GKE实例中无法查看已挂载Persistent Volume的问题求助

Let's break down what's going wrong here and fix it step by step. I see a few key issues in your StatefulSet configuration that are preventing your GCE disk from showing up when you run df -hT.

First: Your StatefulSet is using an emptyDir instead of your persistent volume #

Looking at your StatefulSet YAML, you've defined a volumes section with emptyDir: {} for my-volume. This creates a temporary, ephemeral directory that gets deleted when the pod restarts—this has nothing to do with your GCE disk or the PV/PVC you created. Worse, your volumeMounts are pointing to this emptyDir volume, so your persistent disk isn't being mounted at all.

Second: Your manual PV isn't being used by your PVC (or StatefulSet) #

When you created your PVC my-claim, it automatically bound to a dynamically provisioned PV (the ones starting with pvc-...) instead of your manual my-volume PV. That's because your manual PV doesn't specify a storageClassName matching the default one your cluster uses (likely standard), so the PVC opted for dynamic provisioning.

Plus, your StatefulSet is using volumeClaimTemplates which creates separate PVCs for each replica (like my-claim-xxx_0), but your volume mounts aren't referencing these correctly.

Fix Option 1: Use your manually created PV/PVC with the StatefulSet #

If you want to stick with the GCE disk you created manually, follow these steps:

1. Update your manual PV to match the default storage class
   Edit your PV to add a storageClassName so it can bind to your PVC:

   apiVersion: v1
   kind: PersistentVolume
   metadata:
     name: my-volume
   spec:
     capacity:
       storage: 200Gi
     accessModes:
       - ReadWriteOnce
     storageClassName: standard # Match your cluster's default storage class
     gcePersistentDisk:
       pdName: my-disk
       fsType: ext4
   

   Apply the update: kubectl apply -f your-pv.yaml

2. Delete and re-create your PVC to bind to the manual PV
   First delete the existing PVC that's bound to the dynamic PV:

   kubectl delete pvc my-claim
   

   Re-create your PVC, and it should now bind to my-volume instead of creating a new PV.

3. Fix your StatefulSet configuration
   Remove the emptyDir volume and volumeClaimTemplates, then reference your existing PVC directly:

   apiVersion: apps/v1beta2
   kind: StatefulSet
   metadata:
     name: your-statefulset-name # Fill in your actual name
   spec:
     serviceName: your-service-name # Required for StatefulSets
     replicas: 3 # Adjust to your desired number of replicas
     template:
       metadata:
         labels:
           app: your-app-label
       spec:
         containers:
           - name: your-container-name # Fill in your actual container name
             image: your-container-image # Fill in your actual image
             volumeMounts:
               - name: my-volume
                 mountPath: /mnt/disks
         volumes:
           - name: my-volume
             persistentVolumeClaim:
               claimName: my-claim # Reference your existing PVC
   

   Apply the updated StatefulSet: kubectl apply -f statefulset.yaml

Fix Option 2: Let the StatefulSet manage volumes automatically (Recommended) #

StatefulSets are designed to create unique PVCs for each replica automatically via volumeClaimTemplates. If you don't need the specific manual disk you created, this is the simpler approach:

1. Delete your manual PV and existing PVCs

   kubectl delete pv my-volume
   kubectl delete pvc my-claim my-claim-xxx_0 my-claim-xxx_1 # Delete all related PVCs
   

2. Fix your StatefulSet configuration
   Remove the emptyDir volume, and ensure your volumeMounts reference the name in volumeClaimTemplates:

   apiVersion: apps/v1beta2
   kind: StatefulSet
   metadata:
     name: your-statefulset-name
   spec:
     serviceName: your-service-name
     replicas: 3
     template:
       metadata:
         labels:
           app: your-app-label
       spec:
         containers:
           - name: your-container-name
             image: your-container-image
             volumeMounts:
               - name: my-claim # Matches the name in volumeClaimTemplates
                 mountPath: /mnt/disks
     volumeClaimTemplates:
       - metadata:
           name: my-claim
         spec:
           accessModes: [ "ReadWriteOnce" ]
           resources:
             requests:
               storage: 200Gi
           storageClassName: standard # Use your cluster's default storage class
   

   Apply the updated StatefulSet: kubectl apply -f statefulset.yaml

Verify the fix #

After applying the corrected StatefulSet, wait for the pods to restart. Then SSH into an instance and run df -hT—you should see your /mnt/disks mount listed with the ext4 filesystem type, pointing to your GCE disk.

内容的提问来源于stack exchange，提问作者user3908406