[GE users] monitoring spawned processes

Jeroen Kleijer jeroen.kleijer at xs4all.nl
Thu Nov 25 22:56:58 GMT 2004


On Thu, Nov 25, 2004 at 07:55:35PM +0100, Reuti wrote:
> > I'm running into a problem when I try to submit an Abaqus job with SGE.
> > I can start an Abaqus job with either qrsh or qsub and it gets submitted
> > to one of our compute servers fine. On the compute server a process
> > "abq641" gets started and after 5-10 seconds it spawns a number of
> > processes (Python, pre.x, whatever) and the "abq641" process ends, thereby 
> > ending the job as well even though there are still processes running in the 
> > background doing all kinds of calculations.
> 
> Is this "abq641" used to start a job on a local nodes (with out SGE) in the 
> background? Is there any other script to run it in the foreground?
> 
This "abq641" is a binary you run which takes a couple of parameters and
runs different programs depending on the parameters. When you're doing
calculations it almost immediately spawns a "Python" program for
analysis. It's parent process is the shell which was started for the
abq641 command but when the abq641 command finishes it also ends the
shell. The Python process then gets PID 1 as owner (and spawns several
different processes by itself)

> > When I run qstat the abaqus job (abq641) has finished but our compute
> > nodes are still happily running the spawned processes but we have no
> > idea these jobs are still running untill we login on our servers and
> > check this by hand.
> > 
> > Is there any way for SGE to track these kind of spawned processes?
> 
> If SGE is proper setup all the running/spawned processes should be killed, 
> when the main task ends.
> 
I think I've set it up correctly but maybe I should be avoiding the
abq641 command and try to call / run the Python process directly instead
of calling it via abq641.

> CU - Reuti
> 
Cheers,

Jeroen Kleijer

> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
> 

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list