SUMMARY: Stale NFS file handle

Root mail receive (sunny@tet.uni-hannover.de)
Mon, 29 Sep 1997 08:24:24 +0200

Many thanks to all responses to my question:
> Can anyone please help me with this failure message that
> occures unregularily on my system
> and what to do against it?
> (on a sparc20 running Solaris2.5.1)
> asterix unix: NFS write error on host majestix: Stale NFS file handle.
> asterix unix: (file handle: 80000d 2 a0000 c6c52 b84a436b a0000 65400 62b8d638)

conclusion:

1. Explanations:
NFS stales occur when a (auto)mounted filesystem (or file) is either
no longer exported, or the machine where it resides is down or
again the filesystem does not exist any more.
Find out which filesystem it is and do a manual umount of it
on the machine where the stale is.
A good explanation about how to debug the problem can be found at
http://www.Sun.COM:80/sunworldonline/swol-10-1995/swol-10-sysadmin.html

2. Work to do:
If the mount is busy, do a fuser -u <mounted-filesystem>
and either exit or kill all the printed processes (use -k to do
it automatically).
Use the the script fhfind from the above location to find out
the i-node.

3. What happened:
The reason for my problems were, that the configuration of our samba servers were such, that both servers on two different machines used the same directory for there lock files with the same lock file names. In this configuration the one server could open the file while the second server (on the exporting machine) could erase it. When this happened the mounting machine detected the STALE NFS FILE HANDLE with the respective file handle.
With the help of the script fhfind from the above location I was able to find the respective file from the file handle information. From this file I could find the software that was missconfigured.