Snapshotting a Standalone PVC

This guide will show you how to use Stash to snapshot standalone PersistentVolumeClaims and restore it from the snapshot using Kubernetes VolumeSnapshot API. In this guide, we are going to backup the volumes in Google Cloud Platform with the help of GCE Persistent Disk CSI Driver.

Before You Begin

• At first, you need to be familiar with the GCE Persistent Disk CSI Driver.
• Install Stash in your cluster following the steps here.
• If you don’t know how VolumeSnapshot works in Stash, please visit here.

Prepare for VolumeSnapshot

Here, we are going to create StorageClass that uses GCE Persistent Disk CSI Driver.

Below is the YAML of the StorageClass we are going to use,

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: csi-standard
parameters:
  type: pd-standard
provisioner: pd.csi.storage.gke.io
reclaimPolicy: Delete
volumeBindingMode: Immediate

Let’s create the StorageClass we have shown above,

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/storageclass.yaml
storageclass.storage.k8s.io/csi-standard created

We also need a VolumeSnapshotClass. Below is the YAML of the VolumeSnapshotClass we are going to use,

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotClass
metadata:
  name: csi-snapshot-class
driver: pd.csi.storage.gke.io
deletionPolicy: Delete

Here,

• driver field to point to the respective CSI driver that is responsible for taking snapshot. As we are using GCE Persistent Disk CSI Driver, we are going to use pd.csi.storage.gke.io in this field.

Let’s create the volumeSnapshotClass we have shown above,

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/volumesnapshotclass.yaml
volumesnapshotclass.snapshot.storage.k8s.io/csi-snapshot-class created

To keep everything isolated, we are going to use a separate namespace called demo throughout this tutorial.

$ kubectl create ns demo
namespace/demo created

Note: YAML files used in this tutorial are stored in /docs/guides/volumesnapshot/pvc/examples directory of stashed/docs repository.

Take Volume Snapshot

Here, we are going to create a PVC and mount it with a pod and we are going to also generate some sample data on it. Then, we are going to take snapshot of this PVC using Stash.

Create PersistentVolumeClaim :

At first, let’s create a PVC. We are going to mount this PVC in a pod.

Below is the YAML of the sample PVC,

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: source-data
  namespace: demo
spec:
  accessModes:
  - ReadWriteOnce
  storageClassName: standard
  resources:
    requests:
      storage: 1Gi

Let’s create the PVC we have shown above.

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/source-pvc.yaml
persistentvolumeclaim/source-data created

Create Pod :

Now, we are going to deploy a pod that uses the above PVC. This pod will automatically create data.txt file in /source/data directory and write some sample data in it and also mounted the desired PVC in /source/data directory.

Below is the YAML of the pod that we are going to create,

apiVersion: v1
kind: Pod
metadata:
  name: source-pod
  namespace: demo
spec:
  containers:
  - name: busybox
    image: busybox
    command: ["/bin/sh", "-c"]
    args: ["echo sample_data > /source/data/data.txt && sleep 3000"]
    volumeMounts:
    - name: source-data
      mountPath: /source/data
  volumes:
  - name: source-data
    persistentVolumeClaim:
      claimName: source-data
      readOnly: false

Let’s create the Pod we have shown above.

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/source-pod.yaml
pod/source-pod created

Now, wait for the Pod to go into the Running state.

$ kubectl get pod -n demo
NAME         READY   STATUS    RESTARTS   AGE
source-pod   1/1     Running   0          25s

Verify that the sample data has been created in /source/data directory for source-pod pod using the following command,

$ kubectl exec -n demo source-pod -- cat /source/data/data.txt
sample_data

Create BackupConfiguration :

Now, create a BackupConfiguration crd to take snapshot of the source-data PVC.

Below is the YAML of the BackupConfiguration crd that we are going to create,

apiVersion: stash.appscode.com/v1beta1
kind: BackupConfiguration
metadata:
  name: pvc-volume-snapshot
  namespace: demo
spec:
  schedule: "*/5 * * * *"
  driver: VolumeSnapshotter
  target:
    ref:
      apiVersion: v1
      kind: PersistentVolumeClaim
      name: source-data
    snapshotClassName: csi-snapshot-class
  retentionPolicy:
    name: 'keep-last-5'
    keepLast: 5
    prune: true

Here,

• spec.schedule is a cron expression indicates that BackupSession will be created at 5 minute interval.

• spec.driver indicates the name of the agent to use to back up the target. Currently, Stash supports Restic, VolumeSnapshotter drivers. The VolumeSnapshotter is used to backup/restore PVC using VolumeSnapshot API.

