vmunix: NFS write failed for server : RPC: Timed out#

[Jason Chia]

Hi,

Recently I found out some nfs errors messages in our syslog as follows:

vmunix: NFS write failed for server : RPC: Timed out#

Is it due to the network performance problem that at that time the nfs mount point not able to mount whenever the server needs to access the mounted point?

Do anyone encountered this problem before? Please kindly advises. Thanks in advance

[Steven E. Protter]

Okay, there are a few things to check here.

I have had the problem btw.

lanadmin -x 0

the zero is the last part of the lan interface. If your network connection is lan1 replace the zero with a 1.

This will get you information on how your machine think the lan is configured.

If you have a hub, make sure the lanadmin results match the hub's speed and settings.

Do this for both machines.

If you have a switch, check with network administration on the switch settings.

If they match the lanadmin results move on.

On many Cisco switches, autonegotiate setting on the switch causes HP network problems including your error.

HAve network administration hard code the switch to the maximum settings on your lanadmin.

Check with hp that your network card setup is maximized.

There is a file in /etc/rc.config.d called btlan...

... means I can't remember the last part which lets you force your cards to come up at full speed.

Here is a story.

I had problems with Ignite transfers. The switch was set to auto negotiate and the card showed up as 100 BaseT full duplex.

Because of known problems with Cisco swtiches, we had network administration change the switch setting to 100 BaseT full duplex manual(no auto-negotiate).

Reboot the HP-UX server and ran lanadmin -x 1

Came up 100 BaseT Half Duplex.

Then I had to mess with the btlan... file and finally things came up the way they were supposed to on the network side.

So, its kind of important.

This is enough, write back if there are other issues once this is worked out.

SEP

[Jeff Schussele]

Hi Jason,

You now see why NFS is, at best, a necessary evil in Production. Something - literally - timed out in the long NFS chain. Probably either one of the ends - but certainly very possibly in the middle. So you need to check the logs - turn up inetd logging (inetd -l) if needed to gain insight. Follow network stats - UDP especially if using defaults.
Then verify that all NICs & switches are locked in at optimal settings. Verify MTU size through the chain if possible.

Quite simply NFS is a major pain because of all the points-of-failure & because by *DEFAULT* you get network traffic with no error checking at critical layers. Can cause a lot of overhead on already stressed nets.

Good Luck,
Jeff (Remember logs are your friends - poor networks aren't)

[T G Manikandan · Accepted solution]

check this document

http://www1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000066581018

[K.Vijayaragavan.]

See the performance tuning tips for NFS server @ URL >>>


http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B1031-90043/B1031-90043_top.html&con=/hpux/onlinedocs/B1031-90043/00/00/34-con.html&toc=/hpux/onlinedocs/B1031-90043/00/00/34-toc.html&searchterms=server%7cslow%7cNFS&queryid=20030304-095144