Container Compute Service:Mount ephemeral volumes

Why use ephemeral volumes

Generic ephemeral volumes work like emptyDir volumes — they're scoped to a single pod's lifetime and deleted when the pod is deleted. The key difference is that they're backed by a PVC, which gives you the full capabilities of a persistent storage driver:

• Fixed capacity: Set an exact size limit per pod so the application cannot exceed its storage budget.

• Network-attached storage: Back the volume with a cloud disk instead of node-local storage, which is often limited in size and shared across workloads.

• StorageClass support: Configure performance tier, access mode, and other disk properties.

Use ephemeral volumes when your application needs temporary storage that is too large for node-local storage, or when you need a guaranteed capacity per pod. Common scenarios include:

• Caching services: Applications that move infrequently accessed data from memory to disk benefit from an isolated, capacity-bounded volume per pod.

• Log storage: Applications that generate large volumes of log data can write to a dedicated ephemeral volume instead of the shared node filesystem.

• Scratch data: Batch-processing jobs that need high-throughput disk I/O without persisting data after the job completes.

How it works

When Kubernetes creates a pod with a volumeClaimTemplate in its ephemeral volume spec, the ephemeral volume controller creates a PVC in the same namespace as the pod. The PVC name follows a deterministic pattern: {pod-name}-{volume-name}. This predictable naming lets you look up the PVC directly without searching, as long as you know the pod name and volume name.

When the pod is deleted, the Kubernetes garbage collector deletes the PVC based on the resource ownership relationship. Because the default reclaim policy for StorageClasses is Delete, the underlying disk volume is deleted along with the PVC.

PVC naming follows the pattern {pod-name}-{volume-name}. If two pods in the same namespace have names that produce the same combined string (for example, a pod named pod-a with a volume named scratch and a pod named pod with a volume named a-scratch), a naming conflict will occur. Use distinct pod names within a namespace to avoid this.

Prerequisites

Before you begin, ensure that you have:

• An Alibaba Cloud Container Compute Service (ACS) cluster

• kubectl configured to connect to the cluster — see Obtain the kubeconfig file of a cluster and use kubectl to connect to the cluster and Use kubectl on Cloud Shell to manage ACS clusters

Mount an ephemeral volume to a Deployment

1. Create a file named ephemeral-example.yaml with the following content:

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: ephemeral-example
   spec:
     replicas: 2
     selector:
       matchLabels:
         pod: example-pod
     strategy:
       type: Recreate
     template:
       metadata:
         labels:
           pod: example-pod
       spec:
         containers:
           - name: nginx
             image: registry.cn-hangzhou.aliyuncs.com/acs-sample/nginx:latest
             resources:
               requests:
                 cpu: 500m
                 memory: 2Gi
                 ephemeral-storage: 2Gi
             volumeMounts:
             - mountPath: "/scratch"
               name: scratch-volume
         volumes:
           - name: scratch-volume
             ephemeral:
               volumeClaimTemplate:
                 metadata:
                   labels:
                     type: scratch-volume
                 spec:
                   accessModes: [ "ReadWriteOncePod" ]
                   storageClassName: alicloud-disk-topology-alltype
                   resources:
                     requests:
                       storage: 40Gi

   This Deployment creates two pods, each with a dedicated 40 GiB ephemeral volume mounted at /scratch. The ephemeral.volumeClaimTemplate field is supported on Deployments, StatefulSets, and standalone pods. The following table describes the fields inside volumeClaimTemplate.spec:

   Field	Description
   accessModes	Set to ReadWriteOncePod so the volume is mounted by exactly one pod at a time.
   storageClassName	The StorageClass to use. This example uses alicloud-disk-topology-alltype, the default ACS StorageClass, which provisions a performance level 1 (PL1) Enterprise SSD (ESSD). A PL1 ESSD supports 20 to 65,536 GiB. Disks are billed on a pay-as-you-go basis — see Billing of block storage devices and Prices of block storage devices.
   storage	The requested capacity of the ephemeral volume.

2. Deploy the Deployment:

   kubectl create -f ephemeral-example.yaml

3. Check that both pods are running:

   kubectl get pod

   Expected output:

   NAME                                 READY   STATUS    RESTARTS   AGE
   ephemeral-example-64db7b9f5c-hcpz5   1/1     Running   0          108s
   ephemeral-example-64db7b9f5c-qqvl5   1/1     Running   0          108s

4. Check that a PVC was created for each pod:

   kubectl get pvc

   Expected output:

   NAME                                                STATUS   VOLUME                   CAPACITY   ACCESS MODES   STORAGECLASS                     AGE
   ephemeral-example-64db7b9f5c-hcpz5-scratch-volume   Bound    d-uf6h6nl0yf2y1331****   40Gi       RWOP           alicloud-disk-topology-alltype   3m28s
   ephemeral-example-64db7b9f5c-qqvl5-scratch-volume   Bound    d-uf6a3lk85znunou6****   40Gi       RWOP           alicloud-disk-topology-alltype   3m28s

   The PVC names follow the pattern {pod-name}-scratch-volume. The VOLUME field shows the disk ID of the underlying cloud disk.

Verify that ephemeral volumes are deleted with the pod

1. Scale the Deployment down to one replica:

   kubectl scale deploy ephemeral-example --replicas=1

2. Confirm that one pod was deleted:

   kubectl get pod

   Expected output:

   NAME                                 READY   STATUS    RESTARTS   AGE
   ephemeral-example-64db7b9f5c-qqvl5   1/1     Running   0          11m

3. Check that the PV for the deleted pod was removed:

   kubectl get pv

   Expected output (only the PV for the surviving pod remains):

   NAME                     CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                                                       STORAGECLASS                     REASON   AGE
   d-uf6a3lk85znunou6****   40Gi       RWOP           Delete           Bound    default/ephemeral-example-64db7b9f5c-qqvl5-scratch-volume   alicloud-disk-topology-alltype            20m

4. Check that the PVC for the deleted pod was also removed:

   kubectl get pvc

   Expected output (only the PVC for the surviving pod remains):

   NAME                                                STATUS   VOLUME                   CAPACITY   ACCESS MODES   STORAGECLASS                     AGE
   ephemeral-example-64db7b9f5c-qqvl5-scratch-volume   Bound    d-uf6a3lk85znunou6****   40Gi       RWOP           alicloud-disk-topology-alltype   15m

   The PVC and its underlying disk were deleted together with the pod. The ephemeral volume for the remaining pod is still bound.