NFS issue: Irix server, Linux client - inode number mismatch
We have an SGI Irix 6.5-based server running NFS3 with Linux 2.4.7 clients attaching. Whenever the server crashes, the Linux clients have problems with their NFS mounts. Typically, they're unusable until after reboot.
The /var/log/messages file on the Linux client side has the following messages:
Oct 8 16:01:27 lrender2 automount[22485]: expired
/usr/local/sgi/wbfa
Oct 8 16:01:27 lrender2 kernel: nfs_refresh_inode: inode number
mismatch
Oct 8 16:01:27 lrender2 kernel: expected (0x3000007/0x6467c36),
got (0x3000005/0x6467c36)
Oct 8 16:06:27 lrender2 automount[22486]: expired
/usr/local/sgi/wbfa
Oct 8 16:06:27 lrender2 kernel: nfs_refresh_inode: inode number
mismatch
Oct 8 16:06:27 lrender2 kernel: expected (0x3000007/0x6467c36),
got (0x3000005/0x6467c36)
Oct 8 16:11:27 lrender2 automount[22487]: expired
/usr/local/sgi/wbfa
Oct 8 16:11:27 lrender2 kernel: nfs_refresh_inode: inode number
mismatch
Oct 8 16:11:27 lrender2 kernel: expected (0x3000007/0x6467c36),got (0x3000005/0x6467c36)
Can anyone tell me if this is expected behavior, and if not, if the problem is in the Irix NFS server implementation or the Linux client implementation? I don't have the NFS expertise to be able to discern...
BTW, other clients (SGI Irix and Mac OS X workstations) connected to the same mount points do NOT exhibit this
behaviour after a server crash.
I checked the changelog at
http://www.fys.uio.no/~trondmy/src/ChangeLog.NFSv3 but didn't see anything relevant.
Thanks,
Alan Hagge
Replies via cc:, please...
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
" " == Alan Hagge <[email protected]> writes:
> /usr/local/sgi/wbfa
> Oct 8 16:01:27 lrender2 kernel: nfs_refresh_inode:
> inode number
> mismatch
> Oct 8 16:01:27 lrender2 kernel: expected
> (0x3000007/0x6467c36),
> got (0x3000005/0x6467c36)
That means that the device number or 'fsid' suddenly has changed on your server. That's a server bug.
Cheers,
Trond