Back up Persistent Disk storage using volume snapshots


This page shows you how to back up and restore Persistent Disk storage using volume snapshots.

For an introduction, see About Kubernetes volume snapshots.

Requirements

To use volume snapshots on GKE, you must meet the following requirements:

  • Use a CSI driver that supports snapshots. The in-tree Persistent Disk driver does not support snapshots. To create and manage snapshots, you must use the same CSI driver as the underlying PersistentVolumeClaim (PVC).

  • Use control plane versions 1.17 or later. To use the Compute Engine Persistent Disk CSI driver in a VolumeSnapshot, use GKE versions 1.17.6-gke.4 or later.

  • Have an existing PersistentVolumeClaim to use for a snapshot. The PersistentVolume you use for a snapshot source must be managed by a CSI driver. You can verify that you're using a CSI driver by checking that the PersistentVolume spec has a csi section with driver: pd.csi.storage.gke.io or filestore.csi.storage.gke.io. If the PersistentVolume is dynamically provisioned by the CSI driver as described in the following sections, it's managed by the CSI driver.

Limitations

All restrictions for creating a disk snapshot on Compute Engine also apply to GKE.

Best practices

Be sure to follow best practices for Compute Engine disk snapshots when using Persistent Disk Volume snapshots on GKE.

Before you begin

Before you start, make sure you have performed the following tasks:

  • Enable the Google Kubernetes Engine API.
  • Enable Google Kubernetes Engine API
  • If you want to use the Google Cloud CLI for this task, install and then initialize the gcloud CLI. If you previously installed the gcloud CLI, get the latest version by running gcloud components update.

Creating and using a volume snapshot

The examples in this document show you how to do the following tasks:

  1. Create a PersistentVolumeClaim and Deployment.
  2. Add a file to the PersistentVolume that the Deployment uses.
  3. Create a VolumeSnapshotClass to configure the snapshot.
  4. Create a volume snapshot of the PersistentVolume.
  5. Delete the test file.
  6. Restore the PersistentVolume to the snapshot you created.
  7. Verify that the restoration worked.

To use a volume snapshot, you must complete the following steps:

  1. Create a VolumeSnapshotClass object to specify the CSI driver and deletion policy for your snapshot.
  2. Create a VolumeSnapshot object to request a snapshot of an existing PersistentVolumeClaim.
  3. Reference the VolumeSnapshot in a PersistentVolumeClaim to restore a volume to that snapshot or create a new volume using the snapshot.

Create a PersistentVolumeClaim and a Deployment

  1. To create the PersistentVolumeClaim object, save the following manifest as my-pvc.yaml:

    Persistent Disk

     apiVersion: v1
     kind: PersistentVolumeClaim
     metadata:
       name: my-pvc
     spec:
       storageClassName: standard-rwo
       accessModes:
       - ReadWriteOnce
       resources:
         requests:
           storage: 1Gi
    

    This example uses the standard-rwo storage class installed by default with the Compute Engine Persistent Disk CSI driver. To learn more, see Using the Compute Engine Persistent Disk CSI driver.

    For spec.storageClassName, you can specify any storage class that uses a supported CSI driver.

  2. Apply the manifest:

    kubectl apply -f my-pvc.yaml
    
  3. To create a Deployment, save the following manifest as my-deployment.yaml:

    apiVersion: apps/v1
    kind: Deployment
    metadata:
      name: hello-app
    spec:
      selector:
        matchLabels:
          app: hello-app
      template:
        metadata:
          labels:
            app: hello-app
        spec:
          containers:
          - name: hello-app
            image: google/cloud-sdk:slim
            args: [ "sleep", "3600" ]
            volumeMounts:
            - name: sdk-volume
              mountPath: /usr/share/hello/
          volumes:
          - name: sdk-volume
            persistentVolumeClaim:
              claimName: my-pvc
    
