Oracle apps 11.5.9 Hangs when peer NFS server rebooted?

[Liam Curtis]

Hi all. Running two rp7410's with HPUX 11.11. Our production box uses an NFS mount on our dev/test box for simple file and patch shares. When we rebooted the box, i noticed in the syslog about the nfs server not responding, but then at the same time, the concurrent manager hung on our production box. A simple solution is to not automount the nfs filesystem, but was wondering if anyone had run into this situation before?

thanks

[Steven E. Protter]

You really, really should reconfigure oracle so there is no oracle app or database reading any data on an NFS mountpoint.

Its not recommended by Oracle and if you are not careful you will hang a production application some day.

At the very least, take your own suggestion and stop automounting the filesystem.

SEP

[Liam Curtis]

Thanks for replying Steve. That is the thing...it is just a filesystem we mount for patches, etc. and there are no oracle processes reading or writing to this nfs filesystem, so I am just trying to figure out why certain processes are hanging.

[Eric Antunes]

Hi Liam,

A way to find what's bad configured in your APPS is to find files refering that filesystem in your $APPL_TOP as apps user (usually applmgr):

$cd $APPL_TOP
$cd ..
$find $APPL_TOP -name *.env -exec grep -l '' {} \;

Best Regards,

Eric Antunes

[Eric Antunes]

I wanted to write:

$find $APPL_TOP -name *.* -exec grep -l '' {} \;

[Liam Curtis]

Thanks much! I did, in fact find entries, but they were just log files from where patches were located and installed from. I already manually unmounted the nfs mount and took out of /etc/fstab and there are no probs...I was just hoping someone might have seen this scenario.

Still though, I like your approach to troubleshooting and the switches on the find command as I am still a green unix admin ;))

[Liam Curtis]

Steve! I am so sorry! I clicked the button before I assigned points. I can reassign if you want to reply.

[Steven E. Protter · Accepted solution]

NFS is almost always a chicken or egg deal.

If you have two servers with NFS mounts on each other, you can have this kind of trouble.

So long as Oracle is not involved, you are okay.

I would suggest instead of automounter, soft mounting. Soft mounting lets the NFS client re-connect after the server has been rebooted.

Here is a link. Don't worry about point assignment. Give this one what you think it deserves.

Links:
http://www2.itrc.hp.com/service/james/dispDoc.do?docURL=http%3A%2F%2Fdocs.hp.com%2Fen%2FB2355-60105%2Fmount_nfs.1M.html&aid=SEARCH_MANUAL&pil=4&serStr=NFS+soft+mounting
http://www2.itrc.hp.com/service/james/dispDoc.do?docURL=http%3A%2F%2Fdocs.hp.com%2Fen%2FB5140-90011%2Findex.html&aid=SEARCH_MANUAL&pil=8&serStr=NFS+soft+mounting

SEP

[Devender Khatana]

Hi,

If your NFS mount is not in use and you want to reboot your test/dev server then STOP NFS client services on your Production BOX so that it is not searching for it any more.

#/sbin/init.d/nfs.client stop

But it will not be possible if you have multiple mounts from different servers on your production box and you want atleast one of them to be available.

HTH,
Devender

[Liam Curtis]

Thanks all.. weird...could not get the link to work. will have to try again.