vendor update for CSI 0.3.0

vendor/k8s.io/kubernetes/test/e2e/lifecycle/cluster_upgrade.go

@@ -54,6 +54,10 @@ var upgradeTests = []upgrades.Test{
 	&upgrades.AppArmorUpgradeTest{},
 }
 
+var gpuUpgradeTests = []upgrades.Test{
+	&upgrades.NvidiaGPUUpgradeTest{},
+}
+
 var statefulsetUpgradeTests = []upgrades.Test{
 	&upgrades.MySqlUpgradeTest{},
 	&upgrades.EtcdUpgradeTest{},
@@ -256,6 +260,72 @@ var _ = SIGDescribe("ingress Downgrade [Feature:IngressDowngrade]", func() {
 	})
 })
 
+var _ = SIGDescribe("gpu Upgrade [Feature:GPUUpgrade]", func() {
+	f := framework.NewDefaultFramework("gpu-upgrade")
+
+	// Create the frameworks here because we can only create them
+	// in a "Describe".
+	testFrameworks := createUpgradeFrameworks(gpuUpgradeTests)
+	Describe("master upgrade", func() {
+		It("should NOT disrupt gpu pod [Feature:GPUMasterUpgrade]", func() {
+			upgCtx, err := getUpgradeContext(f.ClientSet.Discovery(), framework.TestContext.UpgradeTarget)
+			framework.ExpectNoError(err)
+
+			testSuite := &junit.TestSuite{Name: "GPU master upgrade"}
+			gpuUpgradeTest := &junit.TestCase{Name: "[sig-node] gpu-master-upgrade", Classname: "upgrade_tests"}
+			testSuite.TestCases = append(testSuite.TestCases, gpuUpgradeTest)
+			upgradeFunc := func() {
+				start := time.Now()
+				defer finalizeUpgradeTest(start, gpuUpgradeTest)
+				target := upgCtx.Versions[1].Version.String()
+				framework.ExpectNoError(framework.MasterUpgrade(target))
+				framework.ExpectNoError(framework.CheckMasterVersion(f.ClientSet, target))
+			}
+			runUpgradeSuite(f, gpuUpgradeTests, testFrameworks, testSuite, upgCtx, upgrades.MasterUpgrade, upgradeFunc)
+		})
+	})
+	Describe("cluster upgrade", func() {
+		It("should be able to run gpu pod after upgrade [Feature:GPUClusterUpgrade]", func() {
+			upgCtx, err := getUpgradeContext(f.ClientSet.Discovery(), framework.TestContext.UpgradeTarget)
+			framework.ExpectNoError(err)
+
+			testSuite := &junit.TestSuite{Name: "GPU cluster upgrade"}
+			gpuUpgradeTest := &junit.TestCase{Name: "[sig-node] gpu-cluster-upgrade", Classname: "upgrade_tests"}
+			testSuite.TestCases = append(testSuite.TestCases, gpuUpgradeTest)
+			upgradeFunc := func() {
+				start := time.Now()
+				defer finalizeUpgradeTest(start, gpuUpgradeTest)
+				target := upgCtx.Versions[1].Version.String()
+				framework.ExpectNoError(framework.MasterUpgrade(target))
+				framework.ExpectNoError(framework.CheckMasterVersion(f.ClientSet, target))
+				framework.ExpectNoError(framework.NodeUpgrade(f, target, framework.TestContext.UpgradeImage))
+				framework.ExpectNoError(framework.CheckNodesVersions(f.ClientSet, target))
+			}
+			runUpgradeSuite(f, gpuUpgradeTests, testFrameworks, testSuite, upgCtx, upgrades.ClusterUpgrade, upgradeFunc)
+		})
+	})
+	Describe("cluster downgrade", func() {
+		It("should be able to run gpu pod after downgrade [Feature:GPUClusterDowngrade]", func() {
+			upgCtx, err := getUpgradeContext(f.ClientSet.Discovery(), framework.TestContext.UpgradeTarget)
+			framework.ExpectNoError(err)
+
+			testSuite := &junit.TestSuite{Name: "GPU cluster downgrade"}
+			gpuDowngradeTest := &junit.TestCase{Name: "[sig-node] gpu-cluster-downgrade", Classname: "upgrade_tests"}
+			testSuite.TestCases = append(testSuite.TestCases, gpuDowngradeTest)
+			upgradeFunc := func() {
+				start := time.Now()
+				defer finalizeUpgradeTest(start, gpuDowngradeTest)
+				target := upgCtx.Versions[1].Version.String()
+				framework.ExpectNoError(framework.NodeUpgrade(f, target, framework.TestContext.UpgradeImage))
+				framework.ExpectNoError(framework.CheckNodesVersions(f.ClientSet, target))
+				framework.ExpectNoError(framework.MasterUpgrade(target))
+				framework.ExpectNoError(framework.CheckMasterVersion(f.ClientSet, target))
+			}
+			runUpgradeSuite(f, gpuUpgradeTests, testFrameworks, testSuite, upgCtx, upgrades.ClusterUpgrade, upgradeFunc)
+		})
+	})
+})
+
 var _ = Describe("[sig-apps] stateful Upgrade [Feature:StatefulUpgrade]", func() {
 	f := framework.NewDefaultFramework("stateful-upgrade")
 

