Troubleshoot Managed Lustre volumes

This page includes troubleshooting steps for common issues and errors when using the Google Cloud Managed Lustre CSI driver on Google Kubernetes Engine.

Before following the troubleshooting steps in this section, refer to the limitations when connecting to Managed Lustre from GKE.

Updated minimum instance capacity

The minimum capacity for Managed Lustre instances has been updated to 9000 GiB. To create 9000 GiB instances using the Managed Lustre CSI driver, upgrade your cluster version to 1.34.0-gke.2285000 or later.

Incorrect performance tier for dynamically provisioned Lustre instances

When dynamically provisioning a Lustre instance, the instance creation fails with an InvalidArgument error for PerUnitStorageThroughput, regardless of the perUnitStorageThroughput value specified in the API request. This affects GKE 1.33 versions before 1.33.4-gke.1036000.

Workaround:

Upgrade the GKE cluster to version 1.33.4-gke.1036000 or later. If using the Stable channel, a newer version might not be available yet. In this case, you can manually select a version from the Regular or Rapid channels that includes the fix.

Managed Lustre communication ports

The Managed Lustre CSI driver uses different ports for communication with Managed Lustre instances, depending on your GKE cluster version and existing Managed Lustre configurations.

  • Default port (988): For new GKE clusters that run version 1.33.2-gke.4780000 or later, the driver uses port 988 for Lustre communication by default.

  • Legacy port (6988): The driver uses port 6988 in the following scenarios:

    • Earlier GKE versions: If your GKE cluster runs a version earlier than 1.33.2-gke.4780000, the --enable-legacy-lustre-port flag is required when enabling the CSI driver. Enabling this flag works around a port conflict with the gke-metadata-server on GKE nodes.
    • Existing Managed Lustre instances with GKE support: If you are connecting to an existing Managed Lustre instance that was created with the --gke-support-enabled flag, you must include --enable-legacy-lustre-port when enabling the CSI driver, irrespective of your cluster version. Without this flag, your GKE cluster will fail to mount the existing Lustre instance.

    For more information on enabling the CSI driver with the legacy port, see Lustre communication ports.

Log queries

To check logs, run the following query in Logs Explorer.

To return Managed Lustre CSI driver node server logs:

resource.type="k8s_container"
resource.labels.pod_name=~"lustre-csi-node*"

Troubleshoot volume provisioning

If the PersistentVolumeClaim (PVC) remains in a Pending state and no PersistentVolume (PV) is created after 20-30 minutes, an error might have occurred.

  1. Check the PVC events:

    kubectl describe pvc PVC_NAME

  2. If the error indicates configuration issues or invalid arguments, verify your StorageClass parameters.

  3. Recreate the PVC.

  4. If the issue persists, contact Cloud Customer Care.

Troubleshoot volume mounting

After the Pod is scheduled to a node, the volume is mounted. If this fails, check the Pod events and kubelet logs.

kubectl describe pod POD_NAME

CSI driver enablement issues

Symptom:

MountVolume.MountDevice failed for volume "yyy" : kubernetes.io/csi: attacher.MountDevice failed to create newCsiDriverClient: driver name lustre.csi.storage.gke.io not found in the list of registered CSI drivers

or

MountVolume.SetUp failed for volume "yyy" : kubernetes.io/csi: mounter.SetUpAt failed to get CSI client: driver name lustre.csi.storage.gke.io not found in the list of registered CSI drivers

Cause: The CSI driver is not enabled or not yet running.

Resolution:

  1. Verify the CSI driver is enabled.
  2. If the cluster was recently scaled or upgraded, wait a few minutes for the driver to become functional.
  3. If the error persists, check the lustre-csi-node logs for "Operation not permitted". This indicates that the node version is too old to support Managed Lustre. To resolve this, upgrade your node pool to version 1.33.2-gke.1111000 or later.
  4. If the logs show "LNET_PORT mismatch", upgrade your node pool to ensure compatible Lustre kernel modules are installed.

Mountpoint already exists

Symptom:

MountVolume.MountDevice failed for volume "yyy" : rpc error: code = AlreadyExists
desc = A mountpoint with the same lustre filesystem name "yyy" already exists on
node "yyy". Please mount different lustre filesystems

Cause: Mounting multiple volumes from different Managed Lustre instances with the same file system name on a single node is not supported.

Resolution: Use a unique file system name for each Managed Lustre instance.

Mount failed: No such file or directory

Symptom:

MountVolume.MountDevice failed for volume "yyy" : rpc error: code = Internal desc = Could not mount ... failed: No such file or directory

Cause: The file system name specified is incorrect or does not exist.

Resolution: Verify the fs_name in your StorageClass or PV configuration matches the Managed Lustre instance.

Mount failed: Input/output error

Symptom:

MountVolume.MountDevice failed for volume "yyy" : rpc error: code = Internal desc = Could not mount ... failed: Input/output error

Cause: The cluster cannot connect to the Managed Lustre instance.

Resolution:

  1. Verify the IP address of the Managed Lustre instance.
  2. Ensure the GKE cluster and Managed Lustre instance are in the same VPC network or are correctly peered.

Internal errors

Symptom: rpc error: code = Internal desc = ...

Resolution: If the error persists, contact Cloud Customer Care.

Troubleshoot volume unmounting

Symptom:

UnmountVolume.TearDown failed for volume "yyy" : rpc error: code = Internal desc = ...

Resolution:

  1. Force delete the Pod:

    kubectl delete pod POD_NAME --force

  2. If the issue persists, contact Cloud Customer Care.

Troubleshoot volume deletion

If the PV remains in a "Released" state for an extended period (for example, more than an hour) after deleting the PVC, contact Cloud Customer Care.

Troubleshoot volume expansion

PVC stuck in ExternalExpanding

Symptom: The PVC status does not change to Resizing, and events show ExternalExpanding.

Cause: The allowVolumeExpansion field might be missing or set to false.

Resolution: Ensure the StorageClass has allowVolumeExpansion: true.

kubectl get storageclass STORAGE_CLASS_NAME -o yaml

Expansion fails: Invalid argument

Symptom: VolumeResizeFailed: rpc error: code = InvalidArgument ...

Cause: The requested size is invalid (for example, not a multiple of the step size or outside limits).

Resolution: Check the valid capacity ranges and update the PVC with a valid size.

Expansion fails: Internal error

Symptom: VolumeResizeFailed ... rpc error: code = Internal

Resolution: Retry the expansion by reapplying the PVC. If it fails repeatedly, contact Cloud Customer Care.

Deadline exceeded

Symptom: VolumeResizeFailed with DEADLINE_EXCEEDED.

Cause: The operation is taking longer than expected but may still be in progress.

Resolution: Wait for the operation to complete. The resizer will retry automatically. If it remains stuck for a long time (for example, > 90 minutes), contact support.

Quota exceeded

Symptom: Expansion fails due to quota limits.

Resolution: Request a quota increase or request a smaller capacity increase.