Make e2e pod start timeouts uniform

test/e2e/events.go

@@ -79,9 +79,7 @@ var _ = Describe("Events", func() {
 			Failf("Failed to create pod: %v", err)
 		}
 
-		By("waiting for the pod to start running")
+		expectNoError(waitForPodRunning(c, pod.Name))
-		err := waitForPodRunning(c, pod.Name, 300*time.Second)
-		Expect(err).NotTo(HaveOccurred())
 
 		By("verifying the pod is in kubernetes")
 		pods, err := podClient.List(labels.SelectorFromSet(labels.Set(map[string]string{"time": value})))

test/e2e/kubectl.go

@@ -35,7 +35,6 @@ const (
 	kittenImage         = "kubernetes/update-demo:kitten"
 	updateDemoSelector  = "name=update-demo"
 	updateDemoContainer = "update-demo"
-	validateTimeout     = 10 * time.Minute // TODO: Make this 30 seconds once #4566 is resolved.
 	kubectlProxyPort    = 8011
 )
 
@@ -120,7 +119,7 @@ func validateController(c *client.Client, image string, replicas int) {
 	getImageTemplate := fmt.Sprintf(`--template={{(index .currentState.info "%s").image}}`, updateDemoContainer)
 
 	By(fmt.Sprintf("waiting for all containers in %s pods to come up.", updateDemoSelector))
-	for start := time.Now(); time.Since(start) < validateTimeout; time.Sleep(5 * time.Second) {
+	for start := time.Now(); time.Since(start) < podStartTimeout; time.Sleep(5 * time.Second) {
 		getPodsOutput := runKubectl("get", "pods", "-o", "template", getPodsTemplate, "-l", updateDemoSelector)
 		pods := strings.Fields(getPodsOutput)
 		if numPods := len(pods); numPods != replicas {

test/e2e/pd.go

@@ -76,8 +76,7 @@ var _ = Describe("PD", func() {
 		_, err := podClient.Create(host0Pod)
 		expectNoError(err, fmt.Sprintf("Failed to create host0Pod: %v", err))
 
-		By("waiting up to 180 seconds for host0Pod to start running")
+		expectNoError(waitForPodRunning(c, host0Pod.Name))
-		expectNoError(waitForPodRunning(c, host0Pod.Name, 180*time.Second), "host0Pod not running after 180 seconds")
 
 		By("deleting host0Pod")
 		expectNoError(podClient.Delete(host0Pod.Name), "Failed to delete host0Pod")
@@ -86,8 +85,7 @@ var _ = Describe("PD", func() {
 		_, err = podClient.Create(host1Pod)
 		expectNoError(err, "Failed to create host1Pod")
 
-		By("waiting up to 180 seconds for host1Pod to start running")
+		expectNoError(waitForPodRunning(c, host1Pod.Name))
-		expectNoError(waitForPodRunning(c, host1Pod.Name, 180*time.Second), "host1Pod not running after 180 seconds")
 
 		By("deleting host1Pod")
 		expectNoError(podClient.Delete(host1Pod.Name), "Failed to delete host1Pod")
@@ -128,7 +126,7 @@ var _ = Describe("PD", func() {
 		By("submitting rwPod to ensure PD is formatted")
 		_, err := podClient.Create(rwPod)
 		expectNoError(err, "Failed to create rwPod")
-		expectNoError(waitForPodRunning(c, rwPod.Name, 180*time.Second), "rwPod not running after 180 seconds")
+		expectNoError(waitForPodRunning(c, rwPod.Name))
 		expectNoError(podClient.Delete(rwPod.Name), "Failed to delete host0Pod")
 
 		By("submitting host0ROPod to kubernetes")
@@ -139,11 +137,9 @@ var _ = Describe("PD", func() {
 		_, err = podClient.Create(host1ROPod)
 		expectNoError(err, "Failed to create host1ROPod")
 
-		By("waiting up to 180 seconds for host0ROPod to start running")
+		expectNoError(waitForPodRunning(c, host0ROPod.Name))
-		expectNoError(waitForPodRunning(c, host0ROPod.Name, 180*time.Second), "host0ROPod not running after 180 seconds")
 
-		By("waiting up to 180 seconds for host1ROPod to start running")
+		expectNoError(waitForPodRunning(c, host1ROPod.Name))
-		expectNoError(waitForPodRunning(c, host1ROPod.Name, 180*time.Second), "host1ROPod not running after 180 seconds")
 
 		By("deleting host0ROPod")
 		expectNoError(podClient.Delete(host0ROPod.Name), "Failed to delete host0ROPod")

test/e2e/pods.go

@@ -47,8 +47,7 @@ func runLivenessTest(c *client.Client, podDescr *api.Pod) {
 	// Wait until the pod is not pending. (Here we need to check for something other than
 	// 'Pending' other than checking for 'Running', since when failures occur, we go to
 	// 'Terminated' which can cause indefinite blocking.)
-	By("waiting for the pod to be something other than pending")
+	expectNoError(waitForPodNotPending(c, ns, podDescr.Name),
-	expectNoError(waitForPodNotPending(c, ns, podDescr.Name, 60*time.Second),
 		fmt.Sprintf("starting pod %s in namespace %s", podDescr.Name, ns))
 	By(fmt.Sprintf("Started pod %s in namespace %s", podDescr.Name, ns))
 
