You are looking at the documentation of a prior release. To read the documentation of the latest release, please visit here.

Snapshotting the volumes of a StatefulSet

This guide will show you how to use Stash to snapshot the volumes of a StatefulSets and restore them from 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.
  • You need to enable the Kubernetes VolumeSnapshotDataSource alpha feature via Kubernetes feature gates
    • --feature-gates=VolumeSnapshotDataSource=true
  • 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

If you don’t already have a StorageClass that uses the CSI driver that supports VolumeSnapshot feature, create one first. Here, we are going to create StorageClass that uses GCE Persistent Disk CSI Driver.

Sample StorageClass YAML are given below,

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: 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/v2020.11.17/docs/examples/guides/latest/volumesnapshot/storageclass.yaml
storageclass.storage.k8s.io/standard created

We also need a VolumeSnapshotClass. We are going to use the following VolumeSnapshotClass for this tutorial,

apiVersion: snapshot.storage.k8s.io/v1alpha1
kind: VolumeSnapshotClass
metadata:
  annotations:
    snapshot.storage.kubernetes.io/is-default-class: "true"
  name: default-snapshot-class
snapshotter: pd.csi.storage.gke.io

Here,

  • metadata.annotations annotations are used to set default volumeSnapshotClass.
  • snapshotter 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 set pd.csi.storage.gke.io to this field.

Let’s create the volumeSnapshotClass crd we have shown above,

$ kubectl apply -f https://github.com/stashed/docs/raw/v2020.11.17/docs/examples/guides/latest/volumesnapshot/default-volumesnapshotclass.yaml
volumesnapshotclass.snapshot.storage.k8s.io/default-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/examples/guides/latest/volumesnapshot directory of stashed/docs repository.

Take Volume Snapshot

When you create a Statefulset, there is no need to create PVCs separately, because all replicas in Statefulset use different PVCs to store data. Kubernetes allows us to define a volumeClaimTemplates in Statefulset so that new PVC is created for each replica automatically. We are going to take snapshot of those PVCs using Stash.

Deploy StatefulSet :

Now, we are going to deploy a Statefulset. This Statefulset will automatically generate sample data in /source/data directory.

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

apiVersion: v1
kind: Service
metadata:
  name: svc
  labels:
    app: demo
  namespace: demo
spec:
  ports:
  - port: 80
    name: web
  clusterIP: None
  selector:
    app: stash
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: stash-demo
  namespace: demo
spec:
  selector:
    matchLabels:
      app: stash
  serviceName: svc
  replicas: 3
  template:
    metadata:
      labels:
        app: stash
    spec:
      containers:
      - args: ["echo $(POD_NAME) > /source/data/data.txt && sleep 3000"]
        command: ["/bin/sh", "-c"]
        env:
        - name:  POD_NAME
          valueFrom:
            fieldRef:
              fieldPath:  metadata.name
        name: nginx
        image: nginx
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: source-data
          mountPath: /source/data
  volumeClaimTemplates:
  - metadata:
      name: source-data
      namespace: demo
    spec:
      accessModes:
      - ReadWriteOnce
      storageClassName: standard
      resources:
        requests:
          storage: 1Gi

Let’s create the Statefulset we have shown above.

$ kubectl create -f ./docs/examples/guides/latest/volumesnapshot/statefulset/statefulset.yaml
service/svc created
statefulset.apps/stash-demo created

Now, wait for the pod of Statefulset to go into the Running state.

$ kubectl get pod -n demo
NAME           READY   STATUS    RESTARTS   AGE
stash-demo-0   1/1     Running   0          97s
stash-demo-1   1/1     Running   0          67s
stash-demo-2   1/1     Running   0          39s

Let’s find out the PVCs created for these replicas,

kubectl get pvc -n demo
NAME                       STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
source-data-stash-demo-0   Bound    pvc-760c1734-a6cc-11e9-9f3a-42010a800050   1Gi        RWO            standard       70s
source-data-stash-demo-1   Bound    pvc-86f5b3bd-a6cc-11e9-9f3a-42010a800050   1Gi        RWO            standard       42s
source-data-stash-demo-2   Bound    pvc-9c9f542f-a6cc-11e9-9f3a-42010a800050   1Gi        RWO            standard       5s

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

$ kubectl exec -n demo stash-demo-0 -- cat /source/data/data.txt
stash-demo-0
$ kubectl exec -n demo stash-demo-1 -- cat /source/data/data.txt
stash-demo-1
$ kubectl exec -n demo stash-demo-2 -- cat /source/data/data.txt
stash-demo-2

Create BackupConfiguration :

Now, create a BackupConfiguration crd to take snapshot of the PVCs of stash-demo Statefulset.

