Merge pull request #42667 from vmware/e2eTestUpdate-v5

Automatic merge from submit-queue (batch tested with PRs 42667, 43923)

Adding test to perform volume operations storm

**What this PR does / why we need it**:
Adding new test to perform volume operations storm
Test Steps
1. Create storage class for thin Provisioning.
2. Create 30 PVCs using above storage class in annotation, requesting 2 GB files.
3. Wait until all disks are ready and all PVs and PVCs get bind. (**CreateVolume** storm)
4. Create pod to mount volumes using PVCs created in step 2. (**AttachDisk** storm)
5. Wait for pod status to be running.
6. Verify all volumes accessible and available in the pod.
7. Delete pod.
8. wait until volumes gets detached. (**DetachDisk** storm)
9. Delete all PVCs. This should delete all Disks. (**DeleteVolume** storm)
10. Delete storage class.

This test will help validate issue reported at https://github.com/vmware/kubernetes/issues/71

**Which issue this PR fixes**
fixes #

**Special notes for your reviewer**:
executed test on 1.5.3 release with `VOLUME_OPS_SCALE` set to `5`
Will execute test with the changes made on PR - https://github.com/kubernetes/kubernetes/pull/42422, with the  `VOLUME_OPS_SCALE` set to `30`

**Release note**:

```release-note
None
```

cc: @abrarshivani @BaluDontu @tusharnt @pdhamdhere @luomiao @kerneltime

test/e2e/BUILD

@@ -89,6 +89,7 @@ go_library(
         "volumes.go",
         "vsphere_utils.go",
         "vsphere_volume_diskformat.go",
+        "vsphere_volume_ops_storm.go",
         "vsphere_volume_placement.go",
     ],
     tags = ["automanaged"],

test/e2e/framework/pv_util.go

@@ -422,25 +422,6 @@ func DeletePodWithWait(f *Framework, c clientset.Interface, pod *v1.Pod) {
 	Logf("Ignore \"not found\" error above. Pod %v successfully deleted", pod.Name)
 }
 
-// Create the test pod, wait for (hopefully) success, and then delete the pod.
-func CreateWaitAndDeletePod(f *Framework, c clientset.Interface, ns string, claimName string) {
-
-	Logf("Creating nfs test pod")
-
-	// Make pod spec
-	pod := MakeWritePod(ns, claimName)
-
-	// Instantiate pod (Create)
-	runPod, err := c.CoreV1().Pods(ns).Create(pod)
-	Expect(err).NotTo(HaveOccurred())
-	Expect(runPod).NotTo(BeNil())
-
-	defer DeletePodWithWait(f, c, runPod)
-
-	// Wait for the test pod to complete its lifecycle
-	testPodSuccessOrFail(c, ns, runPod)
-}
-
 // Sanity check for GCE testing.  Verify the persistent disk attached to the node.
 func VerifyGCEDiskAttached(diskName string, nodeName types.NodeName) bool {
 	gceCloud, err := GetGCECloud()
@@ -532,75 +513,6 @@ func MakePersistentVolumeClaim(cfg PersistentVolumeClaimConfig, ns string) *v1.P
 	}
 }
 
