add e2es & fix integration test

Signed-off-by: Britania Rodriguez Reyes <[email protected]>

Author: britaniar

pkg/controllers/updaterun/execution.go

@@ -123,7 +123,7 @@ func (r *Reconciler) executeUpdatingStage(
 	// Go through each cluster in the stage and check if it's updating/succeeded/failed.
 	for i := 0; i < len(updatingStageStatus.Clusters) && clusterUpdatingCount < maxConcurrency; i++ {
 		clusterStatus := &updatingStageStatus.Clusters[i]
-		
+
 		// Process cluster status to determine if we should skip or handle errors
 		processResult := r.processClusterStatus(clusterStatus, updatingStageStatus, updateRunRef)
 		if processResult.skip {
@@ -210,22 +210,22 @@ func (r *Reconciler) processClusterStatus(
 	updateRunRef klog.ObjectRef,
 ) clusterProcessResult {
 	clusterUpdateSucceededCond := meta.FindStatusCondition(clusterStatus.Conditions, string(placementv1beta1.ClusterUpdatingConditionSucceeded))
-	
+
 	if clusterUpdateSucceededCond != nil && clusterUpdateSucceededCond.Status == metav1.ConditionTrue {
 		// The cluster has been updated successfully.
 		return clusterProcessResult{finished: true, skip: true}
 	}
-	
+
 	if clusterUpdateSucceededCond != nil && clusterUpdateSucceededCond.Status == metav1.ConditionFalse {
-		// The cluster is marked as failed to update, this cluster is counted as updating cluster 
+		// The cluster is marked as failed to update, this cluster is counted as updating cluster
 		// since it's not finished to avoid processing more clusters than maxConcurrency in this round.
 		failedErr := fmt.Errorf("the cluster `%s` in the stage %s has failed", clusterStatus.ClusterName, updatingStageStatus.StageName)
 		klog.ErrorS(failedErr, "The cluster has failed to be updated", "updateRun", updateRunRef)
 		return clusterProcessResult{
 			err: fmt.Errorf("%w: %s", errStagedUpdatedAborted, failedErr.Error()),
 		}
 	}
-	
+
 	// The cluster needs to be processed.
 	return clusterProcessResult{}
 }
@@ -242,7 +242,7 @@ func (r *Reconciler) updateBindingForCluster(
 ) error {
 	updateRunRef := klog.KObj(updateRun)
 	updateRunStatus := updateRun.GetUpdateRunStatus()
-	
+
 	if !isBindingSyncedWithClusterStatus(resourceSnapshotName, updateRun, binding, clusterStatus) {
 		klog.V(2).InfoS("Found the first cluster that needs to be updated", "cluster", clusterStatus.ClusterName, "stage", updatingStageStatus.StageName, "updateRun", updateRunRef)
 		// The binding is not up-to-date with the cluster status.
@@ -259,7 +259,7 @@ func (r *Reconciler) updateBindingForCluster(
 		klog.V(2).InfoS("Updated the status of a binding to bound", "binding", klog.KObj(binding), "cluster", clusterStatus.ClusterName, "stage", updatingStageStatus.StageName, "updateRun", updateRunRef)
 		return r.updateBindingRolloutStarted(ctx, binding, updateRun)
 	}
-	
+
 	// The binding is synced but needs other updates
 	klog.V(2).InfoS("Found the first binding that is updating but the cluster status has not been updated", "cluster", clusterStatus.ClusterName, "stage", updatingStageStatus.StageName, "updateRun", updateRunRef)
 	bindingSpec := binding.GetBindingSpec()
@@ -292,12 +292,12 @@ func (r *Reconciler) validateUpdatingCluster(
 	updatingStageStatus *placementv1beta1.StageUpdatingStatus,
 ) error {
 	updateRunRef := klog.KObj(updateRun)
-	
+
 	// Now the cluster has to be updating, the binding should point to the right resource snapshot and the binding should be bound.
 	inSync := isBindingSyncedWithClusterStatus(resourceSnapshotName, updateRun, binding, clusterStatus)
 	rolloutStarted := condition.IsConditionStatusTrue(meta.FindStatusCondition(binding.GetBindingStatus().Conditions, string(placementv1beta1.ResourceBindingRolloutStarted)), binding.GetGeneration())
 	bindingSpec := binding.GetBindingSpec()
-	
+
 	if !inSync || !rolloutStarted || bindingSpec.State != placementv1beta1.BindingStateBound {
 		// This issue mostly happens when there are concurrent updateRuns referencing the same clusterResourcePlacement but releasing different versions.
 		// After the 1st updateRun updates the binding, and before the controller re-checks the binding status, the 2nd updateRun updates the same binding, and thus the 1st updateRun is preempted and observes the binding not matching the desired state.
@@ -310,7 +310,7 @@ func (r *Reconciler) validateUpdatingCluster(
 		markClusterUpdatingFailed(clusterStatus, updateRun.GetGeneration(), preemptedErr.Error())
 		return fmt.Errorf("%w: %s", errStagedUpdatedAborted, preemptedErr.Error())
 	}
-	
+
 	return nil
 }
 
@@ -323,12 +323,12 @@ func (r *Reconciler) handleStageCompletion(
 	updatingStageStatus *placementv1beta1.StageUpdatingStatus,
 ) (time.Duration, error) {
 	updateRunRef := klog.KObj(updateRun)
-	
+
 	// All the clusters in the stage have been updated.
 	markUpdateRunWaiting(updateRun, updatingStageStatus.StageName)
 	markStageUpdatingWaiting(updatingStageStatus, updateRun.GetGeneration())
 	klog.V(2).InfoS("The stage has finished all cluster updating", "stage", updatingStageStatus.StageName, "updateRun", updateRunRef)
-	
+
 	// Check if the after stage tasks are ready.
 	approved, waitTime, err := r.checkAfterStageTasksStatus(ctx, updatingStageIndex, updateRun)
 	if err != nil {

pkg/controllers/updaterun/initialization_integration_test.go

@@ -49,7 +49,7 @@ const (
 	regionWestus = "westus"
 )
 
-var _ = Describe("Updaterun initialization tests", func() {
+var _ = FDescribe("Updaterun initialization tests", func() {
 	var updateRun *placementv1beta1.ClusterStagedUpdateRun
 	var crp *placementv1beta1.ClusterResourcePlacement
 	var policySnapshot *placementv1beta1.ClusterSchedulingPolicySnapshot
@@ -942,12 +942,19 @@ var _ = Describe("Updaterun initialization tests", func() {
 		})
 	})
 
-	It("Should not initialize if updateRun state is Abandoned", func() {
+	It("Should not initialize if updateRun is created with state Abandoned", func() {
 		By("Creating a new clusterStagedUpdateRun in Abandoned state")
 		updateRun.Spec.State = placementv1beta1.StateAbandoned
 		Expect(k8sClient.Create(ctx, updateRun)).To(Succeed())
 
 		By("Validating the updateRun is not initialized")
+		// Populate the cache first.
+		Eventually(func() error {
+			if err := k8sClient.Get(ctx, updateRunNamespacedName, updateRun); err != nil {
+				return err
+			}
+			return nil
+		}, timeout, interval).Should(Succeed(), "failed to get the updateRun")
 		Consistently(func() error {
 			if err := k8sClient.Get(ctx, updateRunNamespacedName, updateRun); err != nil {
 				return err
@@ -960,12 +967,19 @@ var _ = Describe("Updaterun initialization tests", func() {
 		}, duration, interval).Should(Succeed(), "the abandoned updateRun should not be initialized")
 	})
 
-	It("Should not initialize if updateRun state is Stopped", func() {
+	It("Should not initialize if updateRun is created with state Stopped ", func() {
 		By("Creating a new clusterStagedUpdateRun in Stopped state")
 		updateRun.Spec.State = placementv1beta1.StateStopped
 		Expect(k8sClient.Create(ctx, updateRun)).To(Succeed())
 
 		By("Validating the updateRun is not initialized")
+		// Populate the cache first.
+		Eventually(func() error {
+			if err := k8sClient.Get(ctx, updateRunNamespacedName, updateRun); err != nil {
+				return err
+			}
+			return nil
+		}, timeout, interval).Should(Succeed(), "failed to get the updateRun")
 		Consistently(func() error {
 			if err := k8sClient.Get(ctx, updateRunNamespacedName, updateRun); err != nil {
 				return err