A persistent volume claim (PVC) reserves storage for a pod without tying that storage to the pod’s lifetime. The claim requests capacity from a persistent volume (PV); once bound, that PV is unavailable to other claims. This lets an application use persistent data without managing the underlying storage directly. section 1.4.4 section 7.2
For a storage task, work from the request to the running pod:
- Create or identify an available PV, then create a PVC requesting the needed capacity and access mode. A claim need not request the PV’s exact size: a 90 MB request can bind to a 100 MB volume. section 7.1 section 7.2
- Mount the claimed volume in a pod or Deployment. Check the binding with `kubectl get pvc` and `kubectl get pv`, then check the workload and its mount with `kubectl get po` and `kubectl describe po`. section 7.1
- If the task specifies an existing storage class, select it in the PVC instead of creating a PV manually; the volume can be provisioned automatically. section 7.5
Claim and verify a hostPath volume
Create a 100 MB hostPath PV backed by `/mnt/data` on a worker node, then create a PVC requesting 90 MB. Mount the claim in a Deployment at `/usr/share/nginx/html`. The resulting `kubectl get pvc` output shows the claim as `Bound` to the 100Mi PV; `kubectl get pv` shows the same binding from the volume side. section 7.1
root@kind-control-plane:/# kubectl get pvc NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE claim-02833 Bound vol02833 100Mi RWO manual 4m root@kind-control-plane:/# kubectl get pv NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS vol02833 100Mi RWO Retain Bound default/claim-02833 manual
Compare the PVC’s `VOLUME` and `STATUS` columns with the PV’s `CLAIM` and `STATUS` columns to confirm both sides of the binding. section 7.1
A bound claim also protects storage still in use. In the worked deletion test, forcing deletion of the PV leaves it `Terminating` until the PVC is deleted; forcing deletion of the PVC leaves it `Terminating` while the Deployment’s pod still uses it. Remove the workload before expecting the claim and volume deletion to finish. section 7.2
