Error on NFS-mounted filesystem.

[Leif Halvarsson_2]

Hi
Sometimes we get this error when running an application. The application data is located on a NFS-mounted volume (HP-UX 11.11 server). We don't get this error with local data or with NFS mounts on a Linux server.


? FORTRAN Runtime Error:
? Attempt to read past end of record
? READ(UNIT=10,...
Command exited with non-zero status 255
2550.43user 203.71system 57:22.47elapsed 80%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (18085major+128941minor)pagefaults 0swaps
execution complete /home/appl/okg/hot_bird/okg_bird/cas/c169_kjell
Tue Mar 18 10:53:16 CET 2003


Anyone who have an idea about the reason ?

[Paula J Frazer-Campbell]

Hi

Are the mounts soft or hard -

If soft try hard.

Paula

[Leif Halvarsson_2]

Hi,
The filesystem is mounted with default options, as I can find, hard is default.

[U.SivaKumar_2]

Hi,

Increase the timeout .by specifying timeo= option while mounting or in /etc/fstab.

regards,

U.SivaKumar

[Stefan Farrelly]

Did you get a corresponding error in dmesg or syslog when your program errored ? I would expect so if its an nfs problem (connection/performance).

We use nfs mounted filesystems a lot and we have had to do a lot of research into nfs performance. The best source we found is the book "Optimizing NFS performance" by Olker (for HP). Excellent. It points out things like; do NOT use the timeo option!

We get errors ocassionally but always with an error in syslog/dmesg also. Weve found the following nfs mount options the more reliable and for good performance;

rw,bg,hard,intr,retry=999,retrans=999,rsize=32768,wsize=32768,proto=tcp,vers=3

You can confirm you current mount options with;
nfsstat -m

And dont forget to load the latest nfs patches (very important) and change quite a few kernel parameters (even some undoccumented ones for nfs - theyre in the book).