  4. Apply the manifest:

    kubectl apply -f my-deployment.yaml
    
  5. Check the status of the Deployment:

    kubectl get deployment hello-app
    

    It might take some time for the Deployment to become ready. You can run the preceding command until you see an output similar to the following:

    NAME        READY   UP-TO-DATE   AVAILABLE   AGE
    hello-app   1/1     1            1           2m55s
    

Add a test file to the volume

  1. List the Pods in the Deployment:

    kubectl get pods -l app=hello-app
    

    The output is similar to the following:

    NAME                         READY   STATUS    RESTARTS   AGE
    hello-app-6d7b457c7d-vl4jr   1/1     Running   0          2m56s
    
  2. Create a test file in a Pod:

    kubectl exec POD_NAME \
        -- sh -c 'echo "Hello World!" > /usr/share/hello/hello.txt'
    

    Replace POD_NAME with the name of the Pod.

  3. Verify that the file exists:

    kubectl exec POD_NAME \
        -- sh -c 'cat /usr/share/hello/hello.txt'
    

    The output is similar to the following:

    Hello World!
    

Create a VolumeSnapshotClass object

Create a VolumeSnapshotClass object to specify the CSI driver and deletionPolicy for your volume snapshot. You can reference VolumeSnapshotClass objects when you create VolumeSnapshot objects.

  1. Save the following manifest as volumesnapshotclass.yaml.

    Persistent Disk

    Use the v1 API version for clusters running versions 1.21 or later.

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

    In this example:

    • The driver field is used by the CSI driver to provision the snapshot. In this example, pd.csi.storage.gke.io uses the Compute Engine Persistent Disk CSI driver.

    • ThedeletionPolicy field tells GKE what to do with the VolumeSnapshotContent object and the underlying snapshot when the bound VolumeSnapshot object is deleted. Specify Delete to delete the VolumeSnapshotContent object and the underlying snapshot. Specify Retain if you want to keep the VolumeSnapshotContent and the underlying snapshot.

      To use a custom storage location, add a storage-locations parameter to the snapshot class. To use this parameter, your clusters must use version 1.21 or later.

      apiVersion: snapshot.storage.k8s.io/v1
      kind: VolumeSnapshotClass
      metadata:
        name: my-snapshotclass
      parameters:
        storage-locations: us-east2
      driver: pd.csi.storage.gke.io
      deletionPolicy: Delete
      
    • To create a disk image, add the following to the parameters field:

      parameters:
        snapshot-type: images
        image-family: IMAGE_FAMILY
      

      Replace IMAGE_FAMILY with the name of your preferred image family, such as preloaded-data.

  2. Apply the manifest:

    kubectl apply -f volumesnapshotclass.yaml
    

Create a VolumeSnapshot

A VolumeSnapshot object is a request for a snapshot of an existing PersistentVolumeClaim object. When you create a VolumeSnapshot object, GKE automatically creates and binds it with a VolumeSnapshotContent object, which is a resource in your cluster like a PersistentVolume object.

  1. Save the following manifest as volumesnapshot.yaml.

    apiVersion: snapshot.storage.k8s.io/v1
    kind: VolumeSnapshot
    metadata:
      name: my-snapshot
    spec:
      volumeSnapshotClassName: my-snapshotclass
      source:
        persistentVolumeClaimName: my-pvc
    
  2. Apply the manifest:

    kubectl apply -f volumesnapshot.yaml
    

    After you create a Volume snapshot, GKE creates a corresponding VolumeSnapshotContent object in the cluster. This object stores the snapshot and bindings of VolumeSnapshot objects. You don't interact with VolumeSnapshotContents objects directly.