-// Returns a pod definition based on the namespace. The pod references the PVC's
-// name.
-func MakeWritePod(ns string, pvcName string) *v1.Pod {
-	return MakePod(ns, pvcName, true, "touch /mnt/SUCCESS && (id -G | grep -E '\\b777\\b')")
-}
-
-// Returns a pod definition based on the namespace. The pod references the PVC's
-// name.  A slice of BASH commands can be supplied as args to be run by the pod
-func MakePod(ns string, pvcName string, isPrivileged bool, command string) *v1.Pod {
-
-	if len(command) == 0 {
-		command = "while true; do sleep 1; done"
-	}
-	return &v1.Pod{
-		TypeMeta: metav1.TypeMeta{
-			Kind:       "Pod",
-			APIVersion: api.Registry.GroupOrDie(v1.GroupName).GroupVersion.String(),
-		},
-		ObjectMeta: metav1.ObjectMeta{
-			GenerateName: "pvc-tester-",
-			Namespace:    ns,
-		},
-		Spec: v1.PodSpec{
-			Containers: []v1.Container{
-				{
-					Name:    "write-pod",
-					Image:   "gcr.io/google_containers/busybox:1.24",
-					Command: []string{"/bin/sh"},
-					Args:    []string{"-c", command},
-					VolumeMounts: []v1.VolumeMount{
-						{
-							Name:      pvcName,
-							MountPath: "/mnt",
-						},
-					},
-					SecurityContext: &v1.SecurityContext{
-						Privileged: &isPrivileged,
-					},
-				},
-			},
-			RestartPolicy: v1.RestartPolicyOnFailure,
-			Volumes: []v1.Volume{
-				{
-					Name: pvcName,
-					VolumeSource: v1.VolumeSource{
-						PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{
-							ClaimName: pvcName,
-						},
-					},
-				},
-			},
-		},
-	}
-}
-
-// Define and create a pod with a mounted PV.  Pod runs infinite loop until killed.
-func CreateClientPod(c clientset.Interface, ns string, pvc *v1.PersistentVolumeClaim) *v1.Pod {
-	clientPod := MakePod(ns, pvc.Name, true, "")
-	clientPod, err := c.CoreV1().Pods(ns).Create(clientPod)
-	Expect(err).NotTo(HaveOccurred())
-
-	// Verify the pod is running before returning it
-	err = WaitForPodRunningInNamespace(c, clientPod)
-	Expect(err).NotTo(HaveOccurred())
-	clientPod, err = c.CoreV1().Pods(ns).Get(clientPod.Name, metav1.GetOptions{})
-	Expect(apierrs.IsNotFound(err)).To(BeFalse())
-	return clientPod
-}
-
 func CreatePDWithRetry() (string, error) {
 	newDiskName := ""
 	var err error
@@ -703,3 +615,105 @@ func deletePD(pdName string) error {
 		return fmt.Errorf("Provider does not support volume deletion")
 	}
 }
+
+// Create the test pod, wait for (hopefully) success, and then delete the pod.
+func CreateWaitAndDeletePod(f *Framework, c clientset.Interface, ns string, pvc *v1.PersistentVolumeClaim) {
+	Logf("Creating nfs test pod")
+	// Make pod spec
+	pod := MakeWritePod(ns, pvc)
+
+	// Instantiate pod (Create)
+	runPod, err := c.CoreV1().Pods(ns).Create(pod)
+	Expect(err).NotTo(HaveOccurred())
+	Expect(runPod).NotTo(BeNil())
+
+	defer DeletePodWithWait(f, c, runPod)
+
+	// Wait for the test pod to complete its lifecycle
+	testPodSuccessOrFail(c, ns, runPod)
+}
+
+// Returns a pod definition based on the namespace. The pod references the PVC's
+// name.
+func MakeWritePod(ns string, pvc *v1.PersistentVolumeClaim) *v1.Pod {
+	return MakePod(ns, []*v1.PersistentVolumeClaim{pvc}, true, "touch /mnt/volume1/SUCCESS && (id -G | grep -E '\\b777\\b')")
+}
+
+// Returns a pod definition based on the namespace. The pod references the PVC's
+// name.  A slice of BASH commands can be supplied as args to be run by the pod
+func MakePod(ns string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bool, command string) *v1.Pod {
+	if len(command) == 0 {
+		command = "while true; do sleep 1; done"
+	}
+	podSpec := &v1.Pod{
+		TypeMeta: metav1.TypeMeta{
+			Kind:       "Pod",
+			APIVersion: api.Registry.GroupOrDie(v1.GroupName).GroupVersion.String(),
+		},
+		ObjectMeta: metav1.ObjectMeta{
+			GenerateName: "pvc-tester-",
+			Namespace:    ns,
+		},
+		Spec: v1.PodSpec{
+			Containers: []v1.Container{
+				{
+					Name:    "write-pod",
+					Image:   "gcr.io/google_containers/busybox:1.24",
+					Command: []string{"/bin/sh"},
+					Args:    []string{"-c", command},
+					SecurityContext: &v1.SecurityContext{
+						Privileged: &isPrivileged,
+					},
+				},
+			},
+			RestartPolicy: v1.RestartPolicyOnFailure,
+		},
+	}
+	var volumeMounts = make([]v1.VolumeMount, len(pvclaims))
+	var volumes = make([]v1.Volume, len(pvclaims))
+	for index, pvclaim := range pvclaims {
+		volumename := fmt.Sprintf("volume%v", index+1)
+		volumeMounts[index] = v1.VolumeMount{Name: volumename, MountPath: "/mnt/" + volumename}
+		volumes[index] = v1.Volume{Name: volumename, VolumeSource: v1.VolumeSource{PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{ClaimName: pvclaim.Name, ReadOnly: false}}}
+	}
+	podSpec.Spec.Containers[0].VolumeMounts = volumeMounts
+	podSpec.Spec.Volumes = volumes
+	return podSpec
+}
+
+// create pod with given claims
+func CreatePod(client clientset.Interface, namespace string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bool, command string) *v1.Pod {
+	podSpec := MakePod(namespace, pvclaims, isPrivileged, command)
+	pod, err := client.CoreV1().Pods(namespace).Create(podSpec)
+	Expect(err).NotTo(HaveOccurred())
+
+	// Waiting for pod to be running
+	Expect(WaitForPodNameRunningInNamespace(client, pod.Name, namespace)).To(Succeed())
+
+	// get fresh pod info
+	pod, err = client.CoreV1().Pods(namespace).Get(pod.Name, metav1.GetOptions{})
+	Expect(err).NotTo(HaveOccurred())
+	return pod
+}
+
+// Define and create a pod with a mounted PV.  Pod runs infinite loop until killed.
+func CreateClientPod(c clientset.Interface, ns string, pvc *v1.PersistentVolumeClaim) *v1.Pod {
+	return CreatePod(c, ns, []*v1.PersistentVolumeClaim{pvc}, true, "")
+}
+
+// wait until all pvcs phase set to bound
+func WaitForPVClaimBoundPhase(client clientset.Interface, pvclaims []*v1.PersistentVolumeClaim) []*v1.PersistentVolume {
+	var persistentvolumes = make([]*v1.PersistentVolume, len(pvclaims))
+	for index, claim := range pvclaims {
+		err := WaitForPersistentVolumeClaimPhase(v1.ClaimBound, client, claim.Namespace, claim.Name, Poll, ClaimProvisionTimeout)
+		Expect(err).NotTo(HaveOccurred())
+		// Get new copy of the claim
+		claim, err := client.CoreV1().PersistentVolumeClaims(claim.Namespace).Get(claim.Name, metav1.GetOptions{})
+		Expect(err).NotTo(HaveOccurred())
+
+		// Get the bounded PV
+		persistentvolumes[index], err = client.CoreV1().PersistentVolumes().Get(claim.Spec.VolumeName, metav1.GetOptions{})
+		Expect(err).NotTo(HaveOccurred())
+	}
+	return persistentvolumes
+}

