Server for NFS failed to initialize on drive letter

At times we see an event when the Server for NFS service starts up and the clients fail to mount the shares on this drive with error – invalid argument. This happens on a 2 node cluster environment.

In this specific environment, the drive in questions is L: and the WMI traces show some issues when dealing with the inode cache directories. The error string suggested that when the server is trying to read the header of the inode file (0x4a8 bytes long) the read request is only returning 0x488 bytes. That implies the inode file is too short or the file system had some problem reading the file. After investigation we found that this issue is on x64 bit.

 To begin with we checked the permission on the drives and on the ._nfs folders on C: and L: and they look exactly the same.

A simple workaround was given to the customer to change the registry settings to point to C:

Another solution is to rename the “L:\._nfs\DRIVEL.NOD” file to something like “L:\._nfs\DRIVEL.NOD.Original” and then restart NFS. At this point NFS should create a new DRIVEL.NOD file of a length suitable for a 64bit node.

The action item is as following -

1. Change the registry settings to point back to L:

2. Stop the NFS service on this node after moving the resources to the other node

3. Rename the DriveL.NOD file to DriveL.NOD.Original found in L:\._nfs folder

4. Restart the NFS service and verify a new file (DriveL.NOD) has been created in this folder