@@ -190,8 +189,7 @@ var _ = Describe("Pods", func() {
 			Fail(fmt.Sprintf("Failed to create pod: %v", err))
 		}
 
-		By("waiting for the pod to start running")
+		expectNoError(waitForPodRunning(c, pod.Name))
-		expectNoError(waitForPodRunning(c, pod.Name, 300*time.Second))
 
 		By("verifying the pod is in kubernetes")
 		pods, err := podClient.List(labels.SelectorFromSet(labels.Set(map[string]string{"time": value})))
@@ -213,8 +211,7 @@ var _ = Describe("Pods", func() {
 			Fail(fmt.Sprintf("Failed to update pod: %v", err))
 		}
 
-		By("waiting for the updated pod to start running")
+		expectNoError(waitForPodRunning(c, pod.Name))
-		expectNoError(waitForPodRunning(c, pod.Name, 300*time.Second))
 
 		By("verifying the updated pod is in kubernetes")
 		pods, err = podClient.List(labels.SelectorFromSet(labels.Set(map[string]string{"time": value})))
@@ -246,7 +243,7 @@ var _ = Describe("Pods", func() {
 		if err != nil {
 			Fail(fmt.Sprintf("Failed to create serverPod: %v", err))
 		}
-		expectNoError(waitForPodRunning(c, serverPod.Name, 300*time.Second))
+		expectNoError(waitForPodRunning(c, serverPod.Name))
 
 		// This service exposes port 8080 of the test pod as a service on port 8765
 		// TODO(filbranden): We would like to use a unique service name such as:
@@ -305,8 +302,7 @@ var _ = Describe("Pods", func() {
 			Fail(fmt.Sprintf("Failed to create pod: %v", err))
 		}
 
-		// Wait for client pod to complete.
+		expectNoError(waitForPodRunning(c, clientPod.Name))
-		expectNoError(waitForPodRunning(c, clientPod.Name, 60*time.Second))
 
 		// Grab its logs.  Get host first.
 		clientPodStatus, err := c.Pods(api.NamespaceDefault).Get(clientPod.Name)

test/e2e/rc.go

@@ -133,7 +133,7 @@ func ServeImageOrFail(c *client.Client, test string, image string) {
 	// Wait for the pods to enter the running state. Waiting loops until the pods
 	// are running so non-running pods cause a timeout for this test.
 	for _, pod := range pods.Items {
-		err = waitForPodRunning(c, pod.Name, 300*time.Second)
+		err = waitForPodRunning(c, pod.Name)
 		Expect(err).NotTo(HaveOccurred())
 	}
 

test/e2e/secrets.go

@@ -115,7 +115,7 @@ var _ = Describe("Secrets", func() {
 			Failf("Failed to create pod: %v", err)
 		}
 		// Wait for client pod to complete.
-		expectNoError(waitForPodRunning(c, clientPod.Name, 60*time.Second))
+		expectNoError(waitForPodRunning(c, clientPod.Name))
 
 		// Grab its logs.  Get host first.
 		clientPodStatus, err := c.Pods(ns).Get(clientPod.Name)

test/e2e/service.go

@@ -116,12 +116,10 @@ var _ = Describe("Services", func() {
 			Failf("Failed to create %s pod: %v", pod.Name, err)
 		}
 
-		By("waiting for the pod to start running")
+		expectNoError(waitForPodRunning(c, pod.Name))
-		err := waitForPodRunning(c, pod.Name, 300*time.Second)
-		Expect(err).NotTo(HaveOccurred())
 
 		By("retrieving the pod")
-		pod, err = podClient.Get(pod.Name)
+		pod, err := podClient.Get(pod.Name)
 		if err != nil {
 			Failf("Failed to get pod %s: %v", pod.Name, err)
 		}

test/e2e/util.go

@@ -31,6 +31,12 @@ import (
 	. "github.com/onsi/gomega"
 )
 
+const (
+	// Initial pod start can be delayed O(minutes) by slow docker pulls
+	// TODO: Make this 30 seconds once #4566 is resolved.
+	podStartTimeout = 5 * time.Minute
+)
+
 type testContextType struct {
 	authConfig string
 	certDir    string
@@ -50,55 +56,45 @@ func Failf(format string, a ...interface{}) {
 	Fail(fmt.Sprintf(format, a...), 1)
 }
 