test/e2e/persistent_volumes-disruptive.go

@@ -202,7 +202,7 @@ func testVolumeUnmountsFromDeletedPod(c clientset.Interface, f *framework.Framew
 // initTestCase initializes spec resources (pv, pvc, and pod) and returns pointers to be consumed by the test
 func initTestCase(f *framework.Framework, c clientset.Interface, pvConfig framework.PersistentVolumeConfig, pvcConfig framework.PersistentVolumeClaimConfig, ns, nodeName string) (*v1.Pod, *v1.PersistentVolume, *v1.PersistentVolumeClaim) {
 	pv, pvc := framework.CreatePVPVC(c, pvConfig, pvcConfig, ns, false)
-	pod := framework.MakePod(ns, pvc.Name, true, "")
+	pod := framework.MakePod(ns, []*v1.PersistentVolumeClaim{pvc}, true, "")
 	pod.Spec.NodeName = nodeName
 	framework.Logf("Creating nfs client Pod %s on node %s", pod.Name, nodeName)
 	pod, err := c.CoreV1().Pods(ns).Create(pod)

test/e2e/persistent_volumes.go

@@ -40,7 +40,7 @@ func completeTest(f *framework.Framework, c clientset.Interface, ns string, pv *
 	// 2. create the nfs writer pod, test if the write was successful,
 	//    then delete the pod and verify that it was deleted
 	By("Checking pod has write access to PersistentVolume")
-	framework.CreateWaitAndDeletePod(f, c, ns, pvc.Name)
+	framework.CreateWaitAndDeletePod(f, c, ns, pvc)
 
 	// 3. delete the PVC, wait for PV to become "Released"
 	By("Deleting the PVC to invoke the reclaim policy.")
@@ -66,7 +66,7 @@ func completeMultiTest(f *framework.Framework, c clientset.Interface, ns string,
 		_, found := pvols[pvc.Spec.VolumeName]
 		Expect(found).To(BeTrue())
 		// TODO: currently a serialized test of each PV
-		framework.CreateWaitAndDeletePod(f, c, pvcKey.Namespace, pvcKey.Name)
+		framework.CreateWaitAndDeletePod(f, c, pvcKey.Namespace, pvc)
 	}
 