• spec.target.ref refers to the backup target. apiVersion, kind and name refers to the apiVersion, kind and name of the targeted workload respectively. Stash will use this information to create a Volume Snapshotter Job for creating VolumeSnapshot.

• spec.target.snapshotClassName indicates the VolumeSnapshotClass to be used for volume snapshotting.

Let’s create the BackupConfiguration crd we have shown above.

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/backupconfiguration.yaml
backupconfiguration.stash.appscode.com/pvc-volume-snapshot created

Verify CronJob :

If everything goes well, Stash will create a CronJob to take periodic snapshot of the PVC with the schedule specified in spec.schedule field of BackupConfiguration crd.

Check that the CronJob has been created using the following command,

$ kubectl get cronjob -n demo 
NAME                          SCHEDULE      SUSPEND   ACTIVE   LAST SCHEDULE   AGE
pvc-volume-snapshot           */1 * * * *   False     0        39s             2m41s

Wait for BackupSession :

The pvc-volume-snapshot CronJob will trigger a backup on each scheduled time slot by creating a BackupSession crd.

Wait for the next schedule of backup. Run the following command to watch BackupSession crd,

$ watch -n 1 kubectl get backupsession -n demo
Every 1.0s: kubectl get backupsession -n demo                      

NAME                             INVOKER-TYPE          INVOKER-NAME          PHASE       AGE
pvc-volume-snapshot-fnbwz        BackupConfiguration   pvc-volume-snapshot   Succeeded   1m32s

We can see above that the backup session has succeeded. Now, we are going to verify that the VolumeSnapshot has been created and the snapshots have been stored in the respective backend.

Verify Volume Snapshot :

Once a BackupSession crd is created, it creates volume snapshotter Job. Then the Job creates VolumeSnapshot crd for the targeted PVC.

Check that the VolumeSnapshot has been created Successfully.

$ kubectl get volumesnapshot -n demo
NAME                     AGE
source-data-fnbwz        1m30s

Let’s find out the actual snapshot name that will be saved in the Google Cloud by the following command,

kubectl get volumesnapshot source-data-fnbwz  -n demo -o yaml

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshot
metadata:
  creationTimestamp: "2019-07-15T10:31:09Z"
  finalizers:
  - snapshot.storage.kubernetes.io/volumesnapshot-protection
  generation: 4
  name: source-data-fnbwz
  namespace: demo
  resourceVersion: "32098"
  selfLink: /apis/snapshot.storage.k8s.io/v1/namespaces/demo/volumesnapshots/source-data-fnbwz
  uid: a8e8faeb-a6eb-11e9-9f3a-42010a800050
spec:
  source:
    persistentVolumeClaimName: source-data
  volumeSnapshotClassName: csi-snapshot-class
status:
  boundVolumeSnapshotContentName: snapcontent-a8e8faeb-a6eb-11e9-9f3a-42010a800050
  creationTime: "2019-07-15T10:31:10Z"
  readyToUse: true
  restoreSize: 1Gi

Here, spec.snapshotContentName field specifies the name of the VolumeSnapshotContent crd. It also represents the actual snapshot name that has been saved in Google Cloud. If we navigate to the Snapshots tab in the GCP console, we are going to see snapshot snapcontent-a8e8faeb-a6eb-11e9-9f3a-42010a800050 has been stored successfully.

  

Restore PVC from VolumeSnapshot

This section will show you how to restore the PVC from the snapshot we have taken in the earlier section.

Stop Taking Backup of the Old PVC:

At first, let’s stop taking any further backup of the old PVC so that no backup is taken during the restore process. We are going to pause the BackupConfiguration that we created to backup the source-data PVC. Then, Stash will stop taking any further backup for this PVC. You can learn more how to pause a scheduled backup here

Let’s pause the pvc-volume-snapshot BackupConfiguration,

$ kubectl patch backupconfiguration -n demo pvc-volume-snapshot --type="merge" --patch='{"spec": {"paused": true}}'
backupconfiguration.stash.appscode.com/pvc-volume-snapshot patched

Now, wait for a moment. Stash will pause the BackupConfiguration. Verify that the BackupConfiguration has been paused,

$ kubectl get backupconfiguration -n demo
NAME                  TASK   SCHEDULE      PAUSED   AGE
pvc-volume-snapshot          */1 * * * *   true     22m

Notice the PAUSED column. Value true for this field means that the BackupConfiguration has been paused.

Create RestoreSession :

At first, we have to create a RestoreSession crd to restore the PVC from respective snapshot.

