classifierS
 		l := logger.WithValues("cluster", fmt.Sprintf("%s:%s/%s",
 			cluster.Kind, cluster.Namespace, cluster.Name))
 
+		// Cluster may have been deleted since this ClassifierReport was created. Skip it rather
+		// than treating it as a failure; removeStaleClassifierReports removes the stale report
+		// separately, regardless of the report's Spec.Match value.
+		if _, err := clusterproxy.GetCluster(ctx, r.Client, cluster.Namespace, cluster.Name,
+			clusterproxy.GetClusterType(cluster)); err != nil {
+			if apierrors.IsNotFound(err) {
+				l.V(logs.LogDebug).Info("cluster no longer exists, skipping")
+				continue
+			}
+			errorSeen = err
+			continue
+		}
+
 		clusterInfo, err := r.processClassifier(ctx, classifierScope, r.ControlPlaneEndpoint, cluster, f, l)
 		if err != nil {
 			errorSeen = err

controllers/classifier_report_collection_test.go

@@ -181,6 +181,54 @@ var _ = Describe("Classifier Deployer", func() {
 		Expect(len(classifierReportList.Items)).To(BeZero())
 	})
 
+	It("pruneClassifierReportsForDeletedClusters removes reports only for clusters that no longer exist", func() {
+		clusterType := libsveltosv1beta1.ClusterTypeCapi
+
+		// existingCluster still has a Cluster instance: its ClassifierReport must survive
+		existingClusterNamespace := randomString()
+		existingClusterName := randomString()
+		existingCluster := &clusterv1.Cluster{
+			ObjectMeta: metav1.ObjectMeta{
+				Namespace: existingClusterNamespace,
+				Name:      existingClusterName,
+			},
+		}
+		keepClassifierName := randomString()
+		keepReport := getClassifierReport(keepClassifierName, existingClusterNamespace, existingClusterName)
+		keepReport.Labels = libsveltosv1beta1.GetClassifierReportLabels(
+			keepClassifierName, existingClusterName, &clusterType)
+
+		// deletedClusterNamespace/Name has no matching Cluster instance: both its ClassifierReports
+		// (from two different Classifier instances) must be removed, Spec.Match is irrelevant.
+		deletedClusterNamespace := randomString()
+		deletedClusterName := randomString()
+		staleClassifierName1 := randomString()
+		staleReport1 := getClassifierReport(staleClassifierName1, deletedClusterNamespace, deletedClusterName)
+		staleReport1.Labels = libsveltosv1beta1.GetClassifierReportLabels(
+			staleClassifierName1, deletedClusterName, &clusterType)
+		staleClassifierName2 := randomString()
+		staleReport2 := getClassifierReport(staleClassifierName2, deletedClusterNamespace, deletedClusterName)
+		staleReport2.Labels = libsveltosv1beta1.GetClassifierReportLabels(
+			staleClassifierName2, deletedClusterName, &clusterType)
+
+		initObjects := []client.Object{
+			existingCluster,
+			keepReport,
+			staleReport1,
+			staleReport2,
+		}
+
+		c := fake.NewClientBuilder().WithScheme(scheme).WithStatusSubresource(keepReport, staleReport1, staleReport2).
+			WithObjects(initObjects...).Build()
+
+		Expect(controllers.PruneClassifierReportsForDeletedClusters(context.TODO(), c, logger)).To(Succeed())
+
+		classifierReportList := &libsveltosv1beta1.ClassifierReportList{}
+		Expect(c.List(context.TODO(), classifierReportList)).To(Succeed())
+		Expect(len(classifierReportList.Items)).To(Equal(1))
+		Expect(classifierReportList.Items[0].Name).To(Equal(keepReport.Name))
+	})
+
 	It("collectClassifierReports collects ClassifierReports from clusters", func() {
 		cluster := prepareCluster()
 