 	// 2. delete each PVC, wait for its bound PV to reach `expectedPhase`
@@ -262,7 +262,7 @@ var _ = framework.KubeDescribe("PersistentVolumes [Volume]", func() {
 			// (and test) succeed.
 			It("should test that a PV becomes Available and is clean after the PVC is deleted. [Volume] [Flaky]", func() {
 				By("Writing to the volume.")
-				pod := framework.MakeWritePod(ns, pvc.Name)
+				pod := framework.MakeWritePod(ns, pvc)
 				pod, err := c.CoreV1().Pods(ns).Create(pod)
 				Expect(err).NotTo(HaveOccurred())
 				err = framework.WaitForPodSuccessInNamespace(c, pod.Name, ns)
@@ -279,8 +279,7 @@ var _ = framework.KubeDescribe("PersistentVolumes [Volume]", func() {
 				// If a file is detected in /mnt, fail the pod and do not restart it.
 				By("Verifying the mount has been cleaned.")
 				mount := pod.Spec.Containers[0].VolumeMounts[0].MountPath
-				pod = framework.MakePod(ns, pvc.Name, true, fmt.Sprintf("[ $(ls -A %s | wc -l) -eq 0 ] && exit 0 || exit 1", mount))
+				pod = framework.MakePod(ns, []*v1.PersistentVolumeClaim{pvc}, true, fmt.Sprintf("[ $(ls -A %s | wc -l) -eq 0 ] && exit 0 || exit 1", mount))
-
 				pod, err = c.CoreV1().Pods(ns).Create(pod)
 				Expect(err).NotTo(HaveOccurred())
 				err = framework.WaitForPodSuccessInNamespace(c, pod.Name, ns)

test/e2e/upgrades/persistent_volumes.go

@@ -99,7 +99,7 @@ func (t *PersistentVolumeUpgradeTest) Teardown(f *framework.Framework) {
 
 // testPod creates a pod that consumes a pv and prints it out. The output is then verified.
 func (t *PersistentVolumeUpgradeTest) testPod(f *framework.Framework, cmd string) {
-	pod := framework.MakePod(f.Namespace.Name, t.pvc.Name, false, cmd)
+	pod := framework.CreatePod(f.ClientSet, f.Namespace.Name, []*v1.PersistentVolumeClaim{t.pvc}, false, cmd)
 	expectedOutput := []string{pvTestData}
 	f.TestContainerOutput("pod consumes pv", pod, 0, expectedOutput)
 }

test/e2e/vsphere_utils.go

@@ -17,15 +17,16 @@ limitations under the License.
 package e2e
 
 import (
+	"fmt"
+	"path/filepath"
 	"strconv"
 	"time"
 
-	"fmt"
-
 	. "github.com/onsi/gomega"
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/types"
+	k8stype "k8s.io/apimachinery/pkg/types"
 	"k8s.io/apimachinery/pkg/util/uuid"
 	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/kubernetes/pkg/api/v1"
@@ -332,3 +333,19 @@ func createEmptyFilesOnVSphereVolume(namespace string, podName string, filePaths
 		Expect(err).NotTo(HaveOccurred())
 	}
 }
+
+// verify volumes are attached to the node and are accessible in pod
+func verifyVSphereVolumesAccessible(pod *v1.Pod, persistentvolumes []*v1.PersistentVolume, vsp *vsphere.VSphere) {
+	nodeName := pod.Spec.NodeName
+	namespace := pod.Namespace
+	for index, pv := range persistentvolumes {
+		// Verify disks are attached to the node
+		isAttached, err := verifyVSphereDiskAttached(vsp, pv.Spec.VsphereVolume.VolumePath, k8stype.NodeName(nodeName))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(isAttached).To(BeTrue(), fmt.Sprintf("disk %v is not attached with the node", pv.Spec.VsphereVolume.VolumePath))
+		// Verify Volumes are accessible
+		filepath := filepath.Join("/mnt/", fmt.Sprintf("volume%v", index+1), "/emptyFile.txt")
+		_, err = framework.LookForStringInPodExec(namespace, pod.Name, []string{"/bin/touch", filepath}, "", time.Minute)
+		Expect(err).NotTo(HaveOccurred())
+	}
+}

