NFS problem HPUX11i <-> SCO

[Kay Liesenfeld]

Hi out there,

since we have installed our new rp5470 with HP-UX 11i we have some troubles with a NFS connection to another UNIX-Machine (SCO on HP-Hardware).

The remote machine is located at one of our customes so we don't have any adminstrative rights on it.

At some time, often at night when no traffic is shared between the servers, the NFS-connection is going down. At the early morning, when our server is trying to get some data from the remote machine, we get a "NFS not responding -- till trying" error.

The only solution is to kill and restart the NFS-server process at the remote computer. (The customer's admin is doing so when we call...)

Or -- if we are able to unmount and remount the file system again the NFS is re-connected too.

Does anybody have an idea what to do that the connection does not break down every night? Our customer has not changed anything at his configuration. We have set every parameter like they were on our old server -- we did not have any problems in the past!

Thanks in advance.

Kay.

[U.SivaKumar_2]

Hi,
Are you using NFS over UDP or TCP in SCO side ?.
Are the servers connected by WAN ?.

regards,
U.SivaKumar

[Kay Liesenfeld]

Hi,

there are no specific options set in the fstab file so I think TCP will be used.

Yes, both servers are connected via WAN. We have alread checked all cables and switches.

Kay.

[U.SivaKumar_2]

Hi,
Try setting "timeo" value of mount in /etc/fstab to higher value in case you are using NFS over TCP .

regards,
U.SivaKumar

[Kay Liesenfeld]

We have set timeo already like

klr:/data/hp_io /nov/klr nfs rw,timeo=5,retry=5 0 0

with no result.

Regards,
Kay.

[sven verhaegen · Accepted solution]

Hi

First some remarks regarding NFS version TCP or UDP , there is a very big difference between the two in terems of the manner the data is transmitted and checked , the fact that it worked fine before could mean it worked fine in UDP , NFS TCP is only active by default in 11.i implementations so check using nfsstat what version your mountpoints run ... , if this is TCP there could be either bug on HP-side or on the SCO-side on the TCP version of NFs , for HP_UX ,make sure your client has the latest NFS/ONC megapatch installed + its dependencies (should be about 15 patches).. further server not responding just means the other side doesn't reply anymore , the NFSD does not service any request from the HP-Ux at that moment because of watherver reason , on th eSCO you should try to find out why the NFSD's just stop reacting , did they find any errors there ??? apparently we dont have a timing issue we just have a deamon on the other side giving up , this means we should concentrate some more on the SCO .. you could try to trace the network on the SCO and the HP and see what happens.. in my experience these kind of error generally mean there is a SW problem with the NFSD's thats causing this

[U.SivaKumar_2]

Hi,
Are you sure that SCO supports NFS over TCP ?.
Have you tried increasing the current timeo value ?.
Can clients from your customer side access the SCO NFS server without any problem ?.

regards,
U.SivaKumar

[Kay Liesenfeld]

Hi all,

thank you for your replies.

I'll try the udp-version. Maybe that will work this night.

See you tomorrow!

Kay.

[Kay Liesenfeld]

Hi,

we have tried UDP for several days and it seems to work properly.

Thank you all for your assistance!

Kay.