Use persistent disks with multiple readers

Autopilot Standard

---

This page explains how to add a persistent disk to your cluster using the ReadOnlyMany access mode. This mode allows multiple Pods on different nodes to mount the disk for reading.

For more information on this mode, refer to persistent volume access modes.

Requirements

• Your cluster must run GKE version 1.22 or later.
• Your cluster must use the Compute Engine persistent disk CSI driver.

Format and populate a source persistent disk

To use a persistent disk in read-only mode, you must populate a source persistent disk with data, then either clone the volume or use a volume snapshot to move the data into a new ReadOnlyMany PersistentVolume.

1. Create a PersistentVolume using an existing persistent disk or using dynamic provisioning.

2. Format the disk and populate it with data. To format the disk, reference the disk as a ReadWriteOnce PersistentVolume in a Pod. GKE automatically formats the underlying disk, and lets the Pod write data to the disk. When the Pod starts, make sure the Pod writes the data you want to the disk.

Create a ReadOnlyMany PersistentVolume

You can create a ReadOnlyMany PersistentVolume using one of the following methods:

• A volume snapshot of a source PersistentVolume that you populate with data.
• A volume clone of a source PersistentVolume that you populate with data.

• A pre-existing persistent disk that was already populated with data.

Volume snapshot

1. Create a volume snapshot of the source PersistentVolume.

2. Save the following PersistentVolumeClaim manifest as snapshot-pvc.yaml:

   kind: PersistentVolumeClaim
   apiVersion: v1
   metadata:
     namespace: PVC_NAMESPACE
     name: PVC_NAME
   spec:
     dataSource:
       apiGroup: snapshot.storage.k8s.io
       kind: VolumeSnapshot
       name: SNAPSHOT_NAME
     accessModes:
       - ReadOnlyMany
     storageClassName: premium-rwo
     resources:
       requests:
         storage: STORAGE_SIZE
   

   Replace the following:

   • PVC_NAMESPACE: the namespace of the new PersistentVolumeClaim.
   • PVC_NAME: the name of the new PersistentVolumeClaim.
   • SNAPSHOT_NAME:the name of your VolumeSnapshot object.
   • STORAGE_SIZE: the amount of storage to request. This must be the same as the amount requested in the source PersistentVolumeClaim.

3. Apply the manifest to your cluster:

   kubectl apply -f snapshot-pvc.yaml
   

This creates a new PersistentVolumeClaim named PVC_NAME in your cluster, which GKE uses to dynamically provision a new PersistentVolume that contains the data from the snapshot of the source persistent disk.

Volume cloning

Before using volume cloning, make sure to familiarize yourself with the limitations of this approach.

1. Save the following manifest as cloning-pvc.yaml:

   kind: PersistentVolumeClaim
   apiVersion: v1
   metadata:
     namespace: PVC_NAMESPACE
     name: PVC_NAME
   spec:
     dataSource:
       name: SOURCE_PVC
       kind: PersistentVolumeClaim
     accessModes:
       - ReadOnlyMany
     storageClassName: STORAGE_CLASS_NAME
     resources:
       requests:
         storage: STORAGE_SIZE
   

   Replace the following:

   • PVC_NAMESPACE: the namespace of the new PersistentVolumeClaim.
   • PVC_NAME: the name of the new PersistentVolumeClaim.
   • SOURCE_PVC: the name of the source PersistentVolumeClaim populated with data.
   • STORAGE_CLASS_NAME: the storage class for the new PersistentVolumeClaim. This must be the same as the storage class of the source PersistentVolumeClaim.
   • STORAGE_SIZE: the amount of storage for the new PersistentVolumeClaim. This must be the same amount requested by the source PersistentVolumeClaim.

2. Apply the manifest to your cluster:

   kubectl apply -f cloning-pvc.yaml
   

This creates a PersistentVolumeClaim named PVC_NAME that GKE uses to create a new PersistentVolume in ReadOnlyMany mode with the data in the source PersistentVolume.

For more detailed information about Volume Cloning, see Create clones of persistent volumes.

Pre-existing persistent disk

1. Save the following manifest as preexisting-disk-pv-pvc.yaml:

   apiVersion: v1
   kind: PersistentVolume
   metadata:
     name: PV_NAME
   spec:
     storageClassName: "STORAGE_CLASS_NAME"
     capacity:
       storage: DISK_SIZE
     accessModes:
       - ReadOnlyMany
     claimRef:
       namespace: PVC_NAMESPACE
       name: PVC_NAME
     csi:
       driver: pd.csi.storage.gke.io
       volumeHandle: DISK_ID
       fsType: FS_TYPE
       readOnly: true
   ---
   apiVersion: v1
   kind: PersistentVolumeClaim
   metadata:
     namespace: PVC_NAMESPACE
     name: PVC_NAME
   spec:
     storageClassName: "STORAGE_CLASS_NAME"
     volumeName: PV_NAME
     accessModes:
       - ReadOnlyMany
     resources:
       requests:
         storage: DISK_SIZE
   

   Replace the following:

   • PV_NAME: the name of your new PersistentVolume.
   • STORAGE_CLASS_NAME: the name of your new StorageClass.
   • DISK_SIZE: the size of your pre-existing persistent disk. For example, 500G.
   • PVC_NAMESPACE: the namespace of the new PersistentVolumeClaim.
   • PVC_NAME: the name of your new PersistentVolumeClaim.
   • DISK_ID: the identifier of your pre-existing persistent disk. The format is projects/{project_id}/zones/{zone_name}/disks/{disk_name} for Zonal persistent disks, or projects/{project_id}/regions/{region_name}/disks/{disk_name} for Regional persistent disks.
   • FS_TYPE: the filesystem type. You can leave this as the default (ext4), or use xfs. If your clusters use a Windows Server node pool, you must change this to NTFS.

2. Apply the manifest to your cluster:

   kubectl apply -f preexisting-disk-pv-pvc.yaml
   

This creates a PersistentVolumeClaim named PVC_NAME and a PersistentVolume named PV_NAME in ReadOnlyMany mode.

For more detailed information about using a pre-existing persistent disk, see existing persistent disk.

Use the PersistentVolumeClaim in a Pod

You can now reference the new PersistentVolumeClaim in read-only mode in multiple Pods on multiple nodes at the same time. You cannot attach persistent disks in ReadWriteOnce mode to multiple nodes at the same time. For more information, refer to Deployments vs. StatefulSets.

In your Pod specification, you must specify readOnly: true in the volumeMounts section and the volumes section, such as in the following example:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: multi-read
  labels:
    app: web-server
spec:
  replicas: 3
  selector:
    matchLabels:
      app: web-server
  template:
    metadata:
      labels:
        app: web-server
    spec:
      containers:
      - name: web-server
        image: nginx
        volumeMounts:
        - mountPath: /test-mnt
          name: my-volume
          readOnly: true
      volumes:
      - name: my-volume
        persistentVolumeClaim:
          claimName: PVC_NAME
          readOnly: true

Note: Read-only ext3 and ext4 filesystems do not support journal playback. If the source persistent disk filesystem is marked as dirty when you clone or snapshot it, the data in your read-only persistent disk might be inconsistent or corrupt. The Compute Engine persistent disk CSI driver attempts to mount the disk normally. If the inability to playback the journal causes the volume mount to fail, the CSI driver automatically retries the mount with the noload option, which disables playback attempts on ext3 and ext4 filesystems. If the source filesystem was actually inconsistent at the moment it was cloned or snapshotted, the read-only mount will also be inconsistent.

What's next

• Learn more about using pre-existing persistent disks as PersistentVolumes.
• Learn more about Compute Engine persistent disks.