Storage Snapshots and CSI Related Resources

A Kubernetes volume snapshot (VolumeSnapshot) is a read-only copy of a storage volume at a point in time. It is implemented by CSI (Container Storage Interface) storage drivers and is commonly used for:

Scenario	Description
Data backup / recovery	Take scheduled snapshots of the PVCs of important applications, and restore data from a snapshot when a problem occurs
Rollback insurance before an app upgrade	Take a snapshot before an upgrade, so you can quickly return to the pre-upgrade state if the upgrade fails
Clone / replicate environments	Create new volumes from a snapshot for testing, joint debugging, or replicating production data
Data migration	Use snapshots to move data to a new volume / new StorageClass

Snapshots vs. "backups"

Snapshots rely on the storage backend implementation and usually only record the changed data blocks (incremental). They are fast and occupy little space, but they cannot replace off-site disaster recovery backups. Real offline backups still require tools such as Velero to move data out of the cluster.

This article covers the pages under the Config & Storage → Storage group in Kuboard: Storage Snapshots, Storage Snapshot Classes, and four read-only resource pages — CSI Drivers, CSI Nodes, CSI Storage Capacity, and Volume Attachments.

Menu hidden by default

Storage Snapshots, Snapshot Classes, and CSI-related resources are not shown in the left menu by default. If you cannot see these menu items, ask the cluster administrator to enable the corresponding resource items under System Settings → Menu Item Settings; the menu takes effect immediately after being enabled.

Prerequisites

Condition	Description
The CSI driver supports snapshots	The CSI driver of the underlying storage must implement snapshot capability (e.g., Alibaba Cloud disk.csi.aliyun.com); not all storage supports it
Snapshot CRDs installed	The cluster must have the three CRDs from external-snapshotter: volumesnapshots, volumesnapshotclasses, volumesnapshotcontents (snapshot.storage.k8s.io/v1). If the snapshot menu is not visible in the cluster or the resource page reports it is unavailable, deploy the external snapshot controller first
A VolumeSnapshotClass exists	A corresponding VolumeSnapshotClass must exist before creating a snapshot (see below)
Source PVC ready	The PVC to be snapshotted must be in the Bound state, and its StorageClass must be provisioned by a CSI driver that supports snapshots

How to tell whether a driver supports snapshots

Open the Storage → CSI Drivers detail page to view driver features such as volumeLifecycleModes; a more direct basis is whether a usable VolumeSnapshotClass already exists in the cluster, with its driver matching the StorageClass of the source PVC.

Create a VolumeSnapshot

Entry: Config & Storage → Storage → Storage Snapshots, click Create.

Form fields (all required):

Field	Corresponding spec field	Description
Name	metadata.name	The snapshot name, isolated by namespace
Snapshot Class	spec.volumeSnapshotClassName	Which VolumeSnapshotClass to use (text input, e.g. demo-snapshot-class)
Source PVC	spec.source.persistentVolumeClaimName	Which PVC to take a snapshot of (text input, e.g. my-pvc)

Steps:

1. At the top of the list page, select the cluster and namespace (snapshots are namespace-scoped resources);
2. Click Create and fill in the name, snapshot class, and source PVC;
3. Click Save and confirm in the YAML Preview dialog that pops up before creating;
4. Return to the list and wait until the snapshot's ready status becomes Ready before using it.

Snapshots are created asynchronously

A snapshot does not "complete" immediately after creation. K8s calls the CSI driver in the background to take the snapshot. On the detail page you can see the ready status (readyToUse): only after it becomes Ready is the bound VolumeSnapshotContent generated, and only then can it be used for restore.

View the Snapshot List and Details

List page: consistent with other K8s resource lists, supporting cluster / namespace switching, searching by name, and batch deletion.

Detail page contains two sections:

Section	Content shown
Spec	Snapshot class, source PVC
Status	Ready status (readyToUse, Ready shown as a green label); bound snapshot content (boundVolumeSnapshotContentDataObjectRef)

The "bound snapshot content" on the detail page is a read-only reference link; clicking it jumps to the corresponding VolumeSnapshotContent (snapshot content) to view the underlying snapshot data. VolumeSnapshotContent is a cluster-scoped resource without a dedicated menu page; it is shown here read-only to avoid accidental cross-namespace editing.

