[torqueusers] LAM/MPI + Torque

Marcos Ibanez mgi1982 at gmail.com
Thu Jun 28 11:32:29 MDT 2007


Hi everyone, I'm building a 6 sun fire x2200 cluster based on OpenSUSE
and LAM/MPI and want to use Torque on it. I already made the LAM work,
and tested it with DIRAC04, everything runs smooth at that point, the
lam uses all the cores in all the servers.

The problem is with Torque, i'm not sure if it is well configured and
I'm not starting the job well or there is a configuration problem. I
tried to start dirac with the following script, using the comand qsub
run.sh:

run.sh:

#!/bin/tcsh -f
#
#PBS -l nodes=3,mem=300mb
#

setenv PATH ${PATH}:.
setenv PAM_PATH /home/fisica/DIRAC04

cd /home/fisica/xenon

/usr/local/bin/lamboot -v
$PAM_PATH/pam -wrkdir "/tmp/fisica/xenon" XeF2_hivu32 nmr
/usr/local/bin/lamhalt -v

The problem is that torque starts the proccess in only one of the
cores of the last server. I mean, if i tell him to use 3 servers, the
proccess runs all by itself in a core y server 3, if I tell him to use
5 servers, it only uses one core in server 5, get the idea?

Do you have any suggestions about this problem? It's my script ok? Do
you need any configurations files to look at, just let me know.

Please help me with this, I don't know where else to ask.

Regars.
-- 
Marcos Gabriel Ibañez
Linux Registered User 357259
MSN: mgi1982 at hotmail.com
Cel: 03783-15340575
Web: www.mgi1982.com.ar

Fortune time:
------------------------------------------------------------------
Stewie Griffin: Yes, I rather like this God fellow.
He's very theatrical, you know, a pestilence here,
a plague there. Omnipotence. Gotta get me
some of that.
------------------------------------------------------------------


More information about the torqueusers mailing list