-func waitForPodRunning(c *client.Client, id string, tryFor time.Duration) error {
+type podCondition func(pod *api.Pod) (bool, error)
-	trySecs := int(tryFor.Seconds())
-	for i := 0; i <= trySecs; i += 5 {
-		time.Sleep(5 * time.Second)
-		pod, err := c.Pods(api.NamespaceDefault).Get(id)
-		if err != nil {
-			return fmt.Errorf("Get pod %s failed: %v", id, err.Error())
-		}
-		if pod.Status.Phase == api.PodRunning {
-			return nil
-		}
-		Logf("Waiting for pod %s status to be %q (found %q) (%d secs)", id, api.PodRunning, pod.Status.Phase, i)
-	}
-	return fmt.Errorf("Gave up waiting for pod %s to be running after %d seconds", id, trySecs)
-}
 
-// waitForPodNotPending returns false if it took too long for the pod to go out of pending state.
+func waitForPodCondition(c *client.Client, ns, podName, desc string, condition podCondition) error {
-func waitForPodNotPending(c *client.Client, ns, podName string, tryFor time.Duration) error {
+	By(fmt.Sprintf("waiting up to %v for pod %s status to be %s", podStartTimeout, podName, desc))
-	trySecs := int(tryFor.Seconds())
+	for start := time.Now(); time.Since(start) < podStartTimeout; time.Sleep(5 * time.Second) {
-	for i := 0; i <= trySecs; i += 5 {
-		if i > 0 {
-			time.Sleep(5 * time.Second)
-		}
 		pod, err := c.Pods(ns).Get(podName)
-		if err != nil {
-			Logf("Get pod %s in namespace %s failed, ignoring for 5s: %v", podName, ns, err)
-			continue
-		}
-		if pod.Status.Phase != api.PodPending {
-			Logf("Saw pod %s in namespace %s out of pending state (found %q)", podName, ns, pod.Status.Phase)
-			return nil
-		}
-		Logf("Waiting for status of pod %s in namespace %s to be !%q (found %q) (%v secs)", podName, ns, api.PodPending, pod.Status.Phase, i)
-	}
-	return fmt.Errorf("Gave up waiting for status of pod %s in namespace %s to go out of pending after %d seconds", podName, ns, trySecs)
-}
-
-// waitForPodSuccess returns true if the pod reached state success, or false if it reached failure or ran too long.
-func waitForPodSuccess(c *client.Client, podName string, contName string, tryFor time.Duration) error {
-	trySecs := int(tryFor.Seconds())
-	for i := 0; i <= trySecs; i += 5 {
-		if i > 0 {
-			time.Sleep(5 * time.Second)
-		}
-		pod, err := c.Pods(api.NamespaceDefault).Get(podName)
 		if err != nil {
 			Logf("Get pod failed, ignoring for 5s: %v", err)
 			continue
 		}
+		done, err := condition(pod)
+		if done {
+			return err
+		}
+		Logf("Waiting for pod %s status to be %q (found %q) (%d secs)", podName, api.PodRunning, pod.Status.Phase, time.Since(start).Seconds())
+	}
+	return fmt.Errorf("gave up waiting for pod %s to be %s after %v", podName, podStartTimeout)
+}
+
+func waitForPodRunning(c *client.Client, podName string) error {
+	return waitForPodCondition(c, api.NamespaceDefault, podName, "running", func(pod *api.Pod) (bool, error) {
+		return (pod.Status.Phase == api.PodRunning), nil
+	})
+}
+
+// waitForPodNotPending returns an error if it took too long for the pod to go out of pending state.
+func waitForPodNotPending(c *client.Client, ns, podName string) error {
+	return waitForPodCondition(c, ns, podName, "!pending", func(pod *api.Pod) (bool, error) {
+		if pod.Status.Phase != api.PodPending {
+			Logf("Saw pod %s in namespace %s out of pending state (found %q)", podName, ns, pod.Status.Phase)
+			return true, nil
+		}
+		return false, nil
+	})
+}
+
+// waitForPodSuccess returns nil if the pod reached state success, or an error if it reached failure or ran too long.
+func waitForPodSuccess(c *client.Client, podName string, contName string) error {
+	return waitForPodCondition(c, api.NamespaceDefault, podName, "success or failure", func(pod *api.Pod) (bool, error) {
 		// Cannot use pod.Status.Phase == api.PodSucceeded/api.PodFailed due to #2632
 		ci, ok := pod.Status.Info[contName]
 		if !ok {
@@ -107,17 +103,17 @@ func waitForPodSuccess(c *client.Client, podName string, contName string, tryFor
 			if ci.State.Termination != nil {
 				if ci.State.Termination.ExitCode == 0 {
 					By("Saw pod success")
-					return nil
+					return true, nil
 				} else {
-					Logf("Saw pod failure: %+v", ci.State.Termination)
+					return true, fmt.Errorf("pod %s terminated with failure: %+v", podName, ci.State.Termination)
 				}
 				Logf("Waiting for pod %q status to be success or failure", podName)
 			} else {
 				Logf("Nil State.Termination for container %s in pod %s so far", contName, podName)
 			}
 		}
-	}
+		return false, nil
-	return fmt.Errorf("Gave up waiting for pod %q status to be success or failure after %d seconds", podName, trySecs)
+	})
 }
 
 func loadConfig() (*client.Config, error) {