Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

E2E scale test for vSphere Cloud Provider Volume lifecycle operations #54018

Merged
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
15 changes: 9 additions & 6 deletions test/e2e/framework/pv_util.go
Original file line number Diff line number Diff line change
Expand Up @@ -790,12 +790,12 @@ func deletePD(pdName string) error {
// 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')")
return MakePod(ns, nil, []*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 {
func MakePod(ns string, nodeSelector map[string]string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bool, command string) *v1.Pod {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nice addition

if len(command) == 0 {
command = "while true; do sleep 1; done"
}
Expand Down Expand Up @@ -832,6 +832,9 @@ func MakePod(ns string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bool,
}
podSpec.Spec.Containers[0].VolumeMounts = volumeMounts
podSpec.Spec.Volumes = volumes
if nodeSelector != nil {
podSpec.Spec.NodeSelector = nodeSelector
}
return podSpec
}

Expand Down Expand Up @@ -886,9 +889,9 @@ func MakeSecPod(ns string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bo
return podSpec
}

// create pod with given claims
func CreatePod(client clientset.Interface, namespace string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bool, command string) (*v1.Pod, error) {
pod := MakePod(namespace, pvclaims, isPrivileged, command)
// CreatePod with given claims based on node selector
func CreatePod(client clientset.Interface, namespace string, nodeSelector map[string]string, pvclaims []*v1.PersistentVolumeClaim, isPrivileged bool, command string) (*v1.Pod, error) {
pod := MakePod(namespace, nodeSelector, pvclaims, isPrivileged, command)
pod, err := client.CoreV1().Pods(namespace).Create(pod)
if err != nil {
return nil, fmt.Errorf("pod Create API error: %v", err)
Expand Down Expand Up @@ -928,7 +931,7 @@ func CreateSecPod(client clientset.Interface, namespace string, pvclaims []*v1.P

// 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, error) {
return CreatePod(c, ns, []*v1.PersistentVolumeClaim{pvc}, true, "")
return CreatePod(c, ns, nil, []*v1.PersistentVolumeClaim{pvc}, true, "")
}

// wait until all pvcs phase set to bound
Expand Down
4 changes: 2 additions & 2 deletions test/e2e/multicluster/ubernetes_lite.go
Original file line number Diff line number Diff line change
Expand Up @@ -309,8 +309,8 @@ func PodsUseStaticPVsOrFail(f *framework.Framework, podCount int, image string)

By("Creating pods for each static PV")
for _, config := range configs {
podConfig := framework.MakePod(ns, []*v1.PersistentVolumeClaim{config.pvc}, false, "")
config.pod, err = c.CoreV1().Pods(ns).Create(podConfig)
podConfig := framework.MakePod(ns, nil, []*v1.PersistentVolumeClaim{config.pvc}, false, "")
config.pod, err = c.Core().Pods(ns).Create(podConfig)
Expect(err).NotTo(HaveOccurred())
}

Expand Down
1 change: 1 addition & 0 deletions test/e2e/storage/BUILD
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ go_library(
"volume_metrics.go",
"volume_provisioning.go",
"volumes.go",
"vsphere_scale.go",
"vsphere_statefulsets.go",
"vsphere_stress.go",
"vsphere_utils.go",
Expand Down
4 changes: 2 additions & 2 deletions test/e2e/storage/persistent_volumes-disruptive.go
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,7 @@ var _ = SIGDescribe("PersistentVolumes[Disruptive][Flaky]", func() {
framework.ExpectNoError(framework.WaitOnPVandPVC(c, ns, pv2, pvc2))

By("Attaching both PVC's to a single pod")
clientPod, err = framework.CreatePod(c, ns, []*v1.PersistentVolumeClaim{pvc1, pvc2}, true, "")
clientPod, err = framework.CreatePod(c, ns, nil, []*v1.PersistentVolumeClaim{pvc1, pvc2}, true, "")
Expect(err).NotTo(HaveOccurred())
})

