如何使用 PersistentVolumeClaim 在 Deployment/Pod 上挂载持久卷？

Question

I am trying to mount a persistent volume on pods (via a deployment).

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
      - image: ...
        volumeMounts:
        - mountPath: /app/folder
          name: volume
      volumes:
      - name: volume
        persistentVolumeClaim:
          claimName: volume-claim
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: volume-claim
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 2Gi

However, the pod stays in "ContainerCreating" status and the events show the following error message.

Unable to mount volumes for pod "podname": timeout expired waiting for volumes to attach or mount for pod "namespace"/"podname". list of unmounted volumes=[volume]. list of unattached volumes=[volume]

I verified that the persistent volume claim is ok and bound to a persistent volume.

What am I missing here?

Answer 1

If doing this in a cloud provider, the storageClass object will create the respective volume for your persistent volume claim.

If you are trying to do this locally on minikube or in a self managed kubernetes cluster, you need to manually create the storageClass that will provide the volumes for you, or create it manually like this example:

apiVersion: v1
kind: PersistentVolume
metadata:
  name: task-pv-volume
  labels:
    type: local
spec:
  storageClassName: manual
  capacity:
    storage: 2Gi
  accessModes:
    - ReadWriteOnce
  hostPath:
    path: "/mnt/data"

The hostPath variable will mount this data in the current pod node.

Answer 2

When you create a PVC without specifying a PV or type of StorageClass in GKE clusters it will fall back to default option:

• StorageClass: standard
• Provisioner: kubernetes.io/gce-pd
• Type: pd-standard

Please take a look on official documentation: Cloud.google.com: Kubernetes engine persistent volumes

There could be a lot of circumstances that can produce error message encountered.

As it's unknown how many replicas are in your deployment as well as number of nodes and how pods were scheduled on those nodes, I've tried to reproduce your issue and I encountered the same error with following steps (GKE cluster was freshly created to prevent any other dependencies that might affect the behavior).

Steps :

• Create a PVC
• Create a Deployment with replicas > 1
• Check the state of pods
• Additional links

Create a PVC

Below is example YAML definition of a PVC the same as yours:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: volume-claim
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 2Gi

After applying above definition please check if it created successfully. You can do it by using below commands:

• $ kubectl get pvc volume-claim
• $ kubectl get pv
• $ kubectl describe pvc volume-claim
• $ kubectl get pvc volume-claim -o yaml

Create a Deployment with replicas > 1

Below is example YAML definition of deployment with volumeMounts and replicas > 1:

apiVersion: apps/v1 
kind: Deployment
metadata:
  name: ubuntu-deployment
spec:
  selector:
    matchLabels:
      app: ubuntu
  replicas: 10 # amount of pods must be > 1
  template:
    metadata:
      labels:
        app: ubuntu
    spec:
      containers:
      - name: ubuntu
        image: ubuntu
        command:
        - sleep
        - "infinity"
        volumeMounts:
        - mountPath: /app/folder
          name: volume
      volumes:
      - name: volume
        persistentVolumeClaim:
          claimName: volume-claim

Apply it and wait for a while.

Check the state of pods

You can check the state of pods with below command:

$ kubectl get pods -o wide

Output of above command:

NAME                      READY   STATUS              RESTARTS   AGE     IP            NODE                              
ubuntu-deployment-2q64z   0/1     ContainerCreating   0          4m27s   <none>        gke-node-1  
ubuntu-deployment-4tjp2   1/1     Running             0          4m27s   10.56.1.14    gke-node-2   
ubuntu-deployment-5tn8x   0/1     ContainerCreating   0          4m27s   <none>        gke-node-1   
ubuntu-deployment-5tn9m   0/1     ContainerCreating   0          4m27s   <none>        gke-node-3  
ubuntu-deployment-6vkwf   0/1     ContainerCreating   0          4m27s   <none>        gke-node-1  
ubuntu-deployment-9p45q   1/1     Running             0          4m27s   10.56.1.12    gke-node-2  
ubuntu-deployment-lfh7g   0/1     ContainerCreating   0          4m27s   <none>        gke-node-3  
ubuntu-deployment-qxwmq   1/1     Running             0          4m27s   10.56.1.13    gke-node-2 
ubuntu-deployment-r7k2k   0/1     ContainerCreating   0          4m27s   <none>        gke-node-3   
ubuntu-deployment-rnr72   0/1     ContainerCreating   0          4m27s   <none>        gke-node-3

Take a look on above output:

• 3 pods are in Running state
• 7 pods are in ContainerCreating state

All of the Running pods are located on the same gke-node-2

You can get more detailed information why pods are in ContainerCreating state by:

$ kubectl describe pod NAME_OF_POD_WITH_CC_STATE

The Events part in above command shows:

Events:
  Type     Reason              Age                From                                             Message
  ----     ------              ----               ----                                             -------
  Normal   Scheduled           14m                default-scheduler                                Successfully assigned default/ubuntu-deployment-2q64z to gke-node-1
  Warning  FailedAttachVolume  14m                attachdetach-controller                          Multi-Attach error for volume "pvc-7d756147-6434-11ea-a666-42010a9c0058" Volume is already used by pod(s) ubuntu-deployment-qxwmq, ubuntu-deployment-9p45q, ubuntu-deployment-4tjp2
  Warning  FailedMount         92s (x6 over 12m)  kubelet, gke-node-1  Unable to mount volumes for pod "ubuntu-deployment-2q64z_default(9dc28e95-6434-11ea-a666-42010a9c0058)": timeout expired waiting for volumes to attach or mount for pod "default"/"ubuntu-deployment-2q64z". list of unmounted volumes=[volume]. list of unattached volumes=[volume default-token-dnvnj]

Pod cannot pass ContainerCreating state because of failed mounting of a volume . Mentioned volume is already used by other pods on a different node.

ReadWriteOnce: The Volume can be mounted as read-write by a single node.

Additional links

Please take a look at: Cloud.google.com: Access modes of persistent volumes .

There is detailed answer on topic of access mode: Stackoverflow.com: Why can you set multiple accessmodes on a persistent volume

As it's unknown what you are trying to achieve please take a look on comparison between Deployments and Statefulsets: Cloud.google.com: Persistent Volume: Deployments vs statefulsets .