Using VolumeSnapshots in Kubernetes

With VolumeSnapshots, you create a point-in-time snapshot of a PersistentVolumeClaim (PVC) in your Kubernetes cluster. This lets you capture the contents of a volume at a specific moment, for example before performing an application upgrade or before changing data.

In this guide, we explain when to use VolumeSnapshots, what the costs are, how to create a snapshot, and how to restore a snapshot to a new PersistentVolumeClaim.

Before you begin

For this guide, you need the following:

• A Kubernetes cluster with an existing PersistentVolumeClaim.
• kubectl with access to your cluster.
• A volume created with the TransIP CSI driver. For example, use the StorageClass transip-fast-storage or transip-big-storage.

First, check whether VolumeSnapshots are available in your cluster:

kubectl api-resources | grep -i volumesnapshot

Then check which VolumeSnapshotClass is available:

kubectl get volumesnapshotclass

In a TransIP Kubernetes cluster, you use the VolumeSnapshotClass team-blue-k8s. This class uses the driver bs.csi.transip.nl.

Finally, check the PersistentVolumeClaim you want to create a snapshot of:

kubectl get pvc -n <namespace>

Replace <namespace> with the namespace where your PVC is located.

Creating a VolumeSnapshot

In this example, we create a snapshot of a PVC named webdata in the namespace productie. Replace these names with the names from your own cluster.

Step 1

Create a file named volume-snapshot.yaml:

nano volume-snapshot.yaml

Add the following contents to the file:

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshot
metadata:
  name: webdata-before-upgrade
  namespace: productie
spec:
  volumeSnapshotClassName: team-blue-k8s
  source:
    persistentVolumeClaimName: webdata

Replace the following values:

• webdata-before-upgrade: the name of the snapshot.
• productie: the namespace where your PVC is located.
• webdata: the name of the PVC you are creating a snapshot of.

Save the changes and close the file (ctrl + x > y > enter).

Step 2

Create the VolumeSnapshot:

kubectl apply -f volume-snapshot.yaml

Step 3

Check whether the snapshot is ready to use:

kubectl get volumesnapshot -n productie

Wait until READYTOUSE is set to true. You can use this command if needed:

kubectl wait --for=jsonpath='{.status.readyToUse}'=true volumesnapshot/webdata-before-upgrade -n productie --timeout=10m

View additional details with:

kubectl describe volumesnapshot webdata-before-upgrade -n productie

Restoring a VolumeSnapshot

A restore from a VolumeSnapshot creates a new PersistentVolumeClaim. Kubernetes does not overwrite the original PVC. Only connect your application to the restored PVC after you have verified that the data is correct.

Step 1

Check which StorageClass the original PVC uses:

kubectl get pvc webdata -n productie -o jsonpath='{.spec.storageClassName}{"\n"}'

Use the same StorageClass for the restore, for example transip-fast-storage or transip-big-storage.

Step 2

Create a file named volume-snapshot-restore.yaml:

nano volume-snapshot-restore.yaml

Add the following contents to the file:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: webdata-restore
  namespace: productie
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: transip-fast-storage
  resources:
    requests:
      storage: 20Gi
  dataSource:
    name: webdata-before-upgrade
    kind: VolumeSnapshot
    apiGroup: snapshot.storage.k8s.io

Replace the following values:

• webdata-restore: the name of the new PVC.
• productie: the namespace where the snapshot is located.
• transip-fast-storage: the StorageClass of the new PVC.
• 20Gi: the size of the new PVC. Use at least the restoreSize of the snapshot.
• webdata-before-upgrade: the name of the VolumeSnapshot.

Save the changes and close the file (ctrl + x > y > enter).

Step 3

Create the restored PVC:

kubectl apply -f volume-snapshot-restore.yaml

Check whether the PVC has been created:

kubectl get pvc webdata-restore -n productie

Wait until STATUS is set to Bound.

Step 4

Check the data on the restored PVC before changing your production workload. You can do this by creating a temporary pod, for example:

apiVersion: v1
kind: Pod
metadata:
  name: check-restored-volume
  namespace: productie
spec:
  restartPolicy: Never
  containers:
    - name: shell
      image: busybox:1.36
      command: ["sh", "-c", "ls -la /data && sleep 3600"]
      volumeMounts:
        - name: data
          mountPath: /data
  volumes:
    - name: data
      persistentVolumeClaim:
        claimName: webdata-restore

For example, save this file as check-restored-volume.yaml and create the pod:

kubectl apply -f check-restored-volume.yaml

Then view the contents of the volume:

kubectl exec -n productie check-restored-volume -- ls -la /data

Delete the temporary pod after you have checked the data:

kubectl delete pod check-restored-volume -n productie

Connecting your workload to the restored PVC

After you have checked the data, connect your application to the restored PVC. The exact steps depend on your workload and deployment method.

If you use a Deployment with a fixed claimName, scale the Deployment down first:

kubectl scale deployment <deploymentnaam> -n productie --replicas=0

Then update your manifest, Helm chart, or Kustomize configuration so that the workload uses webdata-restore instead of webdata:

volumes:
  - name: data
    persistentVolumeClaim:
      claimName: webdata-restore

Apply the updated configuration and then scale your Deployment back up:

kubectl apply -f deployment.yaml
kubectl scale deployment <deploymentnaam> -n productie --replicas=1

If you use a StatefulSet, first check how your application uses the PVC. A StatefulSet often creates PVCs through volumeClaimTemplates. These PVC names are linked to the StatefulSet and pod name. In that case, first create a restore plan that fits your application, for example by copying data from the restored PVC back to a new volume with the expected name.

Cleaning up VolumeSnapshots

Delete snapshots you no longer need. The VolumeSnapshotClass team-blue-k8s uses deletionPolicy Delete. This means that the underlying snapshot in the storage environment is also deleted when you delete the VolumeSnapshot object.

kubectl delete volumesnapshot webdata-before-upgrade -n productie

Then check whether the snapshot is gone:

kubectl get volumesnapshot -n productie

Note: deleting a VolumeSnapshot does not automatically delete PVCs that you previously restored from that snapshot. Delete restored PVCs separately when you no longer use them.

Common issues

• kubectl does not recognize VolumeSnapshot: check whether you are connected to the correct cluster and whether the VolumeSnapshot CRDs are available.
• The snapshot remains on READYTOUSE false: first wait a few minutes, then check the events with kubectl describe volumesnapshot <snapshotnaam> -n <namespace>.
• The restore PVC remains Pending: check whether the StorageClass exists, whether the requested storage size is at least equal to the restoreSize, and whether your dataSource points to the correct VolumeSnapshot.
• Your workload does not start with the restored PVC: check whether the accessMode matches your workload. A ReadWriteOnce volume is attached to one node at a time.
• You want to restore to a different namespace: use the same namespace for the standard workflow. Cross-namespace restores require additional Kubernetes functionality and are not the basic workflow in this guide.

You have now created, checked, and restored a VolumeSnapshot to a new PersistentVolumeClaim. Use snapshots as a quick restore point for maintenance and changes, and clean up old snapshots as soon as you no longer need them.