Ensure tests which rely on SSH behavior are skipped if unavailable

A number of tests rely on SSH behavior but are not marked with
SkipUnlessSSHKeysPresent(). This means that the tests will run and
hang. This can be confusing for users because they may not know why
the tests failed. Instead, we should be reporting on the known issue
and skipping the test (hence why the helper Skip... function exists).

test/e2e/apimachinery/etcd_failure.go

@@ -44,6 +44,7 @@ var _ = SIGDescribe("Etcd failure [Disruptive]", func() {
 		// ... so the provider check should be identical to the intersection of
 		// providers that provide those capabilities.
 		framework.SkipUnlessProviderIs("gce")
+		framework.SkipUnlessSSHKeyPresent()
 
 		err := framework.RunRC(testutils.RCConfig{
 			Client:    f.ClientSet,

test/e2e/apps/network_partition.go

@@ -130,6 +130,7 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 		ginkgo.Context("should return to running and ready state after network partition is healed", func() {
 			ginkgo.BeforeEach(func() {
 				framework.SkipUnlessNodeCountIsAtLeast(2)
+				framework.SkipUnlessSSHKeyPresent()
 			})
 
 			// What happens in this test:
@@ -238,6 +239,7 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 	framework.KubeDescribe("[ReplicationController]", func() {
 		ginkgo.It("should recreate pods scheduled on the unreachable node "+
 			"AND allow scheduling of pods on a node after it rejoins the cluster", func() {
+			framework.SkipUnlessSSHKeyPresent()
 
 			// 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
@@ -304,6 +306,8 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 		})
 
 		ginkgo.It("should eagerly create replacement pod during network partition when termination grace is non-zero", func() {
+			framework.SkipUnlessSSHKeyPresent()
+
 			// 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-net"
@@ -394,6 +398,8 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 		})
 
 		ginkgo.It("should not reschedule stateful pods if there is a network partition [Slow] [Disruptive]", func() {
+			framework.SkipUnlessSSHKeyPresent()
+
 			ps := e2esset.NewStatefulSet(psName, ns, headlessSvcName, 3, []v1.VolumeMount{}, []v1.VolumeMount{}, labels)
 			_, err := c.AppsV1().StatefulSets(ns).Create(ps)
 			framework.ExpectNoError(err)
@@ -425,6 +431,8 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 
 	framework.KubeDescribe("[Job]", func() {
 		ginkgo.It("should create new pods when node is partitioned", func() {
+			framework.SkipUnlessSSHKeyPresent()
+
 			parallelism := int32(2)
 			completions := int32(4)
 			backoffLimit := int32(6) // default value
@@ -482,6 +490,7 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 			// 3. After enough time passess all Pods are evicted from the given Node
 			ginkgo.It("[Feature:TaintEviction] All pods on the unreachable node should be marked as NotReady upon the node turn NotReady "+
 				"AND all pods should be evicted after eviction timeout passes", func() {
+				framework.SkipUnlessSSHKeyPresent()
 				ginkgo.By("choose a node - we will block all network traffic on this node")
 				var podOpts metav1.ListOptions
 				nodes := framework.GetReadySchedulableNodesOrDie(c)

test/e2e/autoscaling/cluster_size_autoscaling.go

@@ -685,6 +685,7 @@ var _ = SIGDescribe("Cluster size autoscaling [Slow]", func() {
 
 	ginkgo.It("should correctly scale down after a node is not needed and one node is broken [Feature:ClusterSizeAutoscalingScaleDown]",
 		func() {
+			framework.SkipUnlessSSHKeyPresent()
 			framework.TestUnderTemporaryNetworkFailure(c, "default", getAnyNode(c), func() { simpleScaleDownTest(1) })
 		})
 
@@ -876,6 +877,8 @@ var _ = SIGDescribe("Cluster size autoscaling [Slow]", func() {
 	})
 
 	ginkgo.It("Shouldn't perform scale up operation and should list unhealthy status if most of the cluster is broken[Feature:ClusterSizeAutoscalingScaleUp]", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		clusterSize := nodeCount
 		for clusterSize < unhealthyClusterThreshold+1 {
 			clusterSize = manuallyIncreaseClusterSize(f, originalSizes)

test/e2e/network/service.go

@@ -404,6 +404,7 @@ var _ = SIGDescribe("Services", func() {
 	ginkgo.It("should work after restarting kube-proxy [Disruptive]", func() {
 		// TODO: use the ServiceTestJig here
 		framework.SkipUnlessProviderIs("gce", "gke")
+		framework.SkipUnlessSSHKeyPresent()
 
 		ns := f.Namespace.Name
 		numPods, servicePort := 3, defaultServeHostnameServicePort
@@ -460,6 +461,7 @@ var _ = SIGDescribe("Services", func() {
 	ginkgo.It("should work after restarting apiserver [Disruptive]", func() {
 		// TODO: use the ServiceTestJig here
 		framework.SkipUnlessProviderIs("gce", "gke")
+		framework.SkipUnlessSSHKeyPresent()
 
 		ns := f.Namespace.Name
 		numPods, servicePort := 3, 80
@@ -1658,6 +1660,8 @@ var _ = SIGDescribe("Services", func() {
 		// This test is for clusters on GCE.
 		// (It restarts kube-controller-manager, which we don't support on GKE)
 		framework.SkipUnlessProviderIs("gce")
+		framework.SkipUnlessSSHKeyPresent()
+
 		clusterID, err := gce.GetClusterID(cs)
 		if err != nil {
 			e2elog.Failf("framework.GetClusterID(cs) = _, %v; want nil", err)

