Two settings decide who may use a volume and what happens when they stop.
Access modes
| Mode | Short | Means |
|---|---|---|
ReadWriteOnce | RWO | One node mounts it read-write |
ReadOnlyMany | ROX | Many nodes mount it read-only |
ReadWriteMany | RWX | Many nodes mount it read-write |
ReadWriteOncePod | RWOP | One Pod, read-write |
A claim binds only to a volume that offers every mode it asks for. Most block disks are RWO only; RWX needs a shared filesystem such as NFS. A claim asking for RWX on a cluster that cannot provide it stays Pending, and describe pvc says so.
Reclaim policies
What happens to a PV when its claim is deleted:
claim deletedkubectl delete pvc
gonevolume and data removed
claim deletedkubectl delete pvc
Releaseddata kept, claimRef still set
Delete is the default for dynamically provisioned volumes. Retain is for data you cannot afford to lose by accident.
Released is not Available
text
learner@shallows:~$ kubectl get pv
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM
ledger-pv 1Gi RWO Retain Released depths/ledgerThe data is on disk. spec.claimRef still names the deleted claim, so no other claim can bind. To reuse the volume, clear the reference:
text
learner@shallows:~$ kubectl patch pv ledger-pv -p '{"spec":{"claimRef":null}}'
persistentvolume/ledger-pv patchedThe phase returns to Available and the next matching claim binds, data included.