Merge pull request kubernetes#84886 from YangLu1031/daemonRestartTest

Verify kubelet & kube-proxy recovery on Windows nodes

Author: k8s-ci-robot

test/e2e/apps/daemon_restart.go

@@ -92,9 +92,12 @@ func (r *RestartDaemonConfig) String() string {
 // waitUp polls healthz of the daemon till it returns "ok" or the polling hits the pollTimeout
 func (r *RestartDaemonConfig) waitUp() {
 	framework.Logf("Checking if %v is up by polling for a 200 on its /healthz endpoint", r)
+	nullDev := "/dev/null"
+	if framework.NodeOSDistroIs("windows") {
+		nullDev = "NUL"
+	}
 	healthzCheck := fmt.Sprintf(
-		"curl -s -o /dev/null -I -w \"%%{http_code}\" http://localhost:%v/healthz", r.healthzPort)
-
+		"curl -s -o %v -I -w \"%%{http_code}\" http://localhost:%v/healthz", nullDev, r.healthzPort)
 	err := wait.Poll(r.pollInterval, r.pollTimeout, func() (bool, error) {
 		result, err := e2essh.NodeExec(r.nodeName, healthzCheck, framework.TestContext.Provider)
 		framework.ExpectNoError(err)
@@ -115,8 +118,12 @@ func (r *RestartDaemonConfig) waitUp() {
 
 // kill sends a SIGTERM to the daemon
 func (r *RestartDaemonConfig) kill() {
+	killCmd := fmt.Sprintf("pgrep %v | xargs -I {} sudo kill {}", r.daemonName)
+	if framework.NodeOSDistroIs("windows") {
+		killCmd = fmt.Sprintf("taskkill /im %v.exe /f", r.daemonName)
+	}
 	framework.Logf("Killing %v", r)
-	_, err := e2essh.NodeExec(r.nodeName, fmt.Sprintf("pgrep %v | xargs -I {} sudo kill {}", r.daemonName), framework.TestContext.Provider)
+	_, err := e2essh.NodeExec(r.nodeName, killCmd, framework.TestContext.Provider)
 	framework.ExpectNoError(err)
 }
 
@@ -319,4 +326,18 @@ var _ = SIGDescribe("DaemonRestart [Disruptive]", func() {
 			framework.Failf("Net container restart count went from %v -> %v after kubelet restart on nodes %v \n\n %+v", preRestarts, postRestarts, badNodes, tracker)
 		}
 	})
+
+	ginkgo.It("Kube-proxy should recover after being killed accidentally", func() {
+		nodeIPs, err := e2enode.GetPublicIps(f.ClientSet)
+		if err != nil {
+			framework.Logf("Unexpected error occurred: %v", err)
+		}
+		for _, ip := range nodeIPs {
+			restarter := NewRestartConfig(
+				ip, "kube-proxy", ports.ProxyHealthzPort, restartPollInterval, restartTimeout)
+			// restart method will kill the kube-proxy process and wait for recovery,
+			// if not able to recover, will throw test failure.
+			restarter.restart()
+		}
+	})
 })