> arch redhat-6.2
I do not think this should matter.
I looked at the following
globus-job-run farm003.hep.phy.cam.ac.uk /usr/pbs/bin/qmgr -c 'p s'
and
globus-job-run farm003.hep.phy.cam.ac.uk /usr/pbs/bin/qstat -n
and
globus-job-run farm003.hep.phy.cam.ac.uk /usr/pbs/bin/pbsnodes -a
and could not see anything wrong with them. I am not sure why your
jobs are not being evenly distributed at the moment.
On the nodes that are not receiving jobs is there anything in the
mom_logs directory of intrest.
I will ask one of the pbs experts around here.
Steve
-----Original Message-----
From: Santanu Das [mailto:[log in to unmask]]
Sent: Friday, August 23, 2002 10:31 AM
To: [log in to unmask]
Subject: Re: PBS queue
Traylen, SM (Steve) wrote:
> Santanu,
>
> Another thing to check. On each of the worker nodes there should be
> a file
> /usr/spool/PBS/server_name
> containing the name of your CE.
yap......it has.
>
> and
>
> /usr/spool/PBS/mom_priv/config
>
> $clienthost gppce05.gridpp.rl.ac.uk
> $cputmult 1.0
> $usecp *:/ /
> $logevent 511
> arch redhat-6.2
only 'arch redhat-6.2' is missing here. Should I add this?
Santanu
>
> Steve
>
> -----Original Message-----
> From: Santanu Das [mailto:[log in to unmask]]
> Sent: Thursday, August 22, 2002 1:59 PM
> To: [log in to unmask]
> Subject: PBS queue
>
>
> hi all,
> I've got a little problem with PBS queues. We Have 16 WNs and running
> two PBS queues. Though we are using all that 16 CPUs, only three or four
> CUPs is being used to execute any particular job keeping rest of the
> CPUs free. It's something like jobs are being shared among three/four
> CUPs only. Apart from the fact that they tend to saturate a limit of 5
> jobs per machine.
>
> Any idea about how to configure the PBS to use all the CPUs?
>
> Thanks,
> Santanu
>
--
Santanu Das
University of Cambridge
HEP, Cavendish Laboratory
Madingley Road
Cambridge, CB3 0HE
United Kingdom
Phone:++44 (0)1223 337236, 37236 (internal)
Fax:++44 (0)1223 353920
E-mail:[log in to unmask]
|