Two automountd's running

[Mark Ellzey]

All,

I have 2 C160's running 10.20. When I reboot them, I end up with 2 automount daemons running. Although the 'Starting nfs client' script shows no error on boot, the /etc/rc.log shows error code 1 for automountd. The message is "Mount failed! Possible local filesystem mount. If local filesystem, then nfs.server will mount is." I've verified that /etc/auto_master, /etc/auto_direct and /etc/fstab do not contain any local filesystems. I've also verified that /etc/exports on one of the machines has only the correct filesystems exported.

Any ideas?

Mark

[Massimo Bianchi]

Hi,
try doing another reboot and disablign the automatic start of automount.


in /etc/rc.config.d/nfsconf

put the proper variable to 0.

After the reboot you will check, if there is already an automount, then some other script is starting it; if not, then start it manually.

HTH,
Massimo

[Mark Ellzey]

Massimo,

Tried your method. On reboot, no automount daemons were running after setting nfsconf AUTOMOUNT=0. /etc/rc.log shows:
mounting remote NFS file systems ...
mount:: Device busy
FAILURE CODE: 1
Mount failed! Possible local filesystem ...

/etc/rc.log does not show automount starting.

When manually starting automount as such:
/usr/sbin/automountd -f /etc/auto_master
2 automount daemons start.

When executing /sbin/init.d/nfs.client stop:
killing biod
killing automount

ps -ef | grep auto: shows 2 automount daemons still running.

Can kill automount daemons with kill -2.

[Massimo Bianchi]

Hi,
first check your /etc/fstab, looks like you have some nfs_mounts that may interfere with other mounts.
You said: "verified that /etc/auto_master, /etc/auto_direct and /etc/fstab do not contain any local filesystems"

cat you post the files mentioned ?


second: i don't remember exactly ont thing, in 10.20 automount is a script calling the automountd or it is directly starting the files ?


Are you using nis or something ?

You said that thre are two automountd: are they exactly the same ? no switch, no flag, nothing different ? Are they son of each other or are son of init (1) ?

It may worth to start the tracing for the automount:



-T Trace. Expand each NFS call and log it in
/var/adm/automount.log file.

-v Verbose. Log status messages to the system log
file (see syslogd(1M)).




HTH,
Massimo

[Dagmar Boelen]

Hi,

Could you post you nfsconf-file? Boot with the automountd daemon enable and post your rc.log again.

Regards,

D.

[Shannon Petry]

I have a hard time understanding how /etc/fstab,or /etc/auto_master or /etc/auto_home would make any difference.

The automounter is started by init at boot time. Irregardless of the contents of auto_home/auto_master.

What is most likely occuring is that someone has made a duplicate entry for /sbin/init.d/nfs.client in one of the run level directories. I.E. /sbin/rc2.d...


I do remember a quirk with 10.20, that if you put an executable bit on any of the config files (I.e. auto_master) that it would cause strange problems.

make sure your config files are all 644 or 444 permissions.

Sincerely,
Shannon

[Mark Ellzey]

Attached is a listing of the files requested.
/etc/fstab
/etc/inittab
/etc/auto_master
/etc/auto_direct
/etc/rc.config.d/nfsconf
/etc/rc.log (pertinent parts only)

Shannon,
I've verified that all the config files are 644 permissions, and checked /sbin/rc2.d and /sbin/rc3.d for any executable files. There were none, all are links to /sbin/init.d.

Mark

[Shannon Petry]

Is it possible that someone duplicated the startup entries in a different file somehow?

The beauty and curse of Unix is how flexible it is. A previous admin may not have known how to use "proper" methods of starting NFS, so could have modified a boot script, or even thrown an entry into /etc/inittab.

I have seen stranger things before. There is nothing more difficult than taking over someone elses work.

Regards,
Shannon

[Massimo Bianchi]

Hi,
if you don't need them, tear away the
/net

part in /etc/auto.direct

This will allow the automatic mount under /net , and my worry is that there may be some interference.


Another check that you can do: you are using the automount with the V2 option, but there are entries for the autofs too.

So i think that is right, for you , to have
two entries, but slighty different. Here are mine.

root 491 0 0 Apr 25 console 4:26 autofs_proc
root 488 1 0 Apr 25 ? 2:44 /usr/lib/netsvc/fs/autofs/automountd

I use autofs. In ytou case, you wuold see and automount insted of the autofs_proc.


Second: old automount had many problem if it found already the link from previous automount, and were not cleaned well.



So:
reboot with AUTOMOUNT=0

go to the directory that should be mounted. they should apper as link to /tmp_mnt/something.


remove them all, i think that this is the real problem

remove also the /tmp_mnt, it will be re-created automatically by the automount.


issue a "mount -v" to check that the mount table is clean


Then start nfs.client and cross your finger. I admit i hate automounter... Try thinking to use autofs, it is slightly better, and the sintaz is the same.

HTH,
Massimo

[Mark Ellzey]

All,

I've tried Massimo's trick of removing all the dirs under /tmp_mnt, and verifying the /net is empty. I still have both of the daemons running, but everything seems to work ok. I'm just going to leave it, until I can load 11i on the machine(s).

Thanks for all your help,
Mark

[Caesar_3]

Hello!

When you will migrate to 11i then you would
have automount it will be auto_fs

Caesar