NFS not responding

[firas_1]

K460 server is the nfs client, K570 is nfs server. from time to time this message is displayed " NFS not responding" i need to solve it, as for 2 months all trials fails.

[Clemens van Everdingen]

Hi,

Have you checked your network availablity at these moments ?
Patch level on NFS / Networking ?

C.

[firas_1]

Yes,
PHNE_25234 NFS cum. (+ dep. PHCO_25640 libc.cum).

[Vincent Stedema · Accepted solution]

Hi,

Could you pls post the output of "nfsstat -c" and "nfsstat -m" on the client and "nfsstat -s" on the server? These can help in determining the source of your problem.

At what speed do the nic's in these system's operate?

Regards,

Vincent

[kish_1]

if is working nfs is working on udp , try configuring through tcp, it will work but perfomance will be matter
try configuring through tcp,

ypu can use following command to find out
rpcinfo -p |grep nfs

[firas_1]

Check the attachement.

[Vincent Stedema]

Hi,

The link to the attachment doesn't work...

Regards,

Vincent

[firas_1]

attachement again, nfs.txt

[Vincent Stedema]

Hi,

As you can see at the end of the attachment, the number of retransmissions ("retrans") is exceptionally high. Same goes for the number of timeouts.

You would probably want to experiment with adding specific mount options in /etc/fstab, instead of using the defaults.

Because of the frequent timeouts, try setting the "timeout" mount options to 14. The client will then wait longer before it decides that the NFS call has failed. This helps if your network performance is slow.

You could also try and set the NFS version to 3 instead of 2. Add the following mount options to the nfs mount in /etc/fstab: "vers=3,proto=udp".

More information on configuring nfs and the specific mount options can be found at http://docs.hp.com/hpux/onlinedocs/B1031-90043/00/00/9-con.html .

If i were you, I'd not blame these problems solely on your system config. Ask the network admins if the network configuration is optimal. Ask them if they could check for a large number of collisions on the network etc.

Keep your eye on the "retrans" value in the output of "nfsstat -c" on the client system. You can reset the counters by issuing a "nfsstat -z". Then, experiment somewhat with NFS mount options. The number of "retrans" should not become more than 5% of the total number of "calls". In your current setup, they are at 286%!

My guess: network problems. Check if the nic's in your systems are set to the correct speed.

HTH

Regards,

Vincent

[Jim Keeble]

It appears that either the client requests are not reaching the server, or the responses are not reaching the client.

Try pinging with larger packets, ie.,

ping client-host 8000

And on the client, run netstat -s -p ip to see if anything is dropped due to fragment re-assembly timeouts.

Also note whether the client and server are on the same or different subnets.

If these things indicate any problem, you could have a link mis-configured, or a router problem .

[Oleg Zieaev_1]

Hello.

You have very high number of retransmissions.
If you have extensive load on the subnet, you might consider putting NFS on the dedicated network w/dedicated lan cards.

Ask your network admins to put sniffers on your segment and investigate reasons for retransmissions.
netstat -s -p ip is a good idea. However I think based on reports provided, datagram assembly is not the issue here.

If you have intentions - run nettl and analyse IP headers and agreed MTO size for IP packets between two of your hosts.

Hope this helps,
0leg

[firas_1]

Oleg Said:
If you have intentions - run nettl and analyse IP headers and agreed MTO size for IP packets between two of your hosts.

Could you please give more details on using nettl and how to analyze IP heads and MTO.