Problem with mount options on autofs @ UX11.11

[Karsten Radke_1]

Hi fellows,

i currently have a problem with an NFS link between a UX11.11 box (client) and a UX11.23/IA box (server).

The server is exporting, among others, a directory named "/export/usr/sap/trans".
When trying to mount this directory on the 11.11 client via autofs, i am getting garbled and corrupt data when i'm accessing files across this NFS link. I re-checked this by issuing a "cksum *" both on the client and the server where i got different checksums for the same files, which means that the file's contents have been changed from the client's point of view.

An "nfsstat" command on the client revealed no errors except for the fact that the client obviously uses NFSv2.

Well, i thought, so let's stop the autofs and mount this manually via the "mount" command without any further mount options:

mount r3rel:/usr/sap/trans /usr/sap/trans

Then i re-checked the mount options:

root@germanix: mount
/usr/sap/trans on r3rel:/usr/sap/trans rsize=32768,wsize=32768,NFSv3 on Fri May 16 14:22:49 2008

OK, i thought, the "mount" command obviously uses NFSv3 by default, which was also confirmed by the "nfsstat" output. I re-checked the files' checksums across the NFS link, and bingo, the CRCs were the same, and there wasn't any data corruption anymore.

But when i tried to configure AutoFS to use NFSv3, i failed. I'm used the following auto_direct entry for this:
/usr/sap/trans -rsize=32768,wsize=32768,vers=3 r3rel.sap:/export/usr/sap/trans

After restarting the nfs client, the AutoFS link worked, but it again uses NFSv2 regarding "nfsstat", and i again got my corrupt data.

Does anybody know if i simply used a wrong syntax in the mapfile or is it a general problem with the UX11.11 AutoFS?

By the way, if the client is an UX11.23 or a Linux box, i have no data corruption at all with this NFS link.

TIA,
Karsten

[Steven E. Protter]

Shalom Karsten,

The options seem fine to me. Therefore I can only surmise that the problematic system needs to be patched.

Start with the bi-annual patch set, if that doe s not work, build an NFS specific patch set and install that.

SEP

[Dave Olker · Accepted solution]

Hi Karsten,

Are you sure you're not using the old legacy automounter? That version is only capable of mounting with NFS v2. If you're getting an NFS v2 mount on 11.11, I'm pretty sure you're using the old one. You can do a "ps -ef | grep auto" to be sure.

My recommendation to *any* customer running 11.11 is to either update to a newer OS or install the Enhanced AutoFS for 11.11 from here:

http://software.hp.com/portal/swdepot/displayProductInfo.do?productNumber=ENHAUTO

This is the same version of AutoFS that runs on 11.23 systems, which apparently don't have this problem. I strongly urge you to install the latest ENHAUTO product and try again.

Regards,

Dave

[Tim Nelson]

I never liked the automountd.

What if you just do a normal NFS mount and leave it that way?

mount nfshost:/path/ /this/host/path

??

[Dennis Handly]

>by issuing a "cksum *" both on the client and the server where i got different checksums for the same files, which means that the file's contents have been changed from the client's point of view.

Yes, the same thing happened here. Some systems would work fine and others failed. It was the perl executable that hung or corrupted object files.

>Dave: Are you sure you're not using the old legacy automounter?

Yes, Dave helped us with a bunch of systems using the old automounter and files were corrupted.

[Karsten Radke_1]

Hello everyone,

so far, a heartly thanks to your efforts and feedback!


To SEP:
both client and server have the current standard patch bundles dated from December 2007. I even did an online update with the "software assistant" for both boxes. Nevetheless, i will contact our HP guy for special NFS patches that might not be already covered by the standard bundles.

To Tim:
in our environment, the automounter proved to be stable and superior to a manual mount in most situations, especially when the NFS server is temporarily unreachable due to network problems. I also would not recommend to use AutoFS in every case.

To Dave:
i'm pretty sure that we are currently using the "classic" automounter. I will act on your suggestion to install and check out the enhanced AutoFS here.

I will provide feedback and also assign points later on :-) Thanks again so far!

Regards,
Karsten

[Karsten Radke_1]

The problem has been resolved by installing and using the "Enhanced AutoFS".

Thanks again to everyone for your efforts and crontributions!

Regards,
Karsten