💡 Quick Answer: Delete removes the backing storage when PVC is deleted (default for dynamic provisioning). Retain keeps the volume data for manual recovery. Never use Recycle (deprecated).

The Problem

When a PVC is deleted, what happens to the data? Without understanding reclaim policies:

Production data accidentally deleted with the PVC
Orphaned cloud volumes accumulating costs
Inability to recover data from released PVs
Confusion about why a PV shows “Released” but can’t be rebound

The Solution

Check Current Policy

kubectl get pv -o custom-columns=NAME:.metadata.name,POLICY:.spec.persistentVolumeReclaimPolicy,STATUS:.status.phase

Change Policy on Existing PV

# Change from Delete to Retain (protect data)
kubectl patch pv pv-data-01 -p '{"spec":{"persistentVolumeReclaimPolicy":"Retain"}}'

StorageClass with Retain Policy

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: retain-ssd
provisioner: ebs.csi.aws.com
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
parameters:
  type: gp3
  encrypted: "true"

Recover a Released PV

# 1. PV is in "Released" state after PVC deletion
kubectl get pv pv-data-01
# STATUS: Released

# 2. Remove the claimRef to make it Available
kubectl patch pv pv-data-01 --type=json -p='[{"op":"remove","path":"/spec/claimRef"}]'

# 3. Create a new PVC that binds to this PV

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: recovered-data
spec:
  accessModes: ["ReadWriteOnce"]
  storageClassName: ""  # Empty = manual binding
  volumeName: pv-data-01
  resources:
    requests:
      storage: 50Gi

Static PV with Retain

apiVersion: v1
kind: PersistentVolume
metadata:
  name: nfs-data
spec:
  capacity:
    storage: 100Gi
  accessModes: ["ReadWriteMany"]
  persistentVolumeReclaimPolicy: Retain
  nfs:
    server: 192.168.1.100
    path: /exports/data

stateDiagram-v2
    [*] --> Available: PV Created
    Available --> Bound: PVC Claims PV
    Bound --> Released: PVC Deleted (Retain)
    Bound --> [*]: PVC Deleted (Delete)
    Released --> Available: Remove claimRef
    Released --> [*]: Admin deletes PV

Common Issues

PV stuck in “Released” — can’t bind new PVC The claimRef still references the deleted PVC. Remove it:

kubectl patch pv <pv-name> --type=json -p='[{"op":"remove","path":"/spec/claimRef"}]'

Dynamic PVs defaulting to Delete Most CSI drivers and cloud StorageClasses default to Delete. Create a Retain StorageClass for important data.

Orphaned cloud volumes after PVC deletion with Retain Retained PVs keep the cloud volume. Clean up manually:

# List Released PVs
kubectl get pv --field-selector=status.phase=Released
# Delete PV (doesn't delete cloud volume)
kubectl delete pv <pv-name>
# Delete cloud volume separately
aws ec2 delete-volume --volume-id vol-abc123

Best Practices

Use Retain for production databases and stateful workloads
Use Delete for ephemeral/reproducible data (caches, build artifacts)
Patch existing PVs to Retain before dangerous operations
Create a retain-* StorageClass variant for important workloads
Monitor Released PVs to prevent storage cost creep
Use Velero or CSI snapshots as backup — don’t rely on Retain alone
Document recovery procedures for your team

Key Takeaways

Delete: PV and backing storage removed when PVC is deleted
Retain: PV moves to “Released” state, data preserved, manual recovery possible
Recycle: Deprecated — don’t use (was basic rm -rf)
StorageClass reclaimPolicy sets the default for dynamically provisioned PVs
Released PVs need claimRef removed before rebinding
Always use Retain + snapshots for production stateful data

PersistentVolume Reclaim Policies