getfh failed

Problem:

I had a broken dhcp server. While testing this dhcp server I would repeatedly change the IP reservation for my workstation (which runs RHEL 3ES). Upon concluding this testing, I changed my IP reservation back to the original IP address and rebooted my system. However upon this reboot, my system could no longer mount nfs directories from our two primary fileservers (all running Linux RHEL 3ES).

Note that my workstation had no problem mounting these filesystems when it had a brand-new lease with a different IP address; the failure only occured when the workstation was reverted back to its original IP address.

The error on the server was:

Apr 11 14:06:11 fs0 rpc.mountd: getfh failed: Operation not permitted

Solution:

On the server, clear up stale client mount information with the command

fs0# exportfs -av

Discussion:

When the workstation was originally shut down, it left some stale information on the server's /var/lib/nfs tree, which cannot be trivially edited. This stale information conflicted with the workstation's new request for a mount. To clear this information you must either cycle nfs on the server (BAD, because that makes the server forgets about all client mounts) or run the exportfs command, above.

(All gleaned from a Debian discussion list somewhere)