Fix: gc failure cause workflow restart not working properly (#5242)

Signed-off-by: Somefive <yd219913@alibaba-inc.com>

Signed-off-by: Somefive <yd219913@alibaba-inc.com>

pkg/controller/core.oam.dev/v1alpha2/application/application_controller.go

@@ -302,6 +302,11 @@ func (r *Reconciler) gcResourceTrackers(logCtx monitorContext.Context, handler *
 	}))
 	defer subCtx.Commit("finish gc resourceTrackers")
 
+	statusUpdater := r.patchStatus
+	if isUpdate {
+		statusUpdater = r.updateStatus
+	}
+
 	var options []resourcekeeper.GCOption
 	if !gcOutdated {
 		options = append(options, resourcekeeper.DisableMarkStageGCOption{}, resourcekeeper.DisableGCComponentRevisionOption{}, resourcekeeper.DisableLegacyGCOption{})
@@ -309,8 +314,10 @@ func (r *Reconciler) gcResourceTrackers(logCtx monitorContext.Context, handler *
 	finished, waiting, err := handler.resourceKeeper.GarbageCollect(logCtx, options...)
 	if err != nil {
 		logCtx.Error(err, "Failed to gc resourcetrackers")
-		r.Recorder.Event(handler.app, event.Warning(velatypes.ReasonFailedGC, err))
-		return r.endWithNegativeCondition(logCtx, handler.app, condition.ReconcileError(err), phase)
+		cond := condition.Deleting()
+		cond.Message = fmt.Sprintf("error encountered during garbage collection: %s", err.Error())
+		handler.app.Status.SetConditions(cond)
+		return r.result(statusUpdater(logCtx, handler.app, phase)).ret()
 	}
 	if !finished {
 		logCtx.Info("GarbageCollecting resourcetrackers unfinished")
@@ -319,13 +326,10 @@ func (r *Reconciler) gcResourceTrackers(logCtx monitorContext.Context, handler *
 			cond.Message = fmt.Sprintf("Waiting for %s to delete. (At least %d resources are deleting.)", waiting[0].DisplayName(), len(waiting))
 		}
 		handler.app.Status.SetConditions(cond)
-		return r.result(r.patchStatus(logCtx, handler.app, phase)).requeue(baseGCBackoffWaitTime).ret()
+		return r.result(statusUpdater(logCtx, handler.app, phase)).requeue(baseGCBackoffWaitTime).ret()
 	}
 	logCtx.Info("GarbageCollected resourcetrackers")
-	if isUpdate {
-		return r.result(r.updateStatus(logCtx, handler.app, phase)).ret()
-	}
-	return r.result(r.patchStatus(logCtx, handler.app, phase)).ret()
+	return r.result(statusUpdater(logCtx, handler.app, phase)).ret()
 }
 
