Hello,
I am using autofs-4.1.3-234 on a RHEL 4 machine, using kernel
2.6.9-78.0.1.ELsmp, getting automount maps from LDAP.
Recently I had failures in some hosts when mounting home directories, in
some cases more than one host at a time.
I found this on the NFS client's log:
Jun 30 16:51:21 clnt automount[9458]: >> mount: mount point /home/user does not
exist
Jun 30 16:51:21 clnt automount[9458]: mount(nfs): nfs: mount failure
srvr:/export/home/user on /home/user
Jun 30 16:51:21 clnt automount[9458]: failed to mount /home/user
But on the NFS server, this is all I see:
Jun 30 16:51:21 srvr rpc.mountd: authenticated mount request from clnt.example.com:731 for /export/home/user (/export/home)
This is the same log I have when the mount succeeds.
I believe the NFS server might have been overloaded at that time, and
for that reason the mount request might have taken longer than expected.
From what I've seen, I believe the ">> mount" message is the message
returned by the /bin/mount command spawned from automount, and it means
when it tried to mount the directory, the local /home/user directory was
no longer there.
I also believe this might have happened because it was previously
mounted but not used for the timeout period (I am using 5 minutes as
timeout), so the last mount was expired (and the directory removed) just
before this /bin/mount process was able to run.
I tried to google for it, but all I found was this message:
http://lkml.indiana.edu/hypermail/linux/kernel/9910.3/1105.html
It looks like the problem I am having, however it is very old, from
about 10 years ago...
Is this still a known issue in autofs? Is there a patch for it?
I also know that autofs 5 is in many ways different than autofs 4 (I run
it in my RHEL 5 machines). Now (since RHEL 4.7) I believe autofs 5 is
available for RHEL 4. Is it possible that upgrading to autofs 5 might
fix this issue?
Any help trying to figure out what is going on would be very
appreciated. Also, if you need more logs or for me to execute some
commands or tests I can do it for sure. I was thinking of a way to
trying to reproduce this problem (maybe replace /bin/mount with a script
that does "sleep N" before calling the actual mount) but could not come
up with something good enough to validate my hypothesis, so I decided to
ask the list first.
Thanks!
Filipe
_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs