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

apiVersion: stash.appscode.com/v1beta1
kind: BackupConfiguration
metadata:
  name: statefulset-volume-snapshot
  namespace: demo
spec:
  schedule: "*/5 * * * *"
  driver: VolumeSnapshotter
  target:
    ref:
      apiVersion: apps/v1
      kind: StatefulSet
      name: stash-demo
    replicas : 1
    snapshotClassName: default-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/v2020.11.17/docs/examples/guides/latest/volumesnapshot/statefulset/backupconfiguration.yaml
backupconfiguration.stash.appscode.com/statefulset-volume-snapshot created

Verify CronJob :

If everything goes well, Stash will create a CronJob to take periodic snapshot of stash-demo-0 , stash-demo-1 and stash-demo-2 volumes of the Statefulset 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
statefulset-volume-snapshot   */1 * * * *   False     0        <none>          18s

Wait for BackupSession :

The statefulset-volume-snapshot CronJob will trigger a backup on each schedule by creating a BackupSession crd.

Wait for the next schedule for 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                      suaas-appscode: Tue Jun 18 18:35:41 2019

NAME                                     INVOKER-TYPE          INVOKER-NAME                  PHASE       AGE
statefulset-volume-snapshot-1563177551   BackupConfiguration   statefulset-volume-snapshot   Succeeded   57s

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 has 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.The VolumeSnapshot name follows the following pattern:

 <PVC claim name>-<statefulset name>-<pod ordinal>-<backup session creation timestamp in Unix epoch seconds>

Check that the VolumeSnapshot has been created Successfully.

$ kubectl get volumesnapshot -n demo
NAME                                  AGE
source-data-stash-demo-0-1563177551   115s
source-data-stash-demo-1-1563177551   115s
source-data-stash-demo-2-1563177551   115s

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-stash-demo-0-1563177551 -n demo -o yaml
apiVersion: snapshot.storage.k8s.io/v1alpha1
kind: VolumeSnapshot
metadata:
  creationTimestamp: "2019-07-15T07:59:13Z"
  finalizers:
  - snapshot.storage.kubernetes.io/volumesnapshot-protection
  generation: 4
  name: source-data-stash-demo-0-1563177551
  namespace: demo
  resourceVersion: "18764"
  selfLink: /apis/snapshot.storage.k8s.io/v1alpha1/namespaces/demo/volumesnapshots/source-data-stash-demo-0-1563177551
  uid: 6f3b49a9-a6d6-11e9-9f3a-42010a800050
spec:
  snapshotClassName: default-snapshot-class
  snapshotContentName: snapcontent-6f3b49a9-a6d6-11e9-9f3a-42010a800050
  source:
    apiGroup: null
    kind: PersistentVolumeClaim
    name: source-data-stash-demo-0
status:
  creationTime: "2019-07-15T07:59:14Z"
  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 should see snapshot snapcontent-6f3b49a9-a6d6-11e9-9f3a-42010a800050 has been stored successfully.

  Stash Backup Flow
Fig: Snapshots in GCE Bucket

Restore PVC from VolumeSnapshot

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

Stop Taking Backup of the Old StatefulSet:

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

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

$ kubectl patch backupconfiguration -n demo statefulset-volume-snapshot --type="merge" --patch='{"spec": {"paused": true}}'
backupconfiguration.stash.appscode.com/statefulset-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
statefulset-volume-snapshot          */1 * * * *   true     20m

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 PVCs from respective the snapshots.

