Mount persistent storage for workloads

​

Reference a PVC in a Workload

• Reference the PVC through persistentVolumeClaim.claimName in spec.volumes.
• Set the volume name and mount directory in the container’s volumeMounts.

apiVersion: v1
kind: Pod
metadata:
  name: training-task
  namespace: default
spec:
  containers:
    - name: training
      image: nginx:1.14.2
      volumeMounts:
        - name: training-data
          mountPath: /data
  volumes:
    - name: training-data
      persistentVolumeClaim:
        claimName: training-data-pvc

​

Submit and confirm the mount result

1. Log in to the GPU Cloud console and go to the Clusters page.
2. Click the name of the target cluster to go to the cluster workspace.
3. In the left navigation pane, go to the Workloads page, select the appropriate workload type, and submit the YAML.
4. Verify that the workload is running properly and that the container can read and write data through the target mount directory. If a Pod remains in Pending for an extended period or a container cannot start, check whether the PVC is Bound, whether the PVC and workload are in the same namespace, and whether claimName is correct.

​

Unmount PVC

• For workloads managed by controllers such as Deployment, StatefulSet, and Job, update the Pod template and remove the corresponding volumeMounts and volumes configurations.
• For directly created Pods, stop and delete the Pod, or recreate it using a configuration that no longer references the PVC.