NFS error scenarios

This topic provides information on how to verify and resolve NFS errors.

NFS client cannot access exported NFS data

Problem

The NFS client cannot access the exported data even though the export is mounted. This often results in errors to occur while writing data, creating files, or traversing the directory hierarchy (permission denied).

Determination

The error itself occurs on the NFS client side. Additionally, and based on the nature of the problem, the server-side NFS logs can provide more details about the origin of the error.

Solution

There are multiple reasons for this problem:

The ACL definition in the file system does not allow the requested operation.

The export and/or client definition of the export do not allow that operation (such as a read only definition).

  1. Verify the ACL definition of the export path in the file system. To check ACL definitions, run:
    mmgetacl Path
  2. Verify the definition of the export and the client (especially the access type). To check the NFS export details, run:
    mmnfs export list -n Path
  3. Unmount and remount the file system on the NFS client:
    umount <Path>
mount <mount_options> CES_IP_address:<export_path> <mount_point>

NFS client I/O temporarily stalled

Problem

The NFS client temporarily encounters stalled I/O or access requests to the export. The problem goes away after a short time (about 1 minute.)

Determination

The error itself occurs on the NFS client side, but due to an action on the NFS server side. The server-side NFS logs can provide more details about the origin of the error (such as a restart of the NFS server) along with the CES logs (such as manual move of a CES IP or a failover condition).

Origin

In grace period, the NFS server might temporarily suspend further access to the export from the NFS client (depending on the type of request). During the grace period, certain NFS operations are not allowed. The NFS server enters in the grace period due to any one of the following conditions:
  1. An explicit restart triggered manually through the CLI by running: mmces service stop / start ...
  2. An explicit move of CES IPs manually through the CLI by running: mmces address move ...
  3. A change in the definition of an existing export.
    Note: Adding or removing NFS exports does not initiate a restart.
  4. The creation of the first export.
  5. A critical error condition that triggers CES failover, which in turn causes IP addresses to move.
  6. A failback of CES IPs (depending on the setting of the address distribution policy).