Below is the YAML of the RestoreSesion crd that we are going to create,

apiVersion: stash.appscode.com/v1beta1
kind: RestoreSession
metadata:
  name: restore-pvc
  namespace: demo
spec:
  driver: VolumeSnapshotter
  target:
    volumeClaimTemplates:
    - metadata:
        name: restore-data
      spec:
        accessModes: [ "ReadWriteOnce" ]
        storageClassName: "csi-standard"
        resources:
          requests:
            storage: 1Gi
        dataSource:
          kind: VolumeSnapshot
          name: source-data-fnbwz
          apiGroup: snapshot.storage.k8s.io

Here,

• spec.target.volumeClaimTemplates:
  • metadata.name is the name of the restored PVC or prefix of the VolumeSnapshot name.
  • spec.dataSource: spec.dataSource specifies the source of the data from where the newly created PVC will be initialized. It requires following fields to be set:
    • apiGroup is the group for resource being referenced. Now, Kubernetes supports only snapshot.storage.k8s.io.
    • kind is resource of the kind being referenced. Now, Kubernetes supports only VolumeSnapshot.
    • name is the VolumeSnapshot resource name. In RestoreSession crd, You must set the VolumeSnapshot name directly.

Let’s create the RestoreSession crd we have shown above.

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/restoresession.yaml
restoresession.stash.appscode.com/restore-pvc created

Once, you have created the RestoreSession crd, Stash will create a job to restore. We can watch the RestoreSession phase to check if the restore process has succeeded or not.

Run the following command to watch RestoreSession phase,

$ watch -n 1 kubectl get restore -n demo
Every 1.0s: kubectl get restore -n demo                     

NAME          REPOSITORY-NAME   PHASE       AGE
restore-pvc                     Running     10s
restore-pvc                     Succeeded   1m

So, we can see from the output of the above command that the restore process succeeded.

Verify Restored PVC :

Once the restore process is complete, we are going to see that new PVC with the name restore-data has been created.

To verify that the PVC has been created, run by the following command,

$ kubectl get pvc -n demo
NAME           STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
restore-data   Bound    pvc-c5f0e7f5-a6ec-11e9-9f3a-42010a800050   1Gi        RWO            standard       52s

Notice the STATUS field. It indicates that the respective PV has been provisioned and initialized from the respective VolumeSnapshot by CSI driver and the PVC has been bound with the PV.

The volumeBindingMode field controls when volume binding and dynamic provisioning should occur. Kubernetes allows Immediate and WaitForFirstConsumer modes for binding volumes. The Immediate mode indicates that volume binding and dynamic provisioning occurs once the PVC is created and WaitForFirstConsumer mode indicates that volume binding and provisioning does not occur until a pod is created that uses this PVC. By default volumeBindingMode is Immediate.

If you use volumeBindingMode: WaitForFirstConsumer, respective PVC will be initialized from respective VolumeSnapshot after you create a workload with that PVC. In this case, Stash will mark the restore session as completed with phase Unknown.

Verify Restored Data :

We are going to create a new pod with the restored PVC to verify whether the backed up data has been restored.

Below, the YAML for the Pod we are going to create.

apiVersion: v1
kind: Pod
metadata:
  name: restored-pod
  namespace: demo
spec:
  containers:
  - name: busybox
    image: busybox
    args:
    - sleep
    - "3600"
    volumeMounts:
    - name: restore-data
      mountPath: /restore/data
  volumes:
  - name: restore-data
    persistentVolumeClaim:
      claimName: restore-data
      readOnly: false

Let’s create the Pod we have shown above.

$ kubectl apply -f https://github.com/stashed/docs/raw/v2024.4.8/docs/guides/volumesnapshot/pvc/examples/restored-pod.yaml
pod/restored-pod created

Now, wait for the Pod to go into the Running state.

$ kubectl get pod -n demo 
NAME           READY   STATUS    RESTARTS   AGE
restored-pod   1/1     Running   0          34s

Verify that the backed up data has been restored in /restore/data directory for restored-pod pod using the following command,

$ kubectl exec -n demo restored-pod -- cat /restore/data/data.txt
sample_data

Cleaning Up

To clean up the Kubernetes resources created by this tutorial, run:

kubectl delete -n demo pod source-pod
kubectl delete -n demo pod restored-pod
kubectl delete -n demo backupconfiguration pvc-volume-snapshot
kubectl delete -n demo restoresession restore-pvc
kubectl delete -n demo storageclass csi-standard
kubectl delete -n demo volumesnapshotclass csi-snapshot-class
kubectl delete -n demo pvc --all