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:
    replicas : 3
    volumeClaimTemplates:
    - metadata:
        name: restore-data-restore-demo-${POD_ORDINAL}
      spec:
        accessModes: [ "ReadWriteOnce" ]
        storageClassName: "standard"
        resources:
          requests:
            storage: 1Gi
        dataSource:
          kind: VolumeSnapshot
#            name: source-data-stash-demo-${POD_ORDINAL}-1563177551
          name: source-data-stash-demo-0-1563181264

Here,

  • spec.target.replicas: spec.target.replicas specify the number of replicas of a StatefulSet whose volumes were backed up and Stash uses this field to dynamically create the desired number of PVCs and initialize them from respective or Specific VolumeSnapShots.
  • spec.target.volumeClaimTemplates:
    • metadata.name is a template for the name of the restored PVC that will be created by Stash. You have to provide this named template to match with the desired PVC of a StatefulSet. For example, if you want to deploy a StatefulSet named stash-demo with volumeClaimTemplate name my-volume, the PVCs of your StatefulSet will be my-volume-stash-demo-0, my-volume-stash-demo-1 and so on. In this case, you have to provide volumeClaimTemplate name in RestoreSession in the following format:

      <volume claim name>-<statefulset name>-${POD_ORDINAL}
      

      So for the above example, volumeClaimTemplate name for RestoreSession will be my-volume-stash-demo-${POD_ORDINAL}.

    • 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 provide the name in the following format:

        <VolumeSnapshot name prefix>-${POD_ORDINAL}-<timestamp in Unix  epoch seconds>
        

        The ${POD_ORDINAL} variable is resolved by Stash. If you don’t provide this variable and specify ordinal manually, all the PVC will be restored from the same VolumeSnapshot.

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

$ kubectl create -f ./docs/examples/guides/latest/volumesnapshot/statefulset/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                      suaas-appscode: Tue Jun 18 18:35:41 2019

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 PVCs with the name restore-data-restore-demo-0 , restore-data-restore-demo-1 and restore-data-restore-demo-2 has been created.

Verify that the PVCs has been created by the following command,

$ kubectl get pvc -n demo
NAME                          STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
restore-data-restore-demo-0   Bound    pvc-ed35c54d-a6dc-11e9-9f3a-42010a800050   1Gi        RWO            standard       13s
restore-data-restore-demo-1   Bound    pvc-ed3bcb82-a6dc-11e9-9f3a-42010a800050   1Gi        RWO            standard       13s
restore-data-restore-demo-2   Bound    pvc-ed3fed79-a6dc-11e9-9f3a-42010a800050   1Gi        RWO            standard       13s

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 Statefulset with the restored PVCs to verify whether the backed up data has been restored.

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

apiVersion: v1
kind: Service
metadata:
  name: restore-svc
  labels:
    app: restore-demo
  namespace: demo
spec:
  ports:
  - port: 80
    name: web
  clusterIP: None
  selector:
    app: restore-demo
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: restore-demo
  namespace: demo
spec:
  selector:
    matchLabels:
      app: restore-demo
  serviceName: svc
  replicas: 3
  template:
    metadata:
      labels:
        app: restore-demo
    spec:
      containers:
      - args:
        - sleep
        - "3600"
        name: nginx
        image: nginx
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: restore-data
          mountPath: /restore/data
  volumeClaimTemplates:
    - metadata:
        name: restore-data
        namespace: demo
      spec:
        accessModes:
        - ReadWriteOnce
        storageClassName: standard
        resources:
          requests:
            storage: 1Gi

Let’s create the Statefulset we have shown above.

$ kubectl create -f ./docs/examples/guides/latest/volumesnapshot/statefulset/restored-statefulset.yaml
service/svc created
statefulset.apps/restore-demo created

Now, wait for the pod of the Statefulset to go into the Running state.

$ kubectl get pod -n demo
NAME             READY   STATUS    RESTARTS   AGE
restore-demo-0   1/1     Running   0          65s
restore-demo-1   1/1     Running   0          46s
restore-demo-2   1/1     Running   0          26s

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

$ kubectl exec -n demo restore-demo-0 -- cat /restore/data/data.txt
stash-demo-0
$ kubectl exec -n demo restore-demo-1 -- cat /restore/data/data.txt
stash-demo-1
$ kubectl exec -n demo restore-demo-2 -- cat /restore/data/data.txt
stash-demo-2

Advance Use-Case

Stash can also backup only single replica or restore same data on all replicas of a StatefulSet. This is particularly useful when all replicas of the StatefulSet contains same data. For example, in MongoDB ReplicaSet all the pod contains same data. In this case backup only single replica is enough. Similarly, it might be useful in some cases where all the replicas need to be initialized with same data.

Backup only Single Replica

This section will show you how to snapshot only a single replica of a Statefulset volume.

Create BackupConfiguration :

Now, create a BackupConfiguration crd to take snapshot of a single PVC of the stash-demo Statefulset.

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

apiVersion: stash.appscode.com/v1beta1
kind: BackupConfiguration
metadata:
  name: statefulset-volume-snapshot
  namespace: demo
spec:
  schedule: "*/5 * * * *"
  driver: VolumeSnapshotter
  target:
    ref:
      apiVersion: apps/v1
      kind: StatefulSet
      name: stash-demo
      replicas : 1
    snapshotClassName: default-snapshot-class
  retentionPolicy:
    name: 'keep-last-5'
    keepLast: 5
    prune: true

Here,

  • spec.replicas specifies the number of replicas (starting from 0th) whose data should be backed up. If it is set to 1, Stash will take snapshot only the volumes of <claim-name>-<statefulset-name>-0 pod. For replica set to 2, Stash will take snapshot only the volumes of <claim-name>-<statefulset-name>-0 and <claim-name>-<statefulset-name>-1 pods and so on.

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

$ kubectl apply -f https://github.com/stashed/docs/raw/v2020.11.17/docs/examples/guides/latest/volumesnapshot/statefulset/backupconfiguration.yaml
backupconfiguration.stash.appscode.com/statefulset-volume-snapshot created

Verify CronJob :

If everything goes well, Stash will create a CronJob to take periodic snapshot of stash-demo-0 volume of the Statefulset 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
statefulset-volume-snapshot   */1 * * * *   False     0        <none>          18s

Wait for BackupSession :

The statefulset-volume-snapshot CronJob will trigger a backup on each schedule by creating a BackupSession crd.

Wait for the next schedule for 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                      suaas-appscode: Tue Jun 18 18:35:41 2019

NAME                                     INVOKER-TYPE          INVOKER-NAME                  PHASE       AGE
statefulset-volume-snapshot-1563181264   BackupConfiguration   statefulset-volume-snapshot   Succeeded   57s

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

Verify Volume Snapshotting and Backup :

Once a BackupSession crd is created, Stash creates a volume snapshotter Job. Then the Job creates VolumeSnapshot crd for the targeted PVC. The VolumeSnapshot name follows the following pattern:

 <PVC claim name>-<backup session creation timestamp in Unix epoch seconds>

Check that the VolumeSnapshot has been created Successfully.

$ kubectl get volumesnapshot -n demo
NAME                                  AGE
source-data-stash-demo-0-1563181264   67s

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

kubectl get volumesnapshot source-data-stash-demo-0-1563181264 -n demo -o yaml
apiVersion: snapshot.storage.k8s.io/v1alpha1
kind: VolumeSnapshot
metadata:
  creationTimestamp: "2019-07-15T09:01:06Z"
  finalizers:
  - snapshot.storage.kubernetes.io/volumesnapshot-protection
  generation: 4
  name: source-data-stash-demo-0-1563181264
  namespace: demo
  resourceVersion: "24310"
  selfLink: /apis/snapshot.storage.k8s.io/v1alpha1/namespaces/demo/volumesnapshots/source-data-stash-demo-0-1563181264
  uid: 14984cd3-a6df-11e9-9f3a-42010a800050
spec:
  snapshotClassName: default-snapshot-class
  snapshotContentName: snapcontent-14984cd3-a6df-11e9-9f3a-42010a800050
  source:
    apiGroup: null
    kind: PersistentVolumeClaim
    name: source-data-stash-demo-0
status:
  creationTime: "2019-07-15T09:01:07Z"
  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 GCP. If we navigate to the Snapshots tab in the GCP console, we should see the snapshot snapcontent-14984cd3-a6df-11e9-9f3a-42010a800050 has been stored successfully.

  Stash Backup Flow
Fig: Snapshot in GCE Bucket

Restore Same Data in all Replicas

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

Stop Taking Backup of the Old StatefulSet:

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

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

$ kubectl patch backupconfiguration -n demo statefulset-volume-snapshot --type="merge" --patch='{"spec": {"paused": true}}'
backupconfiguration.stash.appscode.com/statefulset-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
statefulset-volume-snapshot          */1 * * * *   true     20m

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 PVCs from the 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:
    replicas : 3
    volumeClaimTemplates:
      - metadata:
          name: restore-data-restore-demo-${POD_ORDINAL}
        spec:
          accessModes: [ "ReadWriteOnce" ]
          storageClassName: "standard"
          resources:
            requests:
              storage: 1Gi
          dataSource:
            kind: VolumeSnapshot
            name: source-data-stash-demo-0-1563181264
#            name: source-data-stash-demo-${POD_ORDINAL}-1563177551
            apiGroup: snapshot.storage.k8s.io

Here,

  • spec.dataSource.name: spec.dataSource.name is the VolumeSnapshot resource name. data will be restored in all replica from single VolumeSnapshot.

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

$ kubectl create -f ./docs/examples/guides/latest/volumesnapshot/statefulset/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                      suaas-appscode: Tue Jun 18 18:35:41 2019

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 PVCs with the name restore-data-restore-demo-0 , restore-data-restore-demo-1 and restore-data-restore-demo-2 have been created successfully.

check that the status of the PVCs are bound,

$ kubectl get pvc -n demo
NAME                          STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
restore-data-restore-demo-0   Bound    pvc-745e0f51-a6e0-11e9-9f3a-42010a800050   1Gi        RWO            standard       5m23s
restore-data-restore-demo-1   Bound    pvc-746227e7-a6e0-11e9-9f3a-42010a800050   1Gi        RWO            standard       5m23s
restore-data-restore-demo-2   Bound    pvc-74674656-a6e0-11e9-9f3a-42010a800050   1Gi        RWO            standard       5m23s

Verify Restored Data :

We are going to create a new Statefulset to verify whether the restored data has been restored successfully.

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

apiVersion: v1
kind: Service
metadata:
  name: restore-svc
  labels:
    app: restore-demo
  namespace: demo
spec:
  ports:
    - port: 80
      name: web
  clusterIP: None
  selector:
    app: restore-demo
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: restore-demo
  namespace: demo
spec:
  selector:
    matchLabels:
      app: restore-demo
  serviceName: svc
  replicas: 3
  template:
    metadata:
      labels:
        app: restore-demo
    spec:
      containers:
        - args:
            - sleep
            - "3600"
          name: nginx
          image: nginx
          ports:
            - containerPort: 80
              name: web
          volumeMounts:
            - name: restore-data
              mountPath: /restore/data
  volumeClaimTemplates:
    - metadata:
        name: restore-data
        namespace: demo
      spec:
        accessModes:
          - ReadWriteOnce
        storageClassName: standard
        resources:
          requests:
            storage: 1Gi

Let’s create the Statefulset we have shown above.

$ kubectl create -f ./docs/examples/guides/latest/volumesnapshot/statefulset/restored-statefulset.yaml
service/svc created
statefulset.apps/restore-demo created

Now, wait for the pod of Statefulset to go into the Running state.

$ kubectl get pod -n demo
NAME             READY   STATUS    RESTARTS   AGE
restore-demo-0   1/1     Running   0          3m9s
restore-demo-1   1/1     Running   0          2m50s
restore-demo-2   1/1     Running   0          2m30s

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

$ kubectl exec -n demo restore-demo-0 -- cat /restore/data/data.txt
stash-demo-0
$ kubectl exec -n demo restore-demo-1 -- cat /restore/data/data.txt
stash-demo-0
$ kubectl exec -n demo restore-demo-2 -- cat /restore/data/data.txt
stash-demo-0

Cleaning Up

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

kubectl delete -n demo statefulset stash-demo
kubectl delete -n demo statefulset restore-demo
kubectl delete -n demo backupconfiguration statefulset-volume-snapshot
kubectl delete -n demo restoresession restore-pvc
kubectl delete -n demo storageclass standard
kubectl delete -n demo volumesnapshotclass default-snapshot-class