controllers/export_test.go

@@ -39,29 +39,30 @@ func GroupClassifierReportsByCluster(
 }
 
 var (
-	DeployClassifierCRD                     = deployClassifierCRD
-	DeployClassifierReportCRD               = deployClassifierReportCRD
-	DeployHealthCheckCRD                    = deployHealthCheckCRD
-	DeployHealthCheckReportCRD              = deployHealthCheckReportCRD
-	DeployEventSourceCRD                    = deployEventSourceCRD
-	DeployEventReportCRD                    = deployEventReportCRD
-	DeployReloaderCRD                       = deployReloaderCRD
-	DeployReloaderReportCRD                 = deployReloaderReportCRD
-	DeployDebuggingConfigurationCRD         = deployDebuggingConfigurationCRD
-	DeployClassifierInstance                = deployClassifierInstance
-	DeploySveltosAgentInManagedCluster      = deploySveltosAgentInManagedCluster
-	ClassifierHash                          = classifierHash
-	DeployClassifierInCluster               = deployClassifierInCluster
-	UndeployClassifierFromCluster           = undeployClassifierFromCluster
-	RemoveClassifierReports                 = removeClassifierReports
-	RemoveClusterClassifierReports          = removeClusterClassifierReports
-	CollectClassifierReportsFromCluster     = collectClassifierReportsFromCluster
-	DeploySveltosAgentInManagementCluster   = deploySveltosAgentInManagementCluster
-	RemoveSveltosAgentFromManagementCluster = removeSveltosAgentFromManagementCluster
-	GetSveltosAgentLabels                   = getSveltosAgentLabels
-	GetSveltosAgentNamespace                = getSveltosAgentNamespace
-	GetSveltosAgentPatches                  = getSveltosAgentPatches
-	GetSveltosApplierPatches                = getSveltosApplierPatches
+	DeployClassifierCRD                      = deployClassifierCRD
+	DeployClassifierReportCRD                = deployClassifierReportCRD
+	DeployHealthCheckCRD                     = deployHealthCheckCRD
+	DeployHealthCheckReportCRD               = deployHealthCheckReportCRD
+	DeployEventSourceCRD                     = deployEventSourceCRD
+	DeployEventReportCRD                     = deployEventReportCRD
+	DeployReloaderCRD                        = deployReloaderCRD
+	DeployReloaderReportCRD                  = deployReloaderReportCRD
+	DeployDebuggingConfigurationCRD          = deployDebuggingConfigurationCRD
+	DeployClassifierInstance                 = deployClassifierInstance
+	DeploySveltosAgentInManagedCluster       = deploySveltosAgentInManagedCluster
+	ClassifierHash                           = classifierHash
+	DeployClassifierInCluster                = deployClassifierInCluster
+	UndeployClassifierFromCluster            = undeployClassifierFromCluster
+	RemoveClassifierReports                  = removeClassifierReports
+	RemoveClusterClassifierReports           = removeClusterClassifierReports
+	PruneClassifierReportsForDeletedClusters = pruneClassifierReportsForDeletedClusters
+	CollectClassifierReportsFromCluster      = collectClassifierReportsFromCluster
+	DeploySveltosAgentInManagementCluster    = deploySveltosAgentInManagementCluster
+	RemoveSveltosAgentFromManagementCluster  = removeSveltosAgentFromManagementCluster
+	GetSveltosAgentLabels                    = getSveltosAgentLabels
+	GetSveltosAgentNamespace                 = getSveltosAgentNamespace
+	GetSveltosAgentPatches                   = getSveltosAgentPatches
+	GetSveltosApplierPatches                 = getSveltosApplierPatches
 
 	CreateAccessRequest                        = createAccessRequest
 	GetAccessRequestName                       = getAccessRequestName

controllers/utils.go

@@ -179,6 +179,75 @@ func removeStaleClassifierResources(ctx context.Context, logger logr.Logger) {
 	}
 }
 
+// removeStaleClassifierReports periodically removes ClassifierReports whose referenced cluster
+// no longer exists. This complements the cleanup done in syncAndGetMatchingClusters, which only
+// catches reports with Spec.Match set to true; a report left over with Spec.Match false (or unset)
+// for a since-deleted cluster is otherwise never removed and gets retried by every Classifier
+// reconcile forever.
+func removeStaleClassifierReports(ctx context.Context, logger logr.Logger) {
+	const interval = 5 * time.Minute
+	ticker := time.NewTicker(interval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ctx.Done():
+			logger.Info("stopping detection and removal of stale ClassifierReports")
+			return
+		case <-ticker.C:
+			if err := pruneClassifierReportsForDeletedClusters(ctx, getManagementClusterClient(), logger); err != nil {
+				logger.V(logs.LogInfo).Info(fmt.Sprintf("failed to prune stale ClassifierReports: %v", err))
+			}
+		}
+	}
+}
+
+// pruneClassifierReportsForDeletedClusters removes every ClassifierReport referencing a cluster
+// that no longer exists, regardless of the report's Spec.Match value.
+func pruneClassifierReportsForDeletedClusters(ctx context.Context, c client.Client, logger logr.Logger) error {
+	classifierReportList := &libsveltosv1beta1.ClassifierReportList{}
+	if err := c.List(ctx, classifierReportList); err != nil {
+		return fmt.Errorf("failed to list ClassifierReports: %w", err)
+	}
+
+	// Multiple ClassifierReports (from different Classifier instances) can reference the same
+	// cluster; check each distinct cluster only once per sweep.
+	checked := make(map[corev1.ObjectReference]bool)
+
+	for i := range classifierReportList.Items {
+		report := &classifierReportList.Items[i]
+		if report.Spec.ClusterNamespace == "" {
+			continue
+		}
+
+		cluster := getClusterRefFromClassifierReport(report)
+		if checked[*cluster] {
+			continue
+		}
+		checked[*cluster] = true
+
+		_, err := clusterproxy.GetCluster(ctx, c, cluster.Namespace, cluster.Name, clusterproxy.GetClusterType(cluster))
+		if err == nil {
+			continue
+		}
+		if !apierrors.IsNotFound(err) {
+			logger.V(logs.LogInfo).Info(fmt.Sprintf("failed to get cluster %s:%s/%s: %v",
+				cluster.Kind, cluster.Namespace, cluster.Name, err))
+			continue
+		}
+
+		logger.V(logs.LogInfo).Info(fmt.Sprintf(
+			"cluster %s:%s/%s no longer exists, removing its ClassifierReports",
+			cluster.Kind, cluster.Namespace, cluster.Name))
+		if err := removeClusterClassifierReports(ctx, c, cluster.Namespace, cluster.Name,
+			clusterproxy.GetClusterType(cluster), logger); err != nil {
+			logger.V(logs.LogInfo).Info(fmt.Sprintf("failed to remove stale ClassifierReports: %v", err))
+		}
+	}
+
+	return nil
+}
+
 func deplAssociatedClusterExist(ctx context.Context, c client.Client, depl *appsv1.Deployment,
 	logger logr.Logger) (exist bool, clusterName, clusterNamespace string, clusterType libsveltosv1beta1.ClusterType) {