[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] How to switch jobs from idle to run state ?




Le 4 oct. 2009 à 20:04, Mark Tigges a écrit :

My question is, what do you need condor for?

If you have a core2 duo, I think you should only run 2 concurrent
jobs.  Then since all jobs run on the same machine the only
'scheduling' you need is to run a new job when one job finishes.

A Core 2 Duo is a 4 core machine.


Write a really simple that makes sure that there are always two
threads running.  Wake up every 2 seconds or so ... check to see if a
thread ended, if so, start it up again with the new parameters for the
next job.

Condor is for scheduling jobs across lots of machines, if you only
have one machine, you don't need condor.

This is precisely I don't want to do : manage by myself the job flow on the machine. If condor is able to work with numerous machines, it is able to do with one which is in fact 4 processors. And I don't want to organize the logs and job history either.

On Sun, Oct 4, 2009 at 10:47 AM, Marquette Jean-Baptiste
<marquett@xxxxxx> wrote:
Hello everybody,
I am a newbie in condor I installed a few days ago on my MacPro Core 2 Duo.
I want to manage jobs on that machine which is configured to manage, submit
& execute. The jobs are based on the same shell script and I wish to have a
job per processor running permanently. I use default configuration files
customized to apply to my machine.
To test the system I submit 4 jobs in a row and I got this output from
condor_q :
ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD

 37.0   marquett       10/4  16:49   0+00:00:00 I  0   0.0  scampall.sh

 38.0   marquett       10/4  16:49   0+00:00:00 I  0   0.0  scampall.sh

 39.0   marquett       10/4  16:49   0+00:00:00 I  0   0.0  scampall.sh

 40.0   marquett       10/4  16:49   0+00:00:00 I  0   0.0  scampall.sh
After a while 3 jobs are running :
ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD

 37.0   marquett       10/4  16:49   0+00:06:57 R  0   0.0  scampall.sh

 38.0   marquett       10/4  16:49   0+00:06:57 R  0   0.0  scampall.sh

 39.0   marquett       10/4  16:49   0+00:06:55 R  0   0.0  scampall.sh

 40.0   marquett       10/4  16:49   0+00:00:00 I  0   0.0  scampall.sh
Then the 3rd job returns to idle after 12 mn :
ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD

 37.0   marquett       10/4  16:49   0+00:17:35 R  0   0.0  scampall.sh

 38.0   marquett       10/4  16:49   0+00:17:35 R  0   0.0  scampall.sh

 39.0   marquett       10/4  16:49   0+00:12:46 I  0   0.0  scampall.sh

 40.0   marquett       10/4  16:49   0+00:00:00 I  0   0.0  scampall.sh
I never saw the 4th job switching to running state.
So my question is : what are the parameters to change in the config files to
have 4 jobs running in a row ? How to force the running state ? I dont'
see in the documentation any evident parameter to tweak.
All my attempts to that purpose were unsuccessful.
I have more than 237,000 similar jobs to submit, so I would greatly
appreciate any help.
Thanks in advance,

====================================================================
Bien cordialement - Very truly yours - Mit freundlichen Gruessen,
Jean-Baptiste Marquette
Institut d'Astrophysique de Paris
CNRS - UMR 7095
Université Pierre & Marie Curie
98bis Bd Arago
75014 Paris - France
Tel 33 (0)1 4432 8196
Fax 33 (0)1 4432 8001
====================================================================