Back to Kubernetes Course

The Problem: Containers Are Ephemeral

By default, any data written inside a container's filesystem is lost when the container restarts — and a container restarts far more often than you might expect (crashes, image updates, node maintenance). Kubernetes Volumes solve this by attaching storage to a Pod that outlives individual container restarts, and, with the right volume type, outlives the Pod itself.

---

Volumes (Pod-scoped)

The simplest volume types are tied to the Pod's lifetime — they survive container restarts within the Pod, but are deleted when the Pod itself is deleted.

emptyDir — a blank directory created when the Pod starts, useful for scratch space or sharing files between containers in the same Pod:

spec:
  containers:
    - name: app
      image: myapp:1.0
      volumeMounts:
        - name: cache-volume
          mountPath: /cache
  volumes:
    - name: cache-volume
      emptyDir: {}

hostPath — mounts a file or directory from the node's own filesystem into the Pod:

volumes:
  - name: host-logs
    hostPath:
      path: /var/log
      type: Directory

hostPath ties a Pod to whatever happens to be on that specific node, and is generally discouraged for application data — it's mainly used by infrastructure-level Pods (like log collectors or monitoring agents) that need direct node access, and it carries real security risk (covered later) if misused.

---

PersistentVolumes and PersistentVolumeClaims

For data that must outlive the Pod entirely — a database's data directory, user uploads — Kubernetes separates storage into two objects, so application developers don't need to know the underlying storage details:

• PersistentVolume (PV): A piece of actual storage in the cluster (an AWS EBS volume, an NFS share, a local disk), provisioned either manually by a cluster admin or automatically via a StorageClass. This is a cluster-level resource, not tied to any namespace.
• PersistentVolumeClaim (PVC): A request for storage made by a developer/application — "I need 10Gi of storage, ReadWriteOnce." Kubernetes binds the PVC to a PV that satisfies the request.

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: db-data-claim
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Gi
  storageClassName: standard

apiVersion: v1
kind: Pod
metadata:
  name: postgres-pod
spec:
  containers:
    - name: postgres
      image: postgres:16
      volumeMounts:
        - name: db-storage
          mountPath: /var/lib/postgresql/data
  volumes:
    - name: db-storage
      persistentVolumeClaim:
        claimName: db-data-claim

---

Access Modes

| Mode | Meaning | |---|---| | ReadWriteOnce (RWO) | Mounted as read-write by a single node at a time | | ReadOnlyMany (ROX) | Mounted as read-only by many nodes simultaneously | | ReadWriteMany (RWX) | Mounted as read-write by many nodes simultaneously | | ReadWriteOncePod | Mounted as read-write by a single Pod in the whole cluster (newer, stricter than RWO) |

Most cloud block storage (AWS EBS, Azure Disk, GCP PD) only supports ReadWriteOnce. ReadWriteMany typically requires network file storage like NFS, EFS, or Azure Files.

---

StorageClasses and Dynamic Provisioning

Manually pre-creating PersistentVolumes doesn't scale. A StorageClass lets PVs be provisioned automatically the moment a PVC requests one:

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: fast-ssd
provisioner: ebs.csi.aws.com
parameters:
  type: gp3
reclaimPolicy: Delete
volumeBindingMode: WaitForFirstConsumer

• provisioner identifies the plugin (a CSI — Container Storage Interface driver) responsible for actually creating the underlying disk with the cloud provider.
• reclaimPolicy: Delete means the underlying storage is deleted when the PVC is deleted; Retain keeps the data around for manual recovery — an important choice for production databases.
• volumeBindingMode: WaitForFirstConsumer delays provisioning until a Pod using the PVC is actually scheduled, so the storage is created in the same availability zone as the Pod.
• Most managed Kubernetes clusters (EKS, GKE, AKS) ship with a sensible default StorageClass already installed.

kubectl get storageclass
kubectl get pv
kubectl get pvc

---

Why This Matters for StatefulSets

Databases and other stateful applications rely heavily on PersistentVolumeClaims — each replica of a database needs its own persistent, stable storage that survives Pod rescheduling. We'll see this pattern in detail in the StatefulSets lesson, where each replica automatically gets its own dedicated PVC.