Expand Down Expand Up @@ -309,7 +309,7 @@ func initTestCase(f *framework.Framework, c clientset.Interface, pvConfig framew
}
}()
Expect(err).NotTo(HaveOccurred())
pod := framework.MakePod(ns, []*v1.PersistentVolumeClaim{pvc}, true, "")
pod := framework.MakePod(ns, nil, []*v1.PersistentVolumeClaim{pvc}, true, "")
pod.Spec.NodeName = nodeName
framework.Logf("Creating NFS client pod.")
pod, err = c.CoreV1().Pods(ns).Create(pod)
Expand Down
2 changes: 1 addition & 1 deletion test/e2e/storage/persistent_volumes.go
Original file line number Diff line number Diff line change
Expand Up @@ -292,7 +292,7 @@ var _ = SIGDescribe("PersistentVolumes", 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, []*v1.PersistentVolumeClaim{pvc}, true, fmt.Sprintf("[ $(ls -A %s | wc -l) -eq 0 ] && exit 0 || exit 1", mount))
pod = framework.MakePod(ns, nil, []*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())
framework.ExpectNoError(framework.WaitForPodSuccessInNamespace(c, pod.Name, ns))
Expand Down
4 changes: 2 additions & 2 deletions test/e2e/storage/volume_metrics.go
Original file line number Diff line number Diff line change
Expand Up @@ -87,7 +87,7 @@ var _ = SIGDescribe("[Serial] Volume metrics", func() {

claims := []*v1.PersistentVolumeClaim{pvc}

pod := framework.MakePod(ns, claims, false, "")
pod := framework.MakePod(ns, nil, claims, false, "")
pod, err = c.CoreV1().Pods(ns).Create(pod)
Expect(err).NotTo(HaveOccurred())

Expand Down Expand Up @@ -115,7 +115,7 @@ var _ = SIGDescribe("[Serial] Volume metrics", func() {
Expect(pvc).ToNot(Equal(nil))

claims := []*v1.PersistentVolumeClaim{pvc}
pod := framework.MakePod(ns, claims, false, "")
pod := framework.MakePod(ns, nil, claims, false, "")
pod, err = c.CoreV1().Pods(ns).Create(pod)
Expect(err).NotTo(HaveOccurred())

Expand Down
247 changes: 247 additions & 0 deletions test/e2e/storage/vsphere_scale.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,247 @@
/*
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 storage

import (
"fmt"
"os"
"strconv"

. "github.com/onsi/ginkgo"
. "github.com/onsi/gomega"
"k8s.io/api/core/v1"
storageV1 "k8s.io/api/storage/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
k8stypes "k8s.io/apimachinery/pkg/types"
clientset "k8s.io/client-go/kubernetes"
"k8s.io/kubernetes/pkg/cloudprovider/providers/vsphere"
"k8s.io/kubernetes/test/e2e/framework"
)

/*
Perform vsphere volume life cycle management at scale based on user configurable value for number of volumes.
The following actions will be performed as part of this test.

1. Create Storage Classes of 4 Categories (Default, SC with Non Default Datastore, SC with SPBM Policy, SC with VSAN Storage Capalibilies.)
2. Read VCP_SCALE_VOLUME_COUNT, VCP_SCALE_INSTANCES, VCP_SCALE_VOLUMES_PER_POD, VSPHERE_SPBM_POLICY_NAME, VSPHERE_DATASTORE from System Environment.
3. Launch VCP_SCALE_INSTANCES goroutine for creating VCP_SCALE_VOLUME_COUNT volumes. Each goroutine is responsible for create/attach of VCP_SCALE_VOLUME_COUNT/VCP_SCALE_INSTANCES volumes.
4. Read VCP_SCALE_VOLUMES_PER_POD from System Environment. Each pod will be have VCP_SCALE_VOLUMES_PER_POD attached to it.
5. Once all the go routines are completed, we delete all the pods and volumes.
*/
const (
NodeLabelKey = "vsphere_e2e_label"
)

// NodeSelector holds
type NodeSelector struct {
labelKey string
labelValue string
}

var _ = SIGDescribe("vcp at scale [Feature:vsphere] ", func() {
f := framework.NewDefaultFramework("vcp-at-scale")

var (
client clientset.Interface
namespace string
nodeSelectorList []*NodeSelector
volumeCount int
numberOfInstances int
volumesPerPod int
nodeVolumeMapChan chan map[string][]string
nodes *v1.NodeList
policyName string
datastoreName string
scNames = []string{storageclass1, storageclass2, storageclass3, storageclass4}
err error
)

BeforeEach(func() {
framework.SkipUnlessProviderIs("vsphere")
client = f.ClientSet
namespace = f.Namespace.Name
nodeVolumeMapChan = make(chan map[string][]string)

