Skip to main content

Usage

Dynamic provisioning (RWO)​

Create a PVC using the ipfs-csi storage class from the chart defaults:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: ipfs-pvc
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi
  storageClassName: ipfs-csi

Attach the claim to a pod and write data:

apiVersion: v1
kind: Pod
metadata:
  name: ipfs-app
spec:
  containers:
    - name: app
      image: busybox:1.36
      command: ["sh", "-c", "echo 'hello from ipfs csi' > /data/hello.txt && sleep 3600"]
      volumeMounts:
        - name: data
          mountPath: /data
  volumes:
    - name: data
      persistentVolumeClaim:
        claimName: ipfs-pvc

Read-only CID source​

The driver also supports read-only mount scenarios where content is resolved from a fixed CID source. Use a dedicated StorageClass/PVC workflow from your cluster policy when enabling this mode.

Expand volume​

Resize by increasing PVC request size:

kubectl patch pvc ipfs-pvc -p '{"spec":{"resources":{"requests":{"storage":"2Gi"}}}}'

The controller handles ControllerExpandVolume, and node-side expansion is processed during publish/expand operations.

Snapshots​

See Snapshot and Restore for complete flow and requirements.

Common edge cases​

  • If mount paths already contain stale content, node publish/stage logic cleans and re-materializes content.
  • Snapshot restore requires both snapshot CRDs and running snapshot-controller.
  • Missing or mismatched snapshot class/driver name causes restore provisioning failures.

The E2E suites in test/e2e/suites are the canonical executable reference.