Rok Snapshots¶

Rok snapshots allow you to instantly snapshot your Rok volumes to create a point-in-time copy of your data for local and offsite backups. Rok takes immutable, group consistent snapshots of your applications and stores these snapshots in a backup store, for example, Amazon S3.

Combine this with Rok’s instant cloning, and you can recover from hardware failure in minutes, regardless of the volume’s capacity.

To take a volume snapshot, use the rok volume snapshot class:

apiVersion: snapshot.storage.k8s.io/v1beta1
kind: VolumeSnapshotClass
metadata:
  name: rok
parameters:
  # NOTE: Increase this in case of heavy I/O during snapshots
  rok/snapshot-cow-size: 1GiB
driver: rok.arrikto.com
deletionPolicy: Delete

Parameters¶

If the pre-defined volume snapshot class doesn’t meet your requirements, you can create your own customized volume snapshot class, based on the pre-defined one.

Rok supports the following volume snapshot class parameters to further customize Rok snapshots:

rok/snapshot-chunk-size
- Type: string
- Values: size string
- Default: “128KiB”
The chunk size for snapshots. It defines the unit of Copy On Write for snapshots. The value must be a power of 2 between 4KiB and 512KiB.
rok/snapshot-cow-size
- Type: string
- Values: size string
- Default: “10GiB”
The size of the snapshot Copy On Write space. Increase this in case of heavy I/O during snapshots.
rok/pre-snapshot-verification
- Type: string
- Values: “true” | “false”
- Default: “true”
Verify the integrity of the filesystem on the volume, before taking a snapshot. If the filesystem has errors, Rok refuses to snapshot the volume.
rok/post-snapshot-verification
- Type: string
- Values: “true” | “false”
- Default: “false”
Verify the integrity of the filesystem in the snapshot, after creating it. This is an expensive operation that runs a full filesystem check, and may induce additional costs, since it access the snapshot data in the snapshot store, for example, S3.

Note

To affect only a specific Rok snapshot you can specify any of the volume snapshot class parameters as a VolumeSnapshot annotation.

Snapshot PVC Backed by Rok¶

Create a VolumeSnapshot object and set the VolumeSnapshotClass to rok:

apiVersion: snapshot.storage.k8s.io/v1beta1
kind: VolumeSnapshot
metadata:
  name: mysnap
spec:
  volumeSnapshotClassName: rok
  source:
    persistentVolumeClaimName: mypvc

Create Volume from a Kubernetes Snapshot¶

You can create a PVC from a Kubernetes VolumeSnapshot object. You need to have created this object using the rok volume snapshot class.

Create a PVC object and set dataSource to point to the Rok snapshot:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: mysnap-pvc
spec:
  storageClassName: rok
  dataSource:
    name: mysnap
    kind: VolumeSnapshot
    apiGroup: snapshot.storage.k8s.io
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi

Create PVC from a Rok Resource¶

Create a new PVC and set the rok/origin PVC annotation to the URL of the Rok object:

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: mypvc
  annotations:
    rok/origin: https://example.com/swift/v1/<object-url>
spec:
  storageClassName: rok
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi

Create StatefulSet from a Rok Group Resource¶

Rok supports passing a group resource as a rok/origin for a PVC. This is useful when you want to create a StatefulSet with a volumeClaimTemplate and assign a unique disk from a group resource to each Pod.

An example StatefulSet looks like this:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: web
spec:
  serviceName: "nginx"
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: k8s.gcr.io/nginx-slim:0.8
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: www
          mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: www
      annotations:
        rok/origin: https://example.com/swift/v1/<group-object-url>
    spec:
      storageClassName: rok
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

The order of the resources within the group is important, as each Pod will get the corresponding disk (e.g., web-0 will get the first object from the group, web-1 the second one, etc.).

If the number of replicas of the StatefulSet exceeds the number of the objects in the group, the extra Pods will get an empty disk.