04 - Storage: PVs, PVCs, StorageClasses¶

Storage model¶

OpenShift inherits Kubernetes storage:

• PersistentVolume (PV) - cluster-scoped storage resource (an actual disk / NFS share / cloud volume).
• PersistentVolumeClaim (PVC) - namespace-scoped request for storage.
• StorageClass (SC) - template for dynamically provisioning PVs.
• CSI driver - Container Storage Interface driver that knows how to talk to the actual storage backend.

Access modes¶

EBS / Azure Disk / vSphere volumes are RWO. NFS / Ceph FS / Azure File can be RWX.

Reclaim policies¶

When a PVC is deleted, what happens to the PV / underlying storage:

• Delete - PV and underlying storage destroyed (typical for dynamic provisioning)
• Retain - PV kept; admin must manually clean up
• Recycle - deprecated, do not use

Storage classes (cloud common)¶

oc get storageclass
oc get sc                                # alias
oc describe sc gp3-csi

Set default storage class¶

oc patch storageclass gp3-csi \
    -p '{"metadata":{"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

# Remove default annotation from another
oc patch storageclass gp2-csi \
    -p '{"metadata":{"annotations":{"storageclass.kubernetes.io/is-default-class":"false"}}}'

PersistentVolumeClaim¶

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: data
  namespace: myapp
spec:
  accessModes: [ReadWriteOnce]
  resources:
    requests:
      storage: 5Gi
  storageClassName: gp3-csi              # optional; uses default if omitted

oc apply -f pvc.yaml
oc get pvc
oc describe pvc data

If the StorageClass exists and supports the requested access mode, a PV is dynamically provisioned and the PVC reaches Bound status.

Mounting a PVC in a pod / deployment¶

Method 1: oc set volume¶

oc set volume deploy/web \
    --add --type=pvc \
    --claim-name=data \
    --mount-path=/var/lib/data \
    --name=data-volume

Method 2: Deployment YAML¶

spec:
  template:
    spec:
      containers:
      - name: web
        image: nginx
        volumeMounts:
        - name: data
          mountPath: /var/lib/data
      volumes:
      - name: data
        persistentVolumeClaim:
          claimName: data

Quick PVC + mount (one command)¶

oc set volume deploy/web \
    --add --type=pvc \
    --claim-class=gp3-csi \
    --claim-size=5Gi \
    --mount-path=/var/lib/data

This both creates a PVC and mounts it. Useful for the exam.

Static provisioning (NFS example)¶

Sometimes you'll be given an NFS share to use directly.

apiVersion: v1
kind: PersistentVolume
metadata:
  name: nfs-pv-shared
spec:
  capacity:
    storage: 10Gi
  accessModes: [ReadWriteMany]
  persistentVolumeReclaimPolicy: Retain
  nfs:
    server: nfs.example.com
    path: /exports/shared
  storageClassName: ''                   # empty = no class

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: nfs-shared
  namespace: myapp
spec:
  accessModes: [ReadWriteMany]
  resources:
    requests: { storage: 10Gi }
  storageClassName: ''
  volumeName: nfs-pv-shared              # bind to specific PV

Volume expansion¶

If the StorageClass has allowVolumeExpansion: true, you can grow a PVC:

oc edit pvc data
# Change spec.resources.requests.storage from 5Gi to 10Gi

The CSI driver expands the underlying volume. The pod may need a restart for the filesystem to grow (oc rollout restart deploy/web).

oc get pvc data
# STATUS should be Bound; CAPACITY shows new size after expansion completes

ReadWriteMany (RWX) options¶

When you need many pods to write to the same volume:

For most exam scenarios, RWO with EBS-class storage is what you'll use.

Common storage exam tasks¶

Provision 5 GB persistent storage for a deployment¶

oc set volume deploy/web \
    --add --type=pvc \
    --claim-class=gp3-csi \
    --claim-size=5Gi \
    --claim-mode=ReadWriteOnce \
    --mount-path=/var/lib/data

Make gp3-csi the default storage class¶

oc patch storageclass gp3-csi \
    -p '{"metadata":{"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

# Verify
oc get sc                              # one should have (default) suffix

Mount an existing PVC into a pod¶

spec:
  containers:
  - name: app
    volumeMounts:
    - name: data
      mountPath: /data
  volumes:
  - name: data
    persistentVolumeClaim:
      claimName: existing-pvc

Expand a PVC from 5 GB to 10 GB¶

oc edit pvc data
# change resources.requests.storage to 10Gi

# Or:
oc patch pvc data -p '{"spec":{"resources":{"requests":{"storage":"10Gi"}}}}'

oc rollout restart deploy/web         # so pod sees new size

Set up an NFS mount with a specific PV¶

Apply the static PV + PVC YAML pair above.

Troubleshooting storage¶

PVC stuck in Pending¶

oc describe pvc <name>

Look at Events. Common causes:

• No StorageClass with name <name> exists
• StorageClass exists but doesn't support requested access mode (e.g., asking for RWX from EBS)
• Cluster CSI driver in openshift-cluster-csi-drivers is degraded

Pod stuck in Pending because PVC unbound¶

oc describe pod <name> shows volume not yet bound. Fix the PVC.

Pod stuck in ContainerCreating due to mount failure¶

oc describe pod <name>

Common causes:

• SELinux context wrong on bind mount (use :Z if running locally)
• NFS server unreachable
• AccessMode not satisfied by underlying volume

Filesystem full despite PVC expanded¶

The PVC may show new capacity but the filesystem inside the pod hasn't been resized. Restart the pod:

oc rollout restart deploy/web

Verification¶

After storage changes:

• oc get pvc shows Bound status with expected size
• oc get pv shows the underlying PV
• oc rsh <pod> then df -h /mountpoint shows expected size
• oc rsh <pod> then touch /mountpoint/test creates a file (write works)
• After pod restart (oc rollout restart), the file persists (data survives)