Merge pull request #31760 from Random-Liu/remove-stop-services-option

Automatic merge from submit-queue

Node E2E: Make sure the test framework collect log properly.

Fixes #31537.

It turns out that the service [stop function](https://github.com/kubernetes/kubernetes/blob/master/test/e2e_node/e2e_node_suite_test.go#L135-L139) is never called, because `stop-services` is [always `false`](https://github.com/kubernetes/kubernetes/blob/master/test/e2e_node/runner/remote/run_remote.go#L444). And currently we collect log in the `stop` function.

This PR makes sure the `stop-services` is always true in node e2e. This will fix #31537.

@yujuhong @dchen1107 @vishh

hack/make-rules/test-e2e-node.sh

@@ -168,7 +168,7 @@ else
   # Test using the host the script was run on
   # Provided for backwards compatibility
   go run test/e2e_node/runner/local/run_local.go --ginkgo-flags="$ginkgoflags" \
-    --test-flags="--alsologtostderr --v 2 --report-dir=${report} --node-name $(hostname) \
+    --test-flags="--alsologtostderr --v 4 --report-dir=${report} --node-name $(hostname) \
-    --start-services=true --stop-services=true $test_args" --build-dependencies=true
+    $test_args" --build-dependencies=true
   exit $?
 fi

test/e2e_node/remote/remote.go

@@ -195,8 +195,8 @@ func RunRemote(archive string, host string, cleanup bool, junitFilePrefix string
 	// Run the tests
 	cmd = getSshCommand(" && ",
 		fmt.Sprintf("cd %s", tmp),
-		fmt.Sprintf("timeout -k 30s %fs ./ginkgo %s ./e2e_node.test -- --logtostderr --v 2 --stop-services=%t --node-name=%s --report-dir=%s/results --report-prefix=%s %s",
+		fmt.Sprintf("timeout -k 30s %fs ./ginkgo %s ./e2e_node.test -- --logtostderr --v 4 --node-name=%s --report-dir=%s/results --report-prefix=%s %s",
-			testTimeoutSeconds.Seconds(), ginkgoFlags, cleanup, host, tmp, junitFilePrefix, testArgs),
+			testTimeoutSeconds.Seconds(), ginkgoFlags, host, tmp, junitFilePrefix, testArgs),
 	)
 	aggErrs := []error{}
 

test/e2e_node/services/services.go

@@ -87,6 +87,7 @@ func (e *E2EServices) Start() error {
 		// "--cgroups-per-qos="+strconv.FormatBool(framework.TestContext.CgroupsPerQOS),
 		"--manifest-path", framework.TestContext.ManifestPath,
 		"--eviction-hard", framework.TestContext.EvictionHard,
+		"--logtostderr",
 	)
 	e.services = newServer("services", startCmd, nil, nil, getHealthCheckURLs(), servicesLogFile, false)
 	return e.services.start()
@@ -195,6 +196,7 @@ func (es *e2eService) run() error {
 }
 
 func (es *e2eService) start() error {
+	glog.Info("Starting e2e services...")
 	err := es.startEtcd()
 	if err != nil {
 		return err
@@ -215,7 +217,7 @@ func (es *e2eService) start() error {
 	if err != nil {
 		return nil
 	}
-
+	glog.Info("E2E services started.")
 	return nil
 }
 
@@ -226,6 +228,7 @@ func (es *e2eService) getLogFiles() {
 	if framework.TestContext.ReportDir == "" {
 		return
 	}
+	glog.Info("Fetching log files...")
 	journaldFound := isJournaldAvailable()
 	for targetFileName, logFileData := range es.logFiles {
 		targetLink := path.Join(framework.TestContext.ReportDir, targetFileName)
@@ -234,6 +237,7 @@ func (es *e2eService) getLogFiles() {
 			if len(logFileData.journalctlCommand) == 0 {
 				continue
 			}
+			glog.Infof("Get log file %q with journalctl command %v.", targetFileName, logFileData.journalctlCommand)
 			out, err := exec.Command("sudo", append([]string{"journalctl"}, logFileData.journalctlCommand...)...).CombinedOutput()
 			if err != nil {
 				glog.Errorf("failed to get %q from journald: %v, %v", targetFileName, string(out), err)
@@ -275,6 +279,7 @@ func isJournaldAvailable() bool {
 }
 
 func (es *e2eService) stop() {
+	glog.Info("Stopping e2e services...")
 	es.getLogFiles()
 	// TODO(random-liu): Use a loop to stop all services after introducing service interface.
 	// Stop namespace controller
@@ -306,6 +311,7 @@ func (es *e2eService) stop() {
 			glog.Errorf("Failed to delete directory %s.\n%v", d, err)
 		}
 	}
+	glog.Info("E2E services stopped.")
 }
 
 func (es *e2eService) startEtcd() error {
@@ -496,6 +502,7 @@ func readinessCheck(urls []string, errCh <-chan error) error {
 
 // Note: restartOnExit == true requires len(s.healthCheckUrls) > 0 to work properly.
 func (s *server) start() error {
+	glog.Infof("Start server %q with command %q", s.name, commandToString(s.startCommand))
 	errCh := make(chan error)
 
 	var stopRestartingCh, ackStopRestartingCh chan bool
@@ -634,6 +641,7 @@ func (s *server) start() error {
 }
 
 func (s *server) kill() error {
+	glog.Infof("Kill server %q", s.name)
 	name := s.name
 	cmd := s.startCommand