Chance to release stale nfs mount from non-existent nfs server without rebooting client?

[Ralph Grothe]

Hi, I'm dispairing.

The problem is as written in the subject line.

The nfs server isn't alive anymore:


ping -t 10 $(grep nfs /etc/mnttab|cut -d: -f1) -n 1
PING 10.10.110.133: 64 byte packets

----10.10.110.133 PING Statistics----
1 packets transmitted, 0 packets received, 100% packet loss


Since the nfs mount is stale no wonder that all nfs requests hang (e.g. bdf), and an umount attempt fails.

umount $(awk '$3=="nfs"{print $2}' /etc/mnttab)
nfs umount: nfs_unmount: /tmp/bwdiwo: is busy

Ok, Itried to stop the nfs client, but since it neither can release the stale mount it shrewedly reappears.

/sbin/init.d/nfs.client stop
killing rpc.lockd
killing rpc.statd
killing nfsd
killing rpc.mountd
starting NFS SERVER networking

starting up the rpcbind daemon
rpcbind already started, using pid: 1249
starting up the mount daemon
/usr/sbin/rpc.mountd
starting up the NFS daemons
/usr/sbin/nfsd 4
starting up the Status Monitor daemon
/usr/sbin/rpc.statd
starting up the Lock Manager daemon
/usr/sbin/rpc.lockd
umountall: umount : has failed.
umountall: diagnostics from umount
nfs umount: nfs_unmount: /tmp/bwdiwo: is busy


Needless to mention that games like such also fail.


fuser -ck $(awk '$3=="nfs"{print $2}' /etc/mnttab)

or
perl -i.stale -ne 'print unless /nfs/' /etc/mnttab && rmdir /tmp/bwdiwo


Is there any hope to get rid of this nfs mount without rebooting the damned host?

[Pete Randall · Accepted solution]

Ralph,

It's been my sad experience that the only way to get rid of the stale mounts is to reboot. Sorry!


Pete

[Ramkumar Devanathan]

see if this helps -

http://bizforums.itrc.hp.com/cm/QuestionAnswer/0,,0x46a43a7b3682d611abdb0090277a778c,00.html

- ramd.

[Ravi_8]

Hi,

i too had expereinced stale nfs, ultimately rebooting the m/c is the solution we found

[Stefan Farrelly]

The only thing ive found which helps is to remove /etc/mnttab. This allows you to survive for some time without errors, but in time you /etc/mnttab gets recreated and you will need to remove it again.

Only a reboot will fix it properly, sorry.

[Olivier ROBERT]

Hi Ralph,

Eventually, removing NFS mount points as the NFS server is dead has been a problem for many of us I guess, not only on HP-UX systems, and I wonder why this has never been adressed. Is there any reason for that? For example, I guess we all would appreciate having the ability to remount soft and/or intr if necessary, and then remove the attached processes and the mount point. Wouldn't that be a good idea?

Regards,

Olivier

[Pete Randall]

I seem to recall that, at one time, there was a patch to the umount command that was supposed to allow you to force the unmount but, if that's true, I don't know what ever happened to it because such an option doesn't seem to exist now.

Any HP'ers out there that can offer any insight?


Pete

[Steven E. Protter]

First I umount the fs which almost never works on the client.

Then I fuser -cuk /mount_point ...

Then I shut down the nfs client and nfs core, immediately restarting.

Then I wake up and boot the box, commenting out anything in /etc/fstab that will generate errors.



SEP

[Paul Sperry]

reboot :(

[Olivier ROBERT]

Awww... considering all replies so far, this looks like a lost cause!

Pete, if such an option of umount ever existed, I suppose it was removed because its implementation was not compliant with some standard. But really, I would appreciate hearing of an HP representative on that subject...

Regards,

Olivier