Restore from a Snapshot / Create a New PVC

"Restoring" from a snapshot essentially creates a new PVC using the snapshot as the data source, then attaches the workload to the new PVC:

1. On the Storage → PVC page, click Create to open the creation form;
2. The form has no direct "data source" field; click Save to enter the YAML Preview dialog;
3. In the YAML, add dataSource to spec, pointing to the snapshot to restore from:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: data-restored
  namespace: default
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: <StorageClass using the same driver as the snapshot>
  resources:
    requests:
      storage: 10Gi
  dataSource:
    name: <snapshot name>
    kind: VolumeSnapshot
    apiGroup: snapshot.storage.k8s.io

4. After confirming, create it; the content of the new volume is the data at the moment of the snapshot.

Prerequisites for restore

• The snapshot must be Ready (readyToUse = true);
• The StorageClass of the new PVC must use the same CSI driver as the snapshot (usually the same StorageClass is the safest);
• The size of the restored volume should generally not be smaller than the source volume; the specifics depend on whether the storage backend supports snapshot restore.

VolumeSnapshotClass (Snapshot Class)

A VolumeSnapshotClass is a cluster-scoped resource that defines two pieces of information for snapshots: which CSI driver takes the snapshot, and the behavior when a snapshot is deleted.

Field	Description
Name	The snapshot class name
Driver (driver)	The name of the CSI driver that provides snapshot capability, e.g. disk.csi.aliyun.com
Deletion policy (deletionPolicy)	Delete: deletes the backend snapshot data when the VolumeSnapshot is deleted; Retain: keeps the backend data after the VolumeSnapshot is deleted

Creation steps: Config & Storage → Storage → Storage Snapshot Classes → Create, fill in the name, select the driver and deletion policy (default Delete), then save.

When do you need multiple snapshot classes

Different storage drivers each need their own snapshot class; the same driver can also be split into several by deletion policy (e.g., one for "keep for record" and one for "temporary snapshots"). Creating a VolumeSnapshot requires specifying an existing snapshot class.

Choosing the deletionPolicy

Delete removes the backend data when the snapshot is deleted, which is unrecoverable; for snapshots that need long-term retention, the Retain policy is recommended, or confirm the snapshot is no longer needed before using the Delete policy.

CSI Related Resource Pages

All four pages below are read-only (no create / edit entry) and are used to view the running state of the CSI ecosystem, typically during troubleshooting or capacity planning.

Menu item	Resource object	Content shown	When to use
CSI Drivers	CSIDriver	CSI drivers registered in the cluster and their features: attach required (attachRequired), pod info on mount (podInfoOnMount), storage capacity (storageCapacity), FS group policy (fsGroupPolicy), volume lifecycle modes (volumeLifecycleModes)	Confirm whether a storage supports snapshots / capacity-aware scheduling / specific volume modes; compare driver capabilities
CSI Nodes	CSINode	The CSI drivers registered on each node (node ID, topology keys, allocatable volume count)	Troubleshooting: whether a driver is registered on a node, whether volumes can be scheduled to that node
CSI Storage Capacity	CSIStorageCapacity	The allocatable capacity of a StorageClass in a topology region, max / min volume size, node topology	Under capacity-aware scheduling (the storageCapacity feature), check whether a volume can be created in a specified region
Volume Attachments	VolumeAttachment	Which node a PV is attached to, which driver attached it, whether it is attached (attached), attachment metadata, detach errors (detachError)	Troubleshooting: volumes fail to attach, abnormal volume status after node migration

Look at these pages only when needed

The four CSI resource pages are for "viewing" and are not used in daily storage management; when "volume creation / attach / scheduling" problems occur, troubleshoot in the order Volume Attachments → CSI Nodes → CSI Storage Capacity → CSI Drivers.

Common Notes

Note	Description
Snapshots depend on the CSI driver	When a StorageClass is provisioned by a driver that does not support snapshots, that PVC cannot be snapshotted, and the snapshot will remain not ready after creation
Snapshots consume backend quota	Cloud disk snapshots usually have quantity and capacity limits and keep consuming storage quota; regularly clean up snapshots you no longer need
The deletion policy decides where the data goes	Before deleting a snapshot, confirm whether its snapshot class is Retain (keep) or Delete (delete the backend data along with it)
Restore only after ready	A snapshot that is not ready (readyToUse = false) cannot be used as a PVC data source
Snapshot class and StorageClass must match	The StorageClass of a restored / cloned new volume should use the same CSI driver as the snapshot
Menu hidden by default	All resources on this page are hidden in the menu by default and must be enabled by an administrator under System Settings → Menu Item Settings

Related Pages

• PVC / PV / StorageClass: the "upstream" resources of snapshots; see that page for the relationship between snapshot classes and StorageClasses
• ConfigMap and Secret: sibling resources under the same "Config & Storage" group