This is unreleased documentation for SUSE® Storage 1.13 (Dev).

CSI Volume Group Snapshot

A VolumeGroupSnapshot takes snapshots of a set of Longhorn volumes as one group with a single request. SUSE Storage tracks each group in a SnapshotGroup custom resource in the longhorn-system namespace and creates one Longhorn snapshot per member volume. Kubernetes then returns one VolumeSnapshot per member PVC, all bound to the same group, so at restore time it is clear which snapshots belong together.

Snapshot groups can also be created and managed without CSI, from the SUSE Storage UI or with kubectl. For details, see Create a Snapshot Group.

Consistency boundary: Longhorn snapshots each member volume independently, so member snapshots are taken at slightly different times, not at one instant. A group snapshot is crash-consistent per volume; it does not guarantee application consistency across volumes. To get a recoverable set, include every volume the application needs (for example data, log, and metadata volumes) and quiesce or pause the application while the group snapshot is taken. See #2128 for the planned application-consistent snapshot work that builds on this feature.

Prerequisites

CSI Volume Group Snapshot support must be enabled on your cluster and in SUSE Storage. See Enable CSI Volume Group Snapshot Support.

Create a CSI VolumeGroupSnapshot Associated with a Longhorn Snapshot Group

First, label the PVCs that should be snapshotted together:

kubectl label pvc test-vol-1 test-vol-2 app-group=demo

Create a VolumeGroupSnapshotClass with the parameter type set to snap:

apiVersion: groupsnapshot.storage.k8s.io/v1
kind: VolumeGroupSnapshotClass
metadata:
  name: longhorn-group-snap-vgsc
driver: driver.longhorn.io
deletionPolicy: Delete
parameters:
  type: snap

Then create a VolumeGroupSnapshot in the application namespace with a selector matching the PVC labels:

apiVersion: groupsnapshot.storage.k8s.io/v1
kind: VolumeGroupSnapshot
metadata:
  name: test-group-snapshot
spec:
  volumeGroupSnapshotClassName: longhorn-group-snap-vgsc
  source:
    selector:
      matchLabels:
        app-group: demo

Result: SUSE Storage creates one SnapshotGroup custom resource and one snapshot per member volume. Once the group is ready, the snapshot-controller creates one VolumeSnapshot per member PVC, each bound to the group. The VolumeGroupSnapshotContent records the group as snap://group-name in its volumeGroupSnapshotHandle field; each member VolumeSnapshotContent uses the existing snap://volume-name/snapshot-name format.

You can observe the group’s progress with kubectl:

kubectl -n longhorn-system get snapshotgroups
NAME                                                 PHASE   READYTOUSE   CREATIONTIME           AGE
groupsnapshot-59e29b09-2661-4faf-9862-1571f8e68e94   Ready   true         2026-08-25T02:11:05Z   1m

The group starts in InProgress and becomes Ready when every member snapshot is taken, or Failed if the completion deadline (300 seconds by default) passes first. A failed CSI-created group is cleaned up automatically, and the snapshot-controller retries with a fresh group.

Create a CSI VolumeGroupSnapshot Associated with Longhorn Backups

A VolumeGroupSnapshotClass with the parameter type set to bak requests a group backup instead: the same group snapshot is taken first, then each member snapshot is uploaded to the backup target.

apiVersion: groupsnapshot.storage.k8s.io/v1
kind: VolumeGroupSnapshotClass
metadata:
  name: longhorn-group-bak-vgsc
driver: driver.longhorn.io
deletionPolicy: Delete
parameters:
  type: bak

The group reports ready only when every member backup completes. The group handle uses the bak://group-name format, and each member uses the existing bak://volume-name/backup-name format.

The group label is stored in each member backup’s metadata on the backup target, so a restore from another cluster can find the whole set of backups belonging to one group.

The optional backupMode parameter sets the backup mode for the member backups: incremental (default) or full. For more information about backupMode, see Create A Backup.

The type parameter must be snap or bak. All other class parameters (except the reserved backupMode) are applied to every member snapshot as Longhorn snapshot labels.

Restore a Member PVC

Each member VolumeSnapshot is a normal VolumeSnapshot. Restore it on its own by creating a PVC whose dataSource points to it, exactly as for a per-volume snapshot. For details, see CSI VolumeSnapshot Associated with Longhorn Snapshot and CSI VolumeSnapshot Associated with Longhorn Backup.

Delete a CSI VolumeGroupSnapshot

Delete the VolumeGroupSnapshot object. With a Delete deletion policy, this deletes the VolumeGroupSnapshotContent, the Longhorn SnapshotGroup, and all of its member snapshots (for a bak group, the member backups as well). Kubernetes does not allow deleting the member VolumeSnapshot objects individually while they belong to a group.

Limitations

  • A group is capped at 64 member volumes.

  • The group name is at most 54 characters, because member snapshots are named <group-name>-<8-character-suffix> and the result must fit the 63-character name limit.

  • deadlineSeconds sets the time allowed for taking all member snapshots; the default is 300 and the valid range is 10 to 3600. It does not limit the backup uploads of a bak group.

  • A snapshot group cannot be edited after creation: a group is a point-in-time request.

  • Member volumes must be eligible for a snapshot at creation time: standby (DR), restoring, and faulted volumes are rejected. Detached volumes are eligible and are automatically attached.

  • After a group becomes Ready, deleting one of its member snapshots marks the group Degraded. SUSE Storage does not take a replacement snapshot, because it would not match the group’s point in time.

History

Available since v1.13.0.