vendor/k8s.io/kubernetes/test/e2e/lifecycle/ha_master.go

@@ -78,7 +78,7 @@ func createNewRC(c clientset.Interface, ns string, name string) {
 }
 
 func findRegionForZone(zone string) string {
-	region, err := exec.Command("gcloud", "compute", "zones", "list", zone, "--quiet", "--format=csv[no-heading](region)").CombinedOutput()
+	region, err := exec.Command("gcloud", "compute", "zones", "list", zone, "--quiet", "--format=csv[no-heading](region)").Output()
 	framework.ExpectNoError(err)
 	if string(region) == "" {
 		framework.Failf("Region not found; zone: %s", zone)
@@ -88,7 +88,7 @@ func findRegionForZone(zone string) string {
 
 func findZonesForRegion(region string) []string {
 	output, err := exec.Command("gcloud", "compute", "zones", "list", "--filter=region="+region,
-		"--quiet", "--format=csv[no-heading](name)").CombinedOutput()
+		"--quiet", "--format=csv[no-heading](name)").Output()
 	framework.ExpectNoError(err)
 	zones := strings.Split(string(output), "\n")
 	return zones

vendor/k8s.io/kubernetes/test/e2e/lifecycle/reboot.go

@@ -219,7 +219,11 @@ func printStatusAndLogsForNotReadyPods(c clientset.Interface, ns string, podName
 func rebootNode(c clientset.Interface, provider, name, rebootCmd string) bool {
 	// Setup
 	ns := metav1.NamespaceSystem
-	ps := testutils.NewPodStore(c, ns, labels.Everything(), fields.OneTermEqualSelector(api.PodHostField, name))
+	ps, err := testutils.NewPodStore(c, ns, labels.Everything(), fields.OneTermEqualSelector(api.PodHostField, name))
+	if err != nil {
+		framework.Logf("Couldn't initialize pod store: %v", err)
+		return false
+	}
 	defer ps.Stop()
 
 	// Get the node initially.

vendor/k8s.io/kubernetes/test/e2e/lifecycle/resize_nodes.go

@@ -30,8 +30,6 @@ import (
 	. "github.com/onsi/gomega"
 )
 
-const resizeNodeReadyTimeout = 2 * time.Minute
-
 func resizeRC(c clientset.Interface, ns, name string, replicas int32) error {
 	rc, err := c.CoreV1().ReplicationControllers(ns).Get(name, metav1.GetOptions{})
 	if err != nil {
@@ -65,6 +63,7 @@ var _ = SIGDescribe("Nodes [Disruptive]", func() {
 