  3. Confirm that GKE created the VolumeSnapshotContents object:

    kubectl get volumesnapshotcontents
    

    The output is similar to the following:

    NAME                                               AGE
    snapcontent-cee5fb1f-5427-11ea-a53c-42010a1000da   55s
    

After the Volume snapshot content is created, the CSI driver you specified in the VolumeSnapshotClass creates a snapshot on the corresponding storage system. After GKE creates a snapshot on the storage system and binds it to a VolumeSnapshot object on the cluster, the snapshot is ready to use. You can check the status by running the following command:

kubectl get volumesnapshot \
  -o custom-columns='NAME:.metadata.name,READY:.status.readyToUse'

If the snapshot is ready to use, the output is similar to the following:

NAME               READY
my-snapshot        true

Delete the test file

  1. Delete the test file that you created:

    kubectl exec POD_NAME \
        -- sh -c 'rm /usr/share/hello/hello.txt'
    
  2. Verify that the file no longer exists:

    kubectl exec POD_NAME \
        -- sh -c 'cat /usr/share/hello/hello.txt'
    

    The output is similar to the following:

    cat: /usr/share/hello/hello.txt: No such file or directory
    

Restore the volume snapshot

You can reference a VolumeSnapshot in a PersistentVolumeClaim to provision a new volume with data from an existing volume or restore a volume to a state that you captured in the snapshot.

To reference a VolumeSnapshot in a PersistentVolumeClaim, add the dataSource field to your PersistentVolumeClaim. The same process is used whether the VolumeSnapshotContents refers to a disk image or a snapshot.

In this example, you reference the VolumeSnapshot that you created in a new PersistentVolumeClaim and update the Deployment to use the new claim.

  1. Verify if you're using a disk or image snapshot, which differ as follows:

    • Disk snapshots: Take snapshots frequently and restore infrequently.
    • Image snapshots: Take snapshots infrequently and restore frequently. Image snapshots may also be slower to create than disk snapshots.

    For details see Snapshot frequency limits. Knowing your snapshot type helps if you need to troubleshoot any issues.

    Inspect the VolumeSnapshot:

    kubectl describe volumesnapshot SNAPSHOT_NAME
    

    The volumeSnapshotClassName field specifies the snapshot class.

    kubectl describe volumesnapshotclass SNAPSHOT_CLASS_NAME
    

    The snapshot-type parameter will specify snapshots or images. If it is not given, the default is snapshots.

    If there is no snapshot class (for instance, if the snapshot was statically created), inspect the VolumeSnapshotContents. sh kubectl describe volumesnapshotcontents SNAPSHOT_CONTENTS_NAME The format of a snapshot handle in the output tells you the type of snapshot, as follows: * projects/PROJECT_NAME/global/snapshots/SNAPSHOT_NAME: disk snapshot

    • projects/PROJECT_NAME/global/images/IMAGE_NAME: image snapshot
  1. Save the following manifest as pvc-restore.yaml:

    Persistent Disk

    apiVersion: v1
    kind: PersistentVolumeClaim
    metadata:
      name: pvc-restore
    spec:
      dataSource:
        name: my-snapshot
        kind: VolumeSnapshot
        apiGroup: snapshot.storage.k8s.io
      storageClassName: standard-rwo
      accessModes:
        - ReadWriteOnce
      resources:
        requests:
          storage: 1Gi
    
  2. Apply the manifest:

    kubectl apply -f pvc-restore.yaml
    
  3. Update the my-deployment.yaml file to use the new PersistentVolumeClaim:

    ...
    volumes:
    - name: my-volume
      persistentVolumeClaim:
        claimName: pvc-restore
    
  4. Apply the updated manifest:

    kubectl apply -f my-deployment.yaml
    

Check that the snapshot restored successfully

  1. Get the name of the new Pod that GKE creates for the updated Deployment:

     kubectl get pods -l app=hello-app
    

Verify that the test file exists:

   kubectl exec NEW_POD_NAME \
       -- sh -c 'cat /usr/share/hello/hello.txt'

Replace NEW_POD_NAME with the name of the new Pod that GKE created.

The output is similar to the following:

   Hello World!

Import a pre-existing snapshot

You can use an existing volume snapshot created outside the current cluster to manually provision the VolumeSnapshotContents object. For example, you can populate a volume in GKE with a snapshot of another Google Cloud resource created in a different cluster.