// Read the environment variables
volumeCountStr := os.Getenv("VCP_SCALE_VOLUME_COUNT")
Expect(volumeCountStr).NotTo(BeEmpty(), "ENV VCP_SCALE_VOLUME_COUNT is not set")
volumeCount, err = strconv.Atoi(volumeCountStr)
Expect(err).NotTo(HaveOccurred(), "Error Parsing VCP_SCALE_VOLUME_COUNT")

volumesPerPodStr := os.Getenv("VCP_SCALE_VOLUME_PER_POD")
Expect(volumesPerPodStr).NotTo(BeEmpty(), "ENV VCP_SCALE_VOLUME_PER_POD is not set")
volumesPerPod, err = strconv.Atoi(volumesPerPodStr)
Expect(err).NotTo(HaveOccurred(), "Error Parsing VCP_SCALE_VOLUME_PER_POD")

numberOfInstancesStr := os.Getenv("VCP_SCALE_INSTANCES")
Expect(numberOfInstancesStr).NotTo(BeEmpty(), "ENV VCP_SCALE_INSTANCES is not set")
numberOfInstances, err = strconv.Atoi(numberOfInstancesStr)
Expect(err).NotTo(HaveOccurred(), "Error Parsing VCP_SCALE_INSTANCES")
Expect(numberOfInstances > 5).NotTo(BeTrue(), "Maximum allowed instances are 5")
Expect(numberOfInstances > volumeCount).NotTo(BeTrue(), "Number of instances should be less than the total volume count")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"less than or equal to" based on the condition.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

volumeCount is not validated to be within a required range, thus, this test for numberOfInstances is not reliable. In general, user input should be validated to prevent test flakes. It is ok with me if validating these env var inputs is addressed across all vshpere tests in a separate pr.


policyName = os.Getenv("VSPHERE_SPBM_POLICY_NAME")
datastoreName = os.Getenv("VSPHERE_DATASTORE")
Expect(policyName).NotTo(BeEmpty(), "ENV VSPHERE_SPBM_POLICY_NAME is not set")
Expect(datastoreName).NotTo(BeEmpty(), "ENV VSPHERE_DATASTORE is not set")

nodes = framework.GetReadySchedulableNodesOrDie(client)
if len(nodes.Items) < 2 {
framework.Skipf("Requires at least %d nodes (not %d)", 2, len(nodes.Items))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

good err msg!

}
// Verify volume count specified by the user can be satisfied
if volumeCount > volumesPerNode*len(nodes.Items) {
framework.Skipf("Cannot attach %d volumes to %d nodes. Maximum volumes that can be attached on %d nodes is %d", volumeCount, len(nodes.Items), len(nodes.Items), volumesPerNode*len(nodes.Items))
}
nodeSelectorList = createNodeLabels(client, namespace, nodes)
})

/*
Remove labels from all the nodes
*/
framework.AddCleanupAction(func() {
for _, node := range nodes.Items {
framework.RemoveLabelOffNode(client, node.Name, NodeLabelKey)
}
})

It("vsphere scale tests", func() {
var pvcClaimList []string
nodeVolumeMap := make(map[k8stypes.NodeName][]string)
// Volumes will be provisioned with each different types of Storage Class
scArrays := make([]*storageV1.StorageClass, len(scNames))
for index, scname := range scNames {
// Create vSphere Storage Class
By(fmt.Sprintf("Creating Storage Class : %q", scname))
var sc *storageV1.StorageClass
scParams := make(map[string]string)
var err error
switch scname {
case storageclass1:
scParams = nil
case storageclass2:
scParams[Policy_HostFailuresToTolerate] = "1"
case storageclass3:
scParams[SpbmStoragePolicy] = policyName
case storageclass4:
scParams[Datastore] = datastoreName
}
sc, err = client.StorageV1().StorageClasses().Create(getVSphereStorageClassSpec(scname, scParams))
Expect(sc).NotTo(BeNil(), "Storage class is empty")
Expect(err).NotTo(HaveOccurred(), "Failed to create storage class")
defer client.StorageV1().StorageClasses().Delete(scname, nil)
scArrays[index] = sc
}

vsp, err := vsphere.GetVSphere()
Expect(err).NotTo(HaveOccurred())

volumeCountPerInstance := volumeCount / numberOfInstances
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

where is the sanity checking on numOfInstances? It could be larger than volumeCount. Or it could be too large for the test env to create that many go routines.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed!

for instanceCount := 0; instanceCount < numberOfInstances; instanceCount++ {
if instanceCount == numberOfInstances-1 {
volumeCountPerInstance = volumeCount
}
volumeCount = volumeCount - volumeCountPerInstance
go VolumeCreateAndAttach(client, namespace, scArrays, volumeCountPerInstance, volumesPerPod, nodeSelectorList, nodeVolumeMapChan, vsp)
}

// Get the list of all volumes attached to each node from the go routines by reading the data from the channel
for instanceCount := 0; instanceCount < numberOfInstances; instanceCount++ {
for node, volumeList := range <-nodeVolumeMapChan {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

doesn't this channel need to be explicitly closed to end the loop?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You needn't close every channel when you've finished with it. It's only necessary to close a channel when it is important to tell the receiving go routines that all data have been sent. A channel that the garbage collector determines to be unreachable will have its resources reclaimed whether or not it is closed.