 	// Slow issue #13323 (8 min)
 	Describe("Resize [Slow]", func() {
+		var originalNodeCount int32
 		var skipped bool
 
 		BeforeEach(func() {
@@ -98,7 +97,8 @@ var _ = SIGDescribe("Nodes [Disruptive]", func() {
 			if err := framework.WaitForGroupSize(group, int32(framework.TestContext.CloudConfig.NumNodes)); err != nil {
 				framework.Failf("Couldn't restore the original node instance group size: %v", err)
 			}
-			if err := framework.WaitForReadyNodes(c, framework.TestContext.CloudConfig.NumNodes, 10*time.Minute); err != nil {
+
+			if err := framework.WaitForReadyNodes(c, int(originalNodeCount), 10*time.Minute); err != nil {
 				framework.Failf("Couldn't restore the original cluster size: %v", err)
 			}
 			// Many e2e tests assume that the cluster is fully healthy before they start.  Wait until
@@ -114,17 +114,20 @@ var _ = SIGDescribe("Nodes [Disruptive]", func() {
 			// Create a replication controller for a service that serves its hostname.
 			// The source for the Docker container kubernetes/serve_hostname is in contrib/for-demos/serve_hostname
 			name := "my-hostname-delete-node"
-			replicas := int32(framework.TestContext.CloudConfig.NumNodes)
-			common.NewRCByName(c, ns, name, replicas, nil)
-			err := framework.VerifyPods(c, ns, name, true, replicas)
+			numNodes, err := framework.NumberOfRegisteredNodes(c)
+			Expect(err).NotTo(HaveOccurred())
+			originalNodeCount = int32(numNodes)
+			common.NewRCByName(c, ns, name, originalNodeCount, nil)
+			err = framework.VerifyPods(c, ns, name, true, originalNodeCount)
 			Expect(err).NotTo(HaveOccurred())
 
-			By(fmt.Sprintf("decreasing cluster size to %d", replicas-1))
-			err = framework.ResizeGroup(group, replicas-1)
+			targetNumNodes := int32(framework.TestContext.CloudConfig.NumNodes - 1)
+			By(fmt.Sprintf("decreasing cluster size to %d", targetNumNodes))
+			err = framework.ResizeGroup(group, targetNumNodes)
 			Expect(err).NotTo(HaveOccurred())
-			err = framework.WaitForGroupSize(group, replicas-1)
+			err = framework.WaitForGroupSize(group, targetNumNodes)
 			Expect(err).NotTo(HaveOccurred())
-			err = framework.WaitForReadyNodes(c, int(replicas-1), 10*time.Minute)
+			err = framework.WaitForReadyNodes(c, int(originalNodeCount-1), 10*time.Minute)
 			Expect(err).NotTo(HaveOccurred())
 
 			By("waiting 1 minute for the watch in the podGC to catch up, remove any pods scheduled on " +
@@ -132,7 +135,7 @@ var _ = SIGDescribe("Nodes [Disruptive]", func() {
 			time.Sleep(time.Minute)
 
 			By("verifying whether the pods from the removed node are recreated")
-			err = framework.VerifyPods(c, ns, name, true, replicas)
+			err = framework.VerifyPods(c, ns, name, true, originalNodeCount)
 			Expect(err).NotTo(HaveOccurred())
 		})
 
@@ -142,23 +145,26 @@ var _ = SIGDescribe("Nodes [Disruptive]", func() {
 			// The source for the Docker container kubernetes/serve_hostname is in contrib/for-demos/serve_hostname
 			name := "my-hostname-add-node"
 			common.NewSVCByName(c, ns, name)
-			replicas := int32(framework.TestContext.CloudConfig.NumNodes)
-			common.NewRCByName(c, ns, name, replicas, nil)
-			err := framework.VerifyPods(c, ns, name, true, replicas)
+			numNodes, err := framework.NumberOfRegisteredNodes(c)
+			Expect(err).NotTo(HaveOccurred())
+			originalNodeCount = int32(numNodes)
+			common.NewRCByName(c, ns, name, originalNodeCount, nil)
+			err = framework.VerifyPods(c, ns, name, true, originalNodeCount)
 			Expect(err).NotTo(HaveOccurred())
 
