NFS not responding while copying files

[Martin Mohan]

NFS server running in Tru64 Cluster 5.1A.
'/home' file system is exported on the above system.

We are using the above NFS share in many systems which are Tru64 unix systems as well as Suse Linux systems.

We are facing problem in one of our Suse Linux sysetm. Whenever we copy files from NFS share to the local file system or vice-versa. The Linux system hangs for some time. At that time We are getting some errors in /var/log/kernel.log file in linux system.

Apr 16 14:50:04 b934 kernel: nfs_statfs: statfs error = 5
Apr 16 14:51:51 b934 kernel: nfs: server b640 not responding, timed out

The same NFS share is working well in all other systems.

And the NFS mount from some other system is working fine in the linux system.

The details of the systems are follows:

Server:
OS : Tru64 Cluster 5.1A
(OSF1 bwgc559.ts.siemens.de V5.1 1885 alpha)

Client:
OS : Suse Linux 9.1 ( 2.6.5-7 kernel )

I tried to remount the NFS shares, restarting the nfs client daemons and finally I rebooted the client system. Still getting the same problem. We we are using the below option on the client side.

(rw,bg,soft,retry=5,addr=x.x.x.x)

Can you please guide to resolve the above problem. If any other information is required I can provide.


Thanks & Regards,
Martin Mohan Reemala

[anandganjihal]

I suggest you to NFS mount with "hard" option instead of "soft". For example: (v3, rw, udp, hard, intr)

As you told other NFS clients are doing fine and from the err messages it looks like statfs request failed during CLIENT handle creation failing on this particular NFS client. If this client is slain all further I/O fails.

If soft mount and server is down just try once Interruptable hard mounts get counted at this level.

[Hein van den Heuvel]

I've seen situations like that caused with interface/switch incompatibilities.

You may want to doublecheck and/or report the ethernet adapter used, line setting, switch used, switch settings.
Is the Tru64 system connected to the same switch as the other systems or differently?
100 mb or gb? Jumboframe?

As a workaround you could try connecting using tcp instead of udp

fwiw,
Hein.

[reemala]

Thank you very for giving the prompt reply. I will check the network connections and revert back if still there is problem.

Thanks & Regards,
Martin.