Urgent: automounter screwed file system table

[Ralph Grothe]

Hi,

I've an urgent problem.
Don't know what has ridden me.
I was so insane to have an NFS export mounted through automounter.
After I had finished work I manually umounted the shares (sic!) and shut down nfs.client services through init script.
Now I lost my mnttab, i.e. bdf shows nothing.
A sole "mount" didn't recreate the mnttab.
I also tried restarting NFS client an hand nfs_mount the shares as before.
Automounter left the /tmp_mnt/net/... dirs stale lingering there.
I'm getting nervous.
This is a clustered productive DB server and I'm anxiously wait for the app to fail-over.
Any quick fix?

[Ralph Grothe]

Sigh,
another restop/start of nfs.client broke the deadlock.
Probably it was because I forgot to send automounter another SIGUSR2 to toggle logging off before shutting it down.
It may have been locked in the pipe?

[Fred Ruffet]

Aren't the shares available on the other node ?
Have you tried to disable auomount in /etc/rc.config.d/nfsconf ?

Regards,

Fred

[Fred Ruffet]

Is your problem solved then ?

Fred

[Ralph Grothe]

Fred,

yes I did

# grep ^AUTO /etc/rc.config.d/nfsconf
AUTOMOUNT=0
AUTO_MASTER="/etc/auto_master"
AUTO_OPTIONS="-f $AUTO_MASTER"
AUTOFS=0
AUTOMOUNT_OPTIONS=""
AUTOMOUNTD_OPTIONS=""


What still bothers me that there are the stale mountpoints from automounter left over.
Do you think I should delete them up to /tmp_mnt ?

# ll /tmp_mnt/net/clark/export/
total 0
dr-xr-xr-x 2 root sys 96 Dec 15 18:06 ansic
dr-xr-xr-x 2 root sys 96 Dec 15 18:06 langtools

[Ralph Grothe]

Yes, everything seems ok again :-)

[Fred Ruffet]

Problem with NFS is this "my-server-disappeared" problem. I heard it is to be resolved on future released, but for now, when a server is down clients are unable to umount (or hardly). So I don't know if you will be able to. Even breaking all the stack (nfs.client / nfs.core / RPC) may not resolve the problem. Recently, I had such a problem and it last until next reboot...

Regards,

Fred

[Dave Olker]

Hi Ralph,

If your system is running 11.11, you should definitely consider migrating to the Enhanced AutoFS product and away from the legacy automounter. Enhanced AutoFS can be downloaded for free from the http://software.hp.com site.

Regards,

Dave

[Ralph Grothe]

Hi Dave,

glad to see you here again.

Yes, I know how much my performance issues improved (in fact they vanished, at least for the NFS part) since I upgraded to NFS V3 and EnhancedAutoFS and TCP has taken over the transport, thanks to your many helpful replies.
So I too well know the blessings of an upgrade.
Yet, this was an isolated instance with this box as it doesn't host any NFS stuff in normal operations.
I only abused the suspended NFS services in order to mount the shares of a C compiler on another box to install a few Perl modules that had XS code stubs in them.
Because I was too lazy to manually mount the shares I wanted to rely on the hosts-mapping of automounter, which was overkill for my purpose and only brought me into trouble.