-			By(fmt.Sprintf("increasing cluster size to %d", replicas+1))
-			err = framework.ResizeGroup(group, replicas+1)
+			targetNumNodes := int32(framework.TestContext.CloudConfig.NumNodes + 1)
+			By(fmt.Sprintf("increasing cluster size to %d", targetNumNodes))
+			err = framework.ResizeGroup(group, targetNumNodes)
 			Expect(err).NotTo(HaveOccurred())
-			err = framework.WaitForGroupSize(group, replicas+1)
+			err = framework.WaitForGroupSize(group, targetNumNodes)
 			Expect(err).NotTo(HaveOccurred())
-			err = framework.WaitForReadyNodes(c, int(replicas+1), 10*time.Minute)
+			err = framework.WaitForReadyNodes(c, int(originalNodeCount+1), 10*time.Minute)
 			Expect(err).NotTo(HaveOccurred())
 
-			By(fmt.Sprintf("increasing size of the replication controller to %d and verifying all pods are running", replicas+1))
-			err = resizeRC(c, ns, name, replicas+1)
+			By(fmt.Sprintf("increasing size of the replication controller to %d and verifying all pods are running", originalNodeCount+1))
+			err = resizeRC(c, ns, name, originalNodeCount+1)
 			Expect(err).NotTo(HaveOccurred())
-			err = framework.VerifyPods(c, ns, name, true, replicas+1)
+			err = framework.VerifyPods(c, ns, name, true, originalNodeCount+1)
 			Expect(err).NotTo(HaveOccurred())
 		})
 	})

vendor/k8s.io/kubernetes/test/e2e/lifecycle/restart.go

@@ -26,6 +26,7 @@ import (
 	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/util/wait"
 	kubepod "k8s.io/kubernetes/pkg/kubelet/pod"
+	"k8s.io/kubernetes/test/e2e/common"
 	"k8s.io/kubernetes/test/e2e/framework"
 	testutils "k8s.io/kubernetes/test/utils"
 
@@ -55,10 +56,18 @@ func filterIrrelevantPods(pods []*v1.Pod) []*v1.Pod {
 	return results
 }
 
+func nodeNames(nodes []v1.Node) []string {
+	result := make([]string, 0, len(nodes))
+	for i := range nodes {
+		result = append(result, nodes[i].Name)
+	}
+	return result
+}
+
 var _ = SIGDescribe("Restart [Disruptive]", func() {
 	f := framework.NewDefaultFramework("restart")
 	var ps *testutils.PodStore
-	var originalNodeNames []string
+	var originalNodes []v1.Node
 	var originalPodNames []string
 	var numNodes int
 	var systemNamespace string
@@ -67,15 +76,17 @@ var _ = SIGDescribe("Restart [Disruptive]", func() {
 		// This test requires the ability to restart all nodes, so the provider
 		// check must be identical to that call.
 		framework.SkipUnlessProviderIs("gce", "gke")
-		ps = testutils.NewPodStore(f.ClientSet, metav1.NamespaceSystem, labels.Everything(), fields.Everything())
-		numNodes = framework.TestContext.CloudConfig.NumNodes
+		var err error
+		ps, err = testutils.NewPodStore(f.ClientSet, metav1.NamespaceSystem, labels.Everything(), fields.Everything())
+		Expect(err).NotTo(HaveOccurred())
+		numNodes, err = framework.NumberOfRegisteredNodes(f.ClientSet)
+		Expect(err).NotTo(HaveOccurred())
 		systemNamespace = metav1.NamespaceSystem
 
