Merge pull request #104076 from dims/explicitly-restart-kubelet-to-stabilize-serial-containerd-job

Explicitly restart kubelet to stabilize serial-containerd job

test/e2e_node/memory_manager_test.go

@@ -339,12 +339,17 @@ var _ = SIGDescribe("Memory Manager [Serial] [Feature:MemoryManager]", func() {
 			}, 30*time.Second, framework.Poll).Should(gomega.BeNil())
 
 			ginkgo.By("restarting kubelet to pick up pre-allocated hugepages")
+
 			// stop the kubelet and wait until the server will restart it automatically
 			stopKubelet()
+
 			// wait until the kubelet health check will fail
 			gomega.Eventually(func() bool {
 				return kubeletHealthCheck(kubeletHealthCheckURL)
 			}, time.Minute, time.Second).Should(gomega.BeFalse())
+
+			restartKubelet()
+
 			// wait until the kubelet health check will pass
 			gomega.Eventually(func() bool {
 				return kubeletHealthCheck(kubeletHealthCheckURL)

test/e2e_node/util.go

@@ -397,8 +397,14 @@ func getCRIClient() (internalapi.RuntimeService, internalapi.ImageManagerService
 }
 
 // TODO: Find a uniform way to deal with systemctl/initctl/service operations. #34494
-func findRunningKubletServiceName() string {
-	stdout, err := exec.Command("sudo", "systemctl", "list-units", "*kubelet*", "--state=running").CombinedOutput()
+func findKubletServiceName(running bool) string {
+	cmdLine := []string{
+		"systemctl", "list-units", "*kubelet*",
+	}
+	if running {
+		cmdLine = append(cmdLine, "--state=running")
+	}
+	stdout, err := exec.Command("sudo", cmdLine...).CombinedOutput()
 	framework.ExpectNoError(err)
 	regex := regexp.MustCompile("(kubelet-\\w+)")
 	matches := regex.FindStringSubmatch(string(stdout))
@@ -409,7 +415,7 @@ func findRunningKubletServiceName() string {
 }
 
 func restartKubelet() {
-	kubeletServiceName := findRunningKubletServiceName()
+	kubeletServiceName := findKubletServiceName(false)
 	// reset the kubelet service start-limit-hit
 	stdout, err := exec.Command("sudo", "systemctl", "reset-failed", kubeletServiceName).CombinedOutput()
 	framework.ExpectNoError(err, "Failed to reset kubelet start-limit-hit with systemctl: %v, %v", err, stdout)
@@ -420,7 +426,7 @@ func restartKubelet() {
 
 // stopKubelet will kill the running kubelet, and returns a func that will restart the process again
 func stopKubelet() func() {
-	kubeletServiceName := findRunningKubletServiceName()
+	kubeletServiceName := findKubletServiceName(true)
 
 	// reset the kubelet service start-limit-hit
 	stdout, err := exec.Command("sudo", "systemctl", "reset-failed", kubeletServiceName).CombinedOutput()