test/e2e/vsphere_volume_ops_storm.go

@@ -0,0 +1,123 @@
+/*
+Copyright 2017 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package e2e
+
+import (
+	"fmt"
+	"os"
+	"strconv"
+
+	. "github.com/onsi/ginkgo"
+	. "github.com/onsi/gomega"
+	k8stype "k8s.io/apimachinery/pkg/types"
+	"k8s.io/kubernetes/pkg/api/v1"
+	storage "k8s.io/kubernetes/pkg/apis/storage/v1beta1"
+	"k8s.io/kubernetes/pkg/client/clientset_generated/clientset"
+	vsphere "k8s.io/kubernetes/pkg/cloudprovider/providers/vsphere"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+/*
+	Test to perform Disk Ops storm.
+
+	Steps
+    	1. Create storage class for thin Provisioning.
+    	2. Create 30 PVCs using above storage class in annotation, requesting 2 GB files.
+    	3. Wait until all disks are ready and all PVs and PVCs get bind. (CreateVolume storm)
+    	4. Create pod to mount volumes using PVCs created in step 2. (AttachDisk storm)
+    	5. Wait for pod status to be running.
+    	6. Verify all volumes accessible and available in the pod.
+    	7. Delete pod.
+    	8. wait until volumes gets detached. (DetachDisk storm)
+    	9. Delete all PVCs. This should delete all Disks. (DeleteVolume storm)
+		10. Delete storage class.
+*/
+
+var _ = framework.KubeDescribe("vsphere volume operations storm [Volume]", func() {
+	f := framework.NewDefaultFramework("volume-ops-storm")
+	const DEFAULT_VOLUME_OPS_SCALE = 30
+	var (
+		client            clientset.Interface
+		namespace         string
+		storageclass      *storage.StorageClass
+		pvclaims          []*v1.PersistentVolumeClaim
+		persistentvolumes []*v1.PersistentVolume
+		err               error
+		volume_ops_scale  int
+		vsp               *vsphere.VSphere
+	)
+	BeforeEach(func() {
+		framework.SkipUnlessProviderIs("vsphere")
+		client = f.ClientSet
+		namespace = f.Namespace.Name
+		nodeList := framework.GetReadySchedulableNodesOrDie(f.ClientSet)
+		if len(nodeList.Items) == 0 {
+			framework.Failf("Unable to find ready and schedulable Node")
+		}
+		if os.Getenv("VOLUME_OPS_SCALE") != "" {
+			volume_ops_scale, err = strconv.Atoi(os.Getenv("VOLUME_OPS_SCALE"))
+			Expect(err).NotTo(HaveOccurred())
+		} else {
+			volume_ops_scale = DEFAULT_VOLUME_OPS_SCALE
+		}
+		pvclaims = make([]*v1.PersistentVolumeClaim, volume_ops_scale)
+		vsp, err = vsphere.GetVSphere()
+		Expect(err).NotTo(HaveOccurred())
+	})
+	AfterEach(func() {
+		By("Deleting PVCs")
+		for _, claim := range pvclaims {
+			framework.DeletePersistentVolumeClaim(client, claim.Name, namespace)
+		}
+		By("Deleting StorageClass")
+		err = client.StorageV1beta1().StorageClasses().Delete(storageclass.Name, nil)
+		Expect(err).NotTo(HaveOccurred())
+	})
+
+	It("should create pod with many volumes and verify no attach call fails", func() {
+		By(fmt.Sprintf("Running test with VOLUME_OPS_SCALE: %v", volume_ops_scale))
+		By("Creating Storage Class")
+		scParameters := make(map[string]string)
+		scParameters["diskformat"] = "thin"
+		storageclass, err = client.StorageV1beta1().StorageClasses().Create(getVSphereStorageClassSpec("thinsc", scParameters))
+		Expect(err).NotTo(HaveOccurred())
+
+		By("Creating PVCs using the Storage Class")
+		count := 0
+		for count < volume_ops_scale {
+			pvclaims[count] = framework.CreatePVC(client, namespace, getVSphereClaimSpecWithStorageClassAnnotation(namespace, storageclass))
+			count++
+		}
+
+		By("Waiting for all claims to be in bound phase")
+		persistentvolumes = framework.WaitForPVClaimBoundPhase(client, pvclaims)
+
+		By("Creating pod to attach PVs to the node")
+		pod := framework.CreatePod(client, namespace, pvclaims, false, "")
+
+		By("Verify all volumes are accessible and available in the pod")
+		verifyVSphereVolumesAccessible(pod, persistentvolumes, vsp)
+
+		By("Deleting pod")
+		framework.DeletePodWithWait(f, client, pod)
+
+		By("Waiting for volumes to be detached from the node")
+		for _, pv := range persistentvolumes {
+			waitForVSphereDiskToDetach(vsp, pv.Spec.VsphereVolume.VolumePath, k8stype.NodeName(pod.Spec.NodeName))
+		}
+	})
+})