 		By("ensuring all nodes are ready")
-		var err error
-		originalNodeNames, err = framework.CheckNodesReady(f.ClientSet, framework.NodeReadyInitialTimeout, numNodes)
+		originalNodes, err = framework.CheckNodesReady(f.ClientSet, numNodes, framework.NodeReadyInitialTimeout)
 		Expect(err).NotTo(HaveOccurred())
-		framework.Logf("Got the following nodes before restart: %v", originalNodeNames)
+		framework.Logf("Got the following nodes before restart: %v", nodeNames(originalNodes))
 
 		By("ensuring all pods are running and ready")
 		allPods := ps.List()
@@ -99,20 +110,20 @@ var _ = SIGDescribe("Restart [Disruptive]", func() {
 
 	It("should restart all nodes and ensure all nodes and pods recover", func() {
 		By("restarting all of the nodes")
-		err := restartNodes(f, originalNodeNames)
+		err := common.RestartNodes(f.ClientSet, originalNodes)
 		Expect(err).NotTo(HaveOccurred())
 
 		By("ensuring all nodes are ready after the restart")
-		nodeNamesAfter, err := framework.CheckNodesReady(f.ClientSet, framework.RestartNodeReadyAgainTimeout, numNodes)
+		nodesAfter, err := framework.CheckNodesReady(f.ClientSet, numNodes, framework.RestartNodeReadyAgainTimeout)
 		Expect(err).NotTo(HaveOccurred())
-		framework.Logf("Got the following nodes after restart: %v", nodeNamesAfter)
+		framework.Logf("Got the following nodes after restart: %v", nodeNames(nodesAfter))
 
 		// Make sure that we have the same number of nodes. We're not checking
 		// that the names match because that's implementation specific.
 		By("ensuring the same number of nodes exist after the restart")
-		if len(originalNodeNames) != len(nodeNamesAfter) {
+		if len(originalNodes) != len(nodesAfter) {
 			framework.Failf("Had %d nodes before nodes were restarted, but now only have %d",
-				len(originalNodeNames), len(nodeNamesAfter))
+				len(originalNodes), len(nodesAfter))
 		}
 
 		// Make sure that we have the same number of pods. We're not checking
@@ -158,41 +169,3 @@ func waitForNPods(ps *testutils.PodStore, expect int, timeout time.Duration) ([]
 	}
 	return podNames, nil
 }
-
-func restartNodes(f *framework.Framework, nodeNames []string) error {
-	// List old boot IDs.
-	oldBootIDs := make(map[string]string)
-	for _, name := range nodeNames {
-		node, err := f.ClientSet.CoreV1().Nodes().Get(name, metav1.GetOptions{})
-		if err != nil {
-			return fmt.Errorf("error getting node info before reboot: %s", err)
-		}
-		oldBootIDs[name] = node.Status.NodeInfo.BootID
-	}
-	// Reboot the nodes.
-	args := []string{
-		"compute",
-		fmt.Sprintf("--project=%s", framework.TestContext.CloudConfig.ProjectID),
-		"instances",
-		"reset",
-	}
-	args = append(args, nodeNames...)
-	args = append(args, fmt.Sprintf("--zone=%s", framework.TestContext.CloudConfig.Zone))
-	stdout, stderr, err := framework.RunCmd("gcloud", args...)
-	if err != nil {
-		return fmt.Errorf("error restarting nodes: %s\nstdout: %s\nstderr: %s", err, stdout, stderr)
-	}
-	// Wait for their boot IDs to change.
-	for _, name := range nodeNames {
-		if err := wait.Poll(30*time.Second, 5*time.Minute, func() (bool, error) {
-			node, err := f.ClientSet.CoreV1().Nodes().Get(name, metav1.GetOptions{})
-			if err != nil {
-				return false, fmt.Errorf("error getting node info after reboot: %s", err)
-			}
-			return node.Status.NodeInfo.BootID != oldBootIDs[name], nil
-		}); err != nil {
-			return fmt.Errorf("error waiting for node %s boot ID to change: %s", name, err)
-		}
-	}
-	return nil
-}