However just to not end up consuming more resources, I have explicitly closed the channel.

nodeVolumeMap[k8stypes.NodeName(node)] = append(nodeVolumeMap[k8stypes.NodeName(node)], volumeList...)
}
}
podList, err := client.CoreV1().Pods(namespace).List(metav1.ListOptions{})
for _, pod := range podList.Items {
pvcClaimList = append(pvcClaimList, getClaimsForPod(&pod, volumesPerPod)...)
By("Deleting pod")
err = framework.DeletePodWithWait(f, client, &pod)
Expect(err).NotTo(HaveOccurred())
}
By("Waiting for volumes to be detached from the node")
err = waitForVSphereDisksToDetach(vsp, nodeVolumeMap)
Expect(err).NotTo(HaveOccurred())

for _, pvcClaim := range pvcClaimList {
err = framework.DeletePersistentVolumeClaim(client, pvcClaim, namespace)
Expect(err).NotTo(HaveOccurred())
}
})
})

// Get PVC claims for the pod
func getClaimsForPod(pod *v1.Pod, volumesPerPod int) []string {
pvcClaimList := make([]string, volumesPerPod)
for i, volumespec := range pod.Spec.Volumes {
if volumespec.PersistentVolumeClaim != nil {
pvcClaimList[i] = volumespec.PersistentVolumeClaim.ClaimName
}
}
return pvcClaimList
}

// VolumeCreateAndAttach peforms create and attach operations of vSphere persistent volumes at scale
func VolumeCreateAndAttach(client clientset.Interface, namespace string, sc []*storageV1.StorageClass, volumeCountPerInstance int, volumesPerPod int, nodeSelectorList []*NodeSelector, nodeVolumeMapChan chan map[string][]string, vsp *vsphere.VSphere) {
defer GinkgoRecover()
nodeVolumeMap := make(map[string][]string)
nodeSelectorIndex := 0
for index := 0; index < volumeCountPerInstance; index = index + volumesPerPod {
if (volumeCountPerInstance - index) < volumesPerPod {
volumesPerPod = volumeCountPerInstance - index
}
pvclaims := make([]*v1.PersistentVolumeClaim, volumesPerPod)
for i := 0; i < volumesPerPod; i++ {
By("Creating PVC using the Storage Class")
pvclaim, err := framework.CreatePVC(client, namespace, getVSphereClaimSpecWithStorageClassAnnotation(namespace, "2Gi", sc[index%len(sc)]))
Expect(err).NotTo(HaveOccurred())
pvclaims[i] = pvclaim
}

By("Waiting for claim to be in bound phase")
persistentvolumes, err := framework.WaitForPVClaimBoundPhase(client, pvclaims, framework.ClaimProvisionTimeout)
Expect(err).NotTo(HaveOccurred())

By("Creating pod to attach PV to the node")
nodeSelector := nodeSelectorList[nodeSelectorIndex%len(nodeSelectorList)]
// Create pod to attach Volume to Node
pod, err := framework.CreatePod(client, namespace, map[string]string{nodeSelector.labelKey: nodeSelector.labelValue}, pvclaims, false, "")
Expect(err).NotTo(HaveOccurred())

for _, pv := range persistentvolumes {
nodeVolumeMap[pod.Spec.NodeName] = append(nodeVolumeMap[pod.Spec.NodeName], pv.Spec.VsphereVolume.VolumePath)
}
By("Verify the volume is accessible and available in the pod")
verifyVSphereVolumesAccessible(pod, persistentvolumes, vsp)
nodeSelectorIndex++
}
nodeVolumeMapChan <- nodeVolumeMap
close(nodeVolumeMapChan)
}

func createNodeLabels(client clientset.Interface, namespace string, nodes *v1.NodeList) []*NodeSelector {
var nodeSelectorList []*NodeSelector
for i, node := range nodes.Items {
labelVal := "vsphere_e2e_" + strconv.Itoa(i)
nodeSelector := &NodeSelector{
labelKey: NodeLabelKey,
labelValue: labelVal,
}
nodeSelectorList = append(nodeSelectorList, nodeSelector)
framework.AddOrUpdateLabelOnNode(client, node.Name, NodeLabelKey, labelVal)
}
return nodeSelectorList
}