  1. Locate the name of your snapshot.

    Google Cloud console

    Go to https://console.cloud.google.com/compute/snapshots.

    Google Cloud CLI

    Run the following command:

    gcloud compute snapshots list
    

    The output is similar to the following:

    NAME                                           DISK_SIZE_GB  SRC_DISK                                                     STATUS
    snapshot-5e6af474-cbcc-49ed-b53f-32262959a0a0  1             us-central1-b/disks/pvc-69f80fca-bb06-4519-9e7d-b26f45c1f4aa READY
    
  2. Save the following VolumeSnapshot manifest as restored-snapshot.yaml.

    apiVersion: snapshot.storage.k8s.io/v1
    kind: VolumeSnapshot
    metadata:
      name: restored-snapshot
    spec:
      volumeSnapshotClassName: my-snapshotclass
      source:
        volumeSnapshotContentName: restored-snapshot-content
    
  3. Apply the manifest:

    kubectl apply -f restored-snapshot.yaml
    
  4. Save the following VolumeSnapshotContent manifest as restored-snapshot-content.yaml. Replace the snapshotHandle field with your project ID and snapshot name. Both volumeSnapshotRef.name and volumeSnapshotRef.namespace must point to the previously created VolumeSnapshot for the bi-directional binding to be valid.

    apiVersion: snapshot.storage.k8s.io/v1
    kind: VolumeSnapshotContent
    metadata:
      name: restored-snapshot-content
    spec:
      deletionPolicy: Retain
      driver: pd.csi.storage.gke.io
      source:
        snapshotHandle: projects/PROJECT_ID/global/snapshots/SNAPSHOT_NAME
      volumeSnapshotRef:
        kind: VolumeSnapshot
        name: restored-snapshot
        namespace: default
    
  5. Apply the manifest:

    kubectl apply -f restored-snapshot-content.yaml
    
  6. Save the following PersistentVolumeClaim manifest as restored-pvc.yaml. The Kubernetes storage controller will find a VolumeSnapshot named restored-snapshot and then try to find, or dynamically create, a PersistentVolume as the data source. You can then utilize this PVC in a Pod to access the restored data.

    apiVersion: v1
    kind: PersistentVolumeClaim
    metadata:
      name: restored-pvc
    spec:
      dataSource:
        name: restored-snapshot
        kind: VolumeSnapshot
        apiGroup: snapshot.storage.k8s.io
      storageClassName: standard-rwo
      accessModes:
        - ReadWriteOnce
      resources:
        requests:
          storage: 1Gi
    
  7. Apply the manifest:

    kubectl apply -f restored-pvc.yaml
    
  8. Save the following Pod manifest as restored-pod.yaml referring to the PersistentVolumeClaim. The CSI driver will provision a PersistentVolume and populate it from the snapshot.

    apiVersion: v1
    kind: Pod
    metadata:
      name: restored-pod
    spec:
      containers:
      - name: busybox
        image: busybox
        args:
        - sleep
        - "3600"
        volumeMounts:
        - name: source-data
          mountPath: /demo/data
      volumes:
      - name: source-data
        persistentVolumeClaim:
          claimName: restored-pvc
          readOnly: false
    
  9. Apply the manifest:

    kubectl apply -f restored-pod.yaml
    
  10. Verify that the file has been restored:

    kubectl exec restored-pod -- sh -c 'cat /demo/data/hello.txt'
    

Clean up

To avoid incurring charges to your Google Cloud account for the resources used on this page, follow these steps.

  1. Delete the VolumeSnapshot:

    kubectl delete volumesnapshot my-snapshot
    
  2. Delete the VolumeSnapshotClass:

    kubectl delete volumesnapshotclass my-snapshotclass
    
  3. Delete the Deployment:

    kubectl delete deployments hello-app
    
  4. Delete the PersistentVolumeClaim objects:

    kubectl delete pvc my-pvc pvc-restore
    

What's next