test/e2e/node/node_problem_detector.go

@@ -56,6 +56,8 @@ var _ = SIGDescribe("NodeProblemDetector [DisabledForLargeClusters]", func() {
 	})
 
 	ginkgo.It("should run without error", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		ginkgo.By("Getting all nodes and their SSH-able IP addresses")
 		nodes := framework.GetReadySchedulableNodesOrDie(f.ClientSet)
 		gomega.Expect(len(nodes.Items)).NotTo(gomega.BeZero())

test/e2e/scalability/density.go

@@ -503,6 +503,9 @@ var _ = SIGDescribe("Density", func() {
 	f.NamespaceDeletionTimeout = time.Hour
 
 	ginkgo.BeforeEach(func() {
+		// Gathering the metrics currently uses a path which uses SSH.
+		framework.SkipUnlessSSHKeyPresent()
+
 		var err error
 		c = f.ClientSet
 		ns = f.Namespace.Name

test/e2e/scalability/load.go

@@ -109,6 +109,10 @@ var _ = SIGDescribe("Load capacity", func() {
 	var testPhaseDurations *timer.TestPhaseTimer
 	var profileGathererStopCh chan struct{}
 
+	ginkgo.BeforeEach(func() {
+		framework.SkipUnlessSSHKeyPresent()
+	})
+
 	// Gathers metrics before teardown
 	// TODO add flag that allows to skip cleanup on failure
 	ginkgo.AfterEach(func() {

test/e2e/scheduling/taint_based_evictions.go

@@ -76,6 +76,8 @@ var _ = SIGDescribe("TaintBasedEvictions [Serial]", func() {
 	// 5. node lifecycle manager generate a status change: [NodeReady=true, status=ConditionTrue]
 	// 6. node.kubernetes.io/unreachable=:NoExecute taint is taken off the node
 	ginkgo.It("Checks that the node becomes unreachable", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		// find an available node
 		nodeName := GetNodeThatCanRunPod(f)
 		ginkgo.By("Finding an available node " + nodeName)

test/e2e/storage/detach_mounted.go

@@ -64,6 +64,8 @@ var _ = utils.SIGDescribe("Detaching volumes", func() {
 	})
 
 	ginkgo.It("should not work when mount is in progress [Slow]", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		driver := "attachable-with-long-mount"
 		driverInstallAs := driver + "-" + suffix
 

test/e2e/storage/flexvolume_online_resize.go

@@ -116,6 +116,8 @@ var _ = utils.SIGDescribe("Mounted flexvolume volume expand [Slow] [Feature:Expa
 	})
 
 	ginkgo.It("should be resizable when mounted", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		driver := "dummy-attachable"
 
 		node := nodeList.Items[0]

test/e2e/storage/nfs_persistent_volume-disruptive.go

@@ -175,6 +175,8 @@ var _ = utils.SIGDescribe("NFSPersistentVolumes[Disruptive][Flaky]", func() {
 		})
 
 		ginkgo.It("should delete a bound PVC from a clientPod, restart the kube-control-manager, and ensure the kube-controller-manager does not crash", func() {
+			framework.SkipUnlessSSHKeyPresent()
+
 			ginkgo.By("Deleting PVC for volume 2")
 			err = framework.DeletePersistentVolumeClaim(c, pvc2.Name, ns)
 			framework.ExpectNoError(err)

test/e2e/storage/vsphere/persistent_volumes-vsphere.go

@@ -179,6 +179,7 @@ var _ = utils.SIGDescribe("PersistentVolumes:vsphere", func() {
 		3. Verify that written file is accessible after kubelet restart
 	*/
 	ginkgo.It("should test that a file written to the vspehre volume mount before kubelet restart can be read after restart [Disruptive]", func() {
+		framework.SkipUnlessSSHKeyPresent()
 		utils.TestKubeletRestartsAndRestoresMount(c, f, clientPod)
 	})
 
@@ -194,6 +195,7 @@ var _ = utils.SIGDescribe("PersistentVolumes:vsphere", func() {
 		5. Verify that volume mount not to be found.
 	*/
 	ginkgo.It("should test that a vspehre volume mounted to a pod that is deleted while the kubelet is down unmounts when the kubelet returns [Disruptive]", func() {
+		framework.SkipUnlessSSHKeyPresent()
 		utils.TestVolumeUnmountsFromDeletedPod(c, f, clientPod)
 	})
 

test/e2e/storage/vsphere/vsphere_volume_master_restart.go

@@ -81,6 +81,8 @@ var _ = utils.SIGDescribe("Volume Attach Verify [Feature:vsphere][Serial][Disrup
 	})
 
 	ginkgo.It("verify volume remains attached after master kubelet restart", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		// Create pod on each node
 		for i := 0; i < numNodes; i++ {
 			ginkgo.By(fmt.Sprintf("%d: Creating a test vsphere volume", i))

test/e2e/storage/vsphere/vsphere_volume_vpxd_restart.go

@@ -99,6 +99,8 @@ var _ = utils.SIGDescribe("Verify Volume Attach Through vpxd Restart [Feature:vs
 	})
 
 	ginkgo.It("verify volume remains attached through vpxd restart", func() {
+		framework.SkipUnlessSSHKeyPresent()
+
 		for vcHost, nodes := range vcNodesMap {
 			var (
 				volumePaths  []string