PersistentVolumes & Claims

Decoupling Provisioning from Consumption

PersistentVolumes are cluster-scope storage objects created by an admin or provisioner; PersistentVolumeClaims are namespace-scope requests from workloads. Pods never reference a PV directly — they mount a claim.

Binding matches claim requirements — access mode, capacity, storageClassName, and optional label selectors — against available PVs, and once bound the pair stays locked. Access modes state concurrency:

• ReadWriteOnce — one node read-write.
• ReadOnlyMany — many nodes may read.
• ReadWriteMany — many nodes may write (needs filesystem support such as NFS or EFS).

The separation lets storage admins offer capacity without knowing which team will consume it, and lets developers request storage without knowing the backend.

A Claim and Its Pod

apiVersion: v1
kind: PersistentVolumeClaim
metadata: { name: data }
spec:
  accessModes: [ReadWriteOnce]
  storageClassName: standard
  resources: { requests: { storage: 10Gi } }
---
apiVersion: v1
kind: Pod
metadata: { name: app }
spec:
  containers:
  - name: app
    image: nginx:1.27
    volumeMounts: [{ name: data, mountPath: /var/lib/data }]
  volumes:
  - name: data
    persistentVolumeClaim: { claimName: data }

With dynamic provisioning you need a default StorageClass (annotated storageclass.kubernetes.io/is-default-class: "true") or an explicit storageClassName. Reclaim policy decides fate: Delete removes the backing volume when released, Retain parks it for manual recovery.

Pitfalls and Interview Angles

Pitfall 1 — misreading ReadWriteOnce. RWO restricts to one node, not one pod — several pods on that node may share it, pods elsewhere cannot. Multi-host writers need RWX-capable storage or an operator managing access.

Pitfall 2 — delete order and costs. Deleting a bound PVC while pods mount it blocks on the protection finalizer until they exit; with Retain policy, orphaned cloud volumes keep billing after the claim is gone.

Interview-ready: recite the binding criteria, the three access modes with an example each, and reclaim behavior — plus why databases typically pair RWO claims with StatefulSet volumeClaimTemplates and why WaitForFirstConsumer matters for zonal disks.