Problems with the automount

[Marta_3]

Hi,
I have a problem with my HP-UX 11.0 server , sometimes when I??ll execute any command , for example ( bdf, ls -la (from /),pfs_mount ...)
The next message appears:
NFS server hostname not responding still trying
I have been looking for the reason, and I think that the automount it??s the guilty.
Has anyone any idea?
Thanks in advanced,
Marta

[John Carr_2]

Hi

have you tried stopping and restarting the automounter, you can do this by :

cd /sbin/init.d
./nfs.server stop
./nfs.server start
./nfs.client stop
./nfs.client start

John.

[Tom Geudens]

Hi,
It could be automount ... but it could also be pfs_mount (which I noticed in your list). There are several (more like a hundred :-) threads on pfs_mount and what can happen if you start the pfs-daemons in the wrong way ...

You might want to take a look a those threads too (and hope it's not pfs_mount ... the only solution would then be a reboot).

Regards,
Tom Geudens

[sven verhaegen]

OK how to Start ?

What does it mean:
==================
in general this error is a direct result of a client NFS request to the server either being acknowledged late or not acknowledged at all after a number of retries and a timeoput sequence , in general the problem resides on the server side or the network not the client (however it is possible)

Troubleshooting
===============

are both systems correctly patches ? if not get to the latest patchlevel of the ONC NFS/NIS megapatch + dependencies to avoid any known bugs

running nfsstat on server and client could give you stats on types of errors generated
are it reads failing/writes failing /getattributes etc... , this could lead to looking at the servers disk access times

is a nfs export from the client mounted on itself also slow ?? that could isolate the problem to the server and network also

is a manual mount suffering from the problem ?
that could tell you if automounter is involved or not

ping to the server from the client with a large packet size will tell you if there is packet loss , if so this could be causing you to loose to many packets and that would result in server responce errors

looking on your network devices collision rates could be determined , high collision rates could make everything slow and make your requests timeout and being late on the client

If all fails only a network trace on server and client will help out , using nettl on the HP-UX or a lan analyser are needed then , this requires in depth knowledge of network packets and structures + knowledge of NFS protocols

[Anil C. Sedha]

Marta,

Automount has been always a problem on many occasions. There is a new patch for network performance improvement. PHNE_25234.

However, also do the following for analyzing if your problem is really the automounter. Do the automounter logging by using the options from the document attached herewith.

Also, make sure that you have a crontab entry replacing the automount file every 12 hrs as the file gets really big. Copy the file as automount.old and then replace the original file again.

Regards,
Anil

[Sridhar Bhaskarla]

Hi Marta,

I would stop automount on the fly without unmounting automounted file systems. If you are unable to find the mounted filesystems (bdf hangs), you can get that information by doing a cat /etc/mnttab

This will list out all the nfs mounted filesystem that are responding/not responding. Based on this information verify the nfs servers from where these filesystems are mounting. Look for proper exports and nfs server on them and fix them. Once you fix them, then you can unmount these nfs filesystems. Once they are unmounted, you can safely stop the nfs services.

If you don't need them anymore, take them out from /etc/fstab and disable NFS client in /etc/rc.config.d/nfsconf.

-Sri

[MANOJ SRIVASTAVA]

Hi Marta

In /etc/rc.config.d/nfs.conf change to the following :




NFS_CLIENT="0"
NFS_SERVER="0"
NUM_NFSD=4
NUM_NFSIOD=4
PCNFS_SERVER=0


stop the nfs daemons and you are good to go , most likele in you case if this is not the problem them you will have to boot the server as sometimes the nfs processes dont die , though it wont effect the server perfomnance but it wont come out from ls , bdf commands.
This may also be caused by umounting a pfs_mounted CD using a umount command rather tahn a pfs_umount , or killing pfs processes in the wrong order.

Manoj Srivastava

[pap]

Hi,
YOur error message describes that NFS server is not reachable. That means there is

one of the following

1. Network connection between server and your machine. (try to do ping and veryfy)

2. once the NFS file systems are mounted on client and then connection lost between client and server at this time if you try to access NFS mounted file system, it will simply give the above message. ( You need to do /sbin/init.d/nfs.client stop
/sbin/init.d/nfs.client start)

If possible on the server side stop and start nfs services.

3. you need to delete /tmp_mnt directory.

4. Check the correct IP and host name in your /etc/hosts file

5. If you are using dns for name resolution you need to check dns services working properly and is able to reolve NFS server's name correctly. (Check /etc/resolv.conf and /etc/nsswitch.conf for proper order and entries.)

I think you are all set now.

-pap

[Sridhar Bhaskarla]

There was no response to this thread After April 03 and I don't know if the author is going to look at it.

There was a typo in my posting. Please read "I would stop automount" as "I wouldn't stop automount". You should neither kill automount nor stop nfs client if your bdf is hanging.

-Sri