Merge pull request #26272 from yujuhong/fix_reboot_tests

e2e: fix tests that are broken because of the image prepull pod

test/e2e/framework/util.go

@@ -396,6 +396,14 @@ func PodRunningReady(p *api.Pod) (bool, error) {
 	return true, nil
 }
 
+func PodRunningReadyOrSucceeded(p *api.Pod) (bool, error) {
+	// Check if the phase is succeeded.
+	if p.Status.Phase == api.PodSucceeded {
+		return true, nil
+	}
+	return PodRunningReady(p)
+}
+
 // PodNotReady checks whether pod p's has a ready condition of status false.
 func PodNotReady(p *api.Pod) (bool, error) {
 	// Check the ready condition is false.
@@ -3217,16 +3225,30 @@ func GetSigner(provider string) (ssh.Signer, error) {
 	return sshutil.MakePrivateKeySignerFromFile(key)
 }
 
-// checkPodsRunning returns whether all pods whose names are listed in podNames
+// CheckPodsRunningReady returns whether all pods whose names are listed in
-// in namespace ns are running and ready, using c and waiting at most timeout.
+// podNames in namespace ns are running and ready, using c and waiting at most
+// timeout.
 func CheckPodsRunningReady(c *client.Client, ns string, podNames []string, timeout time.Duration) bool {
-	np, desc := len(podNames), "running and ready"
+	return CheckPodsCondition(c, ns, podNames, timeout, PodRunningReady, "running and ready")
+}
+
+// CheckPodsRunningReadyOrSucceeded returns whether all pods whose names are
+// listed in podNames in namespace ns are running and ready, or succeeded; use
+// c and waiting at most timeout.
+func CheckPodsRunningReadyOrSucceeded(c *client.Client, ns string, podNames []string, timeout time.Duration) bool {
+	return CheckPodsCondition(c, ns, podNames, timeout, PodRunningReadyOrSucceeded, "running and ready, or succeeded")
+}
+
+// CheckPodsCondition returns whether all pods whose names are listed in podNames
+// in namespace ns are in the condition, using c and waiting at most timeout.
+func CheckPodsCondition(c *client.Client, ns string, podNames []string, timeout time.Duration, condition podCondition, desc string) bool {
+	np := len(podNames)
 	Logf("Waiting up to %v for %d pods to be %s: %s", timeout, np, desc, podNames)
 	result := make(chan bool, len(podNames))
 	for ix := range podNames {
 		// Launch off pod readiness checkers.
 		go func(name string) {
-			err := waitForPodCondition(c, ns, name, desc, timeout, PodRunningReady)
+			err := waitForPodCondition(c, ns, name, desc, timeout, condition)
 			result <- err == nil
 		}(podNames[ix])
 	}

test/e2e/reboot.go

@@ -241,11 +241,11 @@ func rebootNode(c *client.Client, provider, name, rebootCmd string) bool {
 			podNames = append(podNames, p.ObjectMeta.Name)
 		}
 	}
-	framework.Logf("Node %s has %d pods: %v", name, len(podNames), podNames)
+	framework.Logf("Node %s has %d assigned pods with no liveness probes: %v", name, len(podNames), podNames)
 
 	// For each pod, we do a sanity check to ensure it's running / healthy
-	// now, as that's what we'll be checking later.
+	// or succeeded now, as that's what we'll be checking later.
-	if !framework.CheckPodsRunningReady(c, ns, podNames, framework.PodReadyBeforeTimeout) {
+	if !framework.CheckPodsRunningReadyOrSucceeded(c, ns, podNames, framework.PodReadyBeforeTimeout) {
 		printStatusAndLogsForNotReadyPods(c, ns, podNames, pods)
 		return false
 	}
@@ -267,8 +267,8 @@ func rebootNode(c *client.Client, provider, name, rebootCmd string) bool {
 	}
 
 	// Ensure all of the pods that we found on this node before the reboot are
-	// running / healthy.
+	// running / healthy, or succeeded.
-	if !framework.CheckPodsRunningReady(c, ns, podNames, rebootPodReadyAgainTimeout) {
+	if !framework.CheckPodsRunningReadyOrSucceeded(c, ns, podNames, rebootPodReadyAgainTimeout) {
 		newPods := ps.List()
 		printStatusAndLogsForNotReadyPods(c, ns, podNames, newPods)
 		return false

test/e2e/resize_nodes.go

@@ -396,7 +396,7 @@ var _ = framework.KubeDescribe("Nodes [Disruptive]", func() {
 			// the cluster is restored to health.
 			By("waiting for system pods to successfully restart")
 
-			err := framework.WaitForPodsRunningReady(api.NamespaceSystem, systemPodsNo, framework.PodReadyBeforeTimeout, map[string]string{})
+			err := framework.WaitForPodsRunningReady(api.NamespaceSystem, systemPodsNo, framework.PodReadyBeforeTimeout, framework.ImagePullerLabels)
 			Expect(err).NotTo(HaveOccurred())
 		})
 

test/e2e/restart.go

@@ -82,8 +82,8 @@ var _ = framework.KubeDescribe("Restart [Disruptive]", func() {
 			podNamesBefore[i] = p.ObjectMeta.Name
 		}
 		ns := api.NamespaceSystem
-		if !framework.CheckPodsRunningReady(f.Client, ns, podNamesBefore, framework.PodReadyBeforeTimeout) {
+		if !framework.CheckPodsRunningReadyOrSucceeded(f.Client, ns, podNamesBefore, framework.PodReadyBeforeTimeout) {
-			framework.Failf("At least one pod wasn't running and ready at test start.")
+			framework.Failf("At least one pod wasn't running and ready or succeeded at test start.")
 		}
 
 		By("restarting all of the nodes")
@@ -111,7 +111,7 @@ var _ = framework.KubeDescribe("Restart [Disruptive]", func() {
 		podNamesAfter, err := waitForNPods(ps, len(podNamesBefore), restartPodReadyAgainTimeout)
 		Expect(err).NotTo(HaveOccurred())
 		remaining := restartPodReadyAgainTimeout - time.Since(podCheckStart)
-		if !framework.CheckPodsRunningReady(f.Client, ns, podNamesAfter, remaining) {
+		if !framework.CheckPodsRunningReadyOrSucceeded(f.Client, ns, podNamesAfter, remaining) {
 			framework.Failf("At least one pod wasn't running and ready after the restart.")
 		}
 	})

test/e2e/scheduler_predicates.go

@@ -196,7 +196,7 @@ var _ = framework.KubeDescribe("SchedulerPredicates [Serial]", func() {
 			}
 		}
 
-		err = framework.WaitForPodsRunningReady(api.NamespaceSystem, int32(systemPodsNo), framework.PodReadyBeforeTimeout, map[string]string{})
+		err = framework.WaitForPodsRunningReady(api.NamespaceSystem, int32(systemPodsNo), framework.PodReadyBeforeTimeout, framework.ImagePullerLabels)
 		Expect(err).NotTo(HaveOccurred())
 
 		for _, node := range nodeList.Items {