 type reconcileResult struct {

test/e2e-multicluster-test/multicluster_test.go

@@ -714,5 +714,87 @@ var _ = Describe("Test multicluster scenario", func() {
 			Expect(cm.Data["cluster"]).Should(Equal("cluster-worker"))
 			Expect(k8sClient.Delete(hubCtx, def)).Should(Succeed())
 		})
+
+		It("Test application with failed gc and restart workflow", func() {
+			By("duplicate cluster")
+			secret := &corev1.Secret{}
+			const secretName = "disconnection-test"
+			Expect(k8sClient.Get(hubCtx, types.NamespacedName{Namespace: kubevelatypes.DefaultKubeVelaNS, Name: WorkerClusterName}, secret)).Should(Succeed())
+			secret.SetName(secretName)
+			secret.SetResourceVersion("")
+			Expect(k8sClient.Create(hubCtx, secret)).Should(Succeed())
+			defer func() {
+				_ = k8sClient.Delete(hubCtx, secret)
+			}()
+
+			By("create cluster normally")
+			bs, err := os.ReadFile("./testdata/app/app-disconnection-test.yaml")
+			Expect(err).Should(Succeed())
+			app := &v1beta1.Application{}
+			Expect(yaml.Unmarshal(bs, app)).Should(Succeed())
+			app.SetNamespace(namespace)
+			Expect(k8sClient.Create(hubCtx, app)).Should(Succeed())
+			key := client.ObjectKeyFromObject(app)
+			Eventually(func(g Gomega) {
+				g.Expect(k8sClient.Get(hubCtx, key, app)).Should(Succeed())
+				g.Expect(app.Status.Phase).Should(Equal(common.ApplicationRunning))
+			}).WithTimeout(10 * time.Second).WithPolling(2 * time.Second).Should(Succeed())
+
+			By("disconnect cluster")
+			Expect(k8sClient.Get(hubCtx, types.NamespacedName{Namespace: kubevelatypes.DefaultKubeVelaNS, Name: secretName}, secret)).Should(Succeed())
+			secret.Data["endpoint"] = []byte("https://1.2.3.4:9999")
+			Expect(k8sClient.Update(hubCtx, secret)).Should(Succeed())
+
+			By("update application")
+			Expect(k8sClient.Get(hubCtx, key, app)).Should(Succeed())
+			app.Spec.Policies = nil
+			Expect(k8sClient.Update(hubCtx, app)).Should(Succeed())
+			Eventually(func(g Gomega) {
+				g.Expect(k8sClient.Get(hubCtx, key, app)).Should(Succeed())
+				g.Expect(app.Status.ObservedGeneration).Should(Equal(app.Generation))
+				g.Expect(app.Status.Phase).Should(Equal(common.ApplicationRunning))
+				rts := &v1beta1.ResourceTrackerList{}
+				g.Expect(k8sClient.List(hubCtx, rts, client.MatchingLabels{oam.LabelAppName: key.Name, oam.LabelAppNamespace: key.Namespace})).Should(Succeed())
+				cnt := 0
+				for _, item := range rts.Items {
+					if item.Spec.Type == v1beta1.ResourceTrackerTypeVersioned {
+						cnt++
+					}
+				}
+				g.Expect(cnt).Should(Equal(2))
+			}).WithTimeout(10 * time.Second).WithPolling(2 * time.Second).Should(Succeed())
+
+			By("try update application again")
+			Expect(k8sClient.Get(hubCtx, key, app)).Should(Succeed())
+			if app.Annotations == nil {
+				app.Annotations = map[string]string{}
+			}
+			app.Annotations[oam.AnnotationPublishVersion] = "test"
+			Expect(k8sClient.Update(hubCtx, app)).Should(Succeed())
+			Eventually(func(g Gomega) {
+				g.Expect(k8sClient.Get(hubCtx, key, app)).Should(Succeed())
+				g.Expect(app.Status.LatestRevision).ShouldNot(BeNil())
+				g.Expect(app.Status.LatestRevision.Revision).Should(Equal(int64(3)))
+				g.Expect(app.Status.ObservedGeneration).Should(Equal(app.Generation))
+				g.Expect(app.Status.Phase).Should(Equal(common.ApplicationRunning))
+			}).WithTimeout(1 * time.Minute).WithPolling(2 * time.Second).Should(Succeed())
+
+			By("clear disconnection cluster secret")
+			Expect(k8sClient.Get(hubCtx, types.NamespacedName{Namespace: kubevelatypes.DefaultKubeVelaNS, Name: secretName}, secret)).Should(Succeed())
+			Expect(k8sClient.Delete(hubCtx, secret)).Should(Succeed())
+
+			By("wait gc application completed")
+			Eventually(func(g Gomega) {
+				rts := &v1beta1.ResourceTrackerList{}
+				g.Expect(k8sClient.List(hubCtx, rts, client.MatchingLabels{oam.LabelAppName: key.Name, oam.LabelAppNamespace: key.Namespace})).Should(Succeed())
+				cnt := 0
+				for _, item := range rts.Items {
+					if item.Spec.Type == v1beta1.ResourceTrackerTypeVersioned {
+						cnt++
+					}
+				}
+				g.Expect(cnt).Should(Equal(1))
+			}).WithTimeout(30 * time.Second).WithPolling(2 * time.Second).Should(Succeed())
+		})
 	})
 })

test/e2e-multicluster-test/testdata/app/app-disconnection-test.yaml

@@ -0,0 +1,17 @@
+apiVersion: core.oam.dev/v1beta1
+kind: Application
+metadata:
+  name: app-disconnection-test
+spec:
+  components:
+    - type: k8s-objects
+      name: app-dis-cm
+      properties:
+        objects:
+          - apiVersion: v1
+            kind: ConfigMap
+  policies:
+    - type: topology
+      name: disconnection-test
+      properties:
+        clusters: ["disconnection-test"]