Stale NFS mounts

[mike adams_4]

What is the best way to resolve stale NFS mounts? This occured when rebiulding the NFS server and now the client is reporting stale mounts.

[G. Vrijhoeven]

Hi Mike,

check this link:

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B1031-90053/B1031-90053_top.html&con=/hpux/onlinedocs/B1031-90053/00/00/47-con.html&toc=/hpux/onlinedocs/B1031-90053/00/00/47-toc.html&searchterms=stale%7cNFS%7cmounts%7cresolve&queryid=20041018-075947

Regards,

Gideon

[Pete Randall]

Mike,

The only way to really resolve stale mounts is to reboot. With the latest versions of HP-UX and NFS, this problem seems to happen less frequently - you might want to consider updating and/or patching NFS.


Pete

[RAC_1]

With new NFS version, there is a provision to forcibly unmount stale NFS moounts. If you not have that, your only option is reboot.

You should be moutning the NFS mounts with soft option, so that you atleast get to work.

By default they are mounted with hard option.

Anil

[Chris Wilshaw]

To refresh the mounts, you'll need to unmount each one from the client, then re-mount them (mount -a will do this as long as they have been defined in /etc/fstab.

Ideally, you should unmount them from the client before working on the server (of course, if the server crashed, you wouldn't have a chance to do this).

[V. Nyga]

Hi,

umount -a
and then
mount -a
should work
If not, then you have to look for the process which still accesses the directory.

Volkmar

[Govind_3]

The best and the easiest way is to reboot the NFS clients, after the NFS server is rebuilt, to remove the stale NFS file handles.
The harder way is to however unmount the file system on the client and remount it.
Cheers,
Govind

[Bill Hassell]

NFS is very difficult to handle when server or network problems exist. In essence, when the NFS server(s) went down, it was as if you yanked out a physical disk. There may be processes with open files, directory searches from the filesystems, user shells cd'ed into these directories, etc. So a lot of code needs special handling for loss of files and directories.

With the next release of HP-UX, an NFS forced umount may be available. But for now, NFS servers can never be shutdown until all clients have umounted their NFS mountpoints. This is one of the reasons to use NFS with caution on production servers.

[Muthukumar_5]

Some human checks or automated scripting there.

Check particular Stale mount and unmount that with umount

Another key one is, if some user is on Staled directory then that will be a more problem. We have to tell the logged users to come out of that directory to unmount that FS there.

Simply removing entry from /etc/mnttab may be also used there.