Skip to content

4. Storage (10%)

Level: Intermediate. These are small, quick tasks. Get them right and bank the points.

4.1 The model

flowchart LR
    Pod -- volumes.persistentVolumeClaim --> PVC[PersistentVolumeClaim<br/>namespaced: what I need]
    PVC -- bound 1:1 --> PV[PersistentVolume<br/>cluster-scoped: actual storage]
    SC[StorageClass<br/>how to provision] -. dynamic provisioning .-> PV
    PVC -. storageClassName .-> SC
  • Static provisioning: An admin creates the PV, and a PVC binds to a matching one (capacity ≥ request, same access mode, same storageClassName).
  • Dynamic provisioning: A PVC names a StorageClass, and the provisioner (CSI driver) creates the PV automatically.

4.2 StorageClass

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: fast
  annotations:
    storageclass.kubernetes.io/is-default-class: "true"   # make it the default
provisioner: rancher.io/local-path          # check what exists: k get sc
reclaimPolicy: Delete                       # Delete (default) | Retain
volumeBindingMode: WaitForFirstConsumer     # bind when a pod uses it (topology-aware)
allowVolumeExpansion: true
parameters: {}                              # provisioner-specific

k get sc                                     # (default) marker shows the default class
k patch sc fast -p '{"metadata":{"annotations":{"storageclass.kubernetes.io/is-default-class":"false"}}}'

Only one StorageClass should be the default. When changing it, remove the annotation from the old one.

4.3 PersistentVolume (static)

apiVersion: v1
kind: PersistentVolume
metadata: {name: pv-data}
spec:
  capacity: {storage: 1Gi}
  accessModes: [ReadWriteOnce]
  persistentVolumeReclaimPolicy: Retain
  storageClassName: manual            # must match the PVC ("" for no class)
  hostPath:
    path: /mnt/data
    type: DirectoryOrCreate
Access mode Short Meaning
ReadWriteOnce RWO Read-write by one node (multiple pods on that node can share it)
ReadOnlyMany ROX Read-only by many nodes
ReadWriteMany RWX Read-write by many nodes (NFS, CephFS, cloud file stores)
ReadWriteOncePod RWOP Read-write by one pod in the whole cluster
Reclaim policy After the PVC is deleted
Retain PV goes to Released. Data is kept, and an admin must clean up and reuse it manually (remove spec.claimRef to make it Available again)
Delete PV and the backing storage are deleted (default for dynamic provisioning)
Recycle Deprecated. Don't use

4.4 PersistentVolumeClaim and using it in a pod

apiVersion: v1
kind: PersistentVolumeClaim
metadata: {name: data, namespace: app}
spec:
  accessModes: [ReadWriteOnce]
  resources:
    requests: {storage: 500Mi}
  storageClassName: manual       # omit to use the default StorageClass
---
apiVersion: v1
kind: Pod
metadata: {name: writer, namespace: app}
spec:
  containers:
  - name: app
    image: busybox:1.36
    command: ["sh", "-c", "echo hello > /data/out.txt; sleep 3600"]
    volumeMounts:
    - {name: data, mountPath: /data}
  volumes:
  - name: data
    persistentVolumeClaim:
      claimName: data

k get pv,pvc -A
k describe pvc data -n app        # Events explain why it's Pending

PVC stuck Pending?

  • No PV matches: check capacity, access mode, and storageClassName.
  • WaitForFirstConsumer: this is normal until a pod uses the claim.
  • No default StorageClass, and the PVC doesn't name one.
  • The provisioner isn't running.

4.5 Expanding a PVC

# StorageClass must have allowVolumeExpansion: true
k patch pvc data -n app -p '{"spec":{"resources":{"requests":{"storage":"2Gi"}}}}'
k get pvc data -n app -w

You can only increase the size. Some drivers need the pod to restart for the filesystem resize.

4.6 Other volume types to recognize

Type Use
emptyDir Scratch space shared by the containers in a pod. Deleted with the pod. medium: Memory for tmpfs
hostPath Node directory. Used for static pods and node agents. Avoid it for apps
configMap / secret / projected Config and credentials as files
persistentVolumeClaim Durable storage
CSI (csi:) Provided by a CSI driver (see Page 5)

StatefulSets create one PVC per replica from volumeClaimTemplates, named <template>-<sts>-<ordinal>. Those PVCs are not deleted when you scale down.

4.7 Checklist

  • Create a StorageClass and make it the default (and un-default the old one)
  • Create a hostPath PV with Retain, plus a matching PVC, and mount it in a pod
  • Explain RWO vs RWOP vs RWX, and Retain vs Delete
  • Diagnose a Pending PVC from its events
  • Expand a PVC

Next: 5. Cluster Architecture, Installation & Configuration →