[gridengine users] Signal sent to processes on requeue

[Reuti]

Am 31.01.2012 um 22:51 schrieb Ben De Luca:

> Strangely im pondering this issue at the moment. If a python process
> is killed any process started with subprocess does not die.

Yes, there is no safety kill in SGE when a job leaves the system.


> The two methods im following,
> 
> 1. a reaper, the grid job starts a python process (parent) that starts
> two other jobs, the task, and the reaper.
> A. The task is the process I want to run.
> B. The reaper, sits waiting to see if the parent python process
> disappears, if it does it cleans up process A.

Why is the process starting subprocesses at all? This might oversubscribe the granted slots for the job.


> When Parent sees A finishes normally it kills B.
> If Parent is killed B kills A.

You could scan for the additional attached group ID for all processes and kill them.

-- Reuti


> 2. I think an epilog can be run to clean up.
> 
> 
> On Tue, Jan 31, 2012 at 8:00 PM, Hugh Macdonald
> <hugh.macdonald at nvizible.com> wrote:
>> Hi,
>> 
>> I've got a python script here that is run through Grid. When I requeue the
>> job, I want to catch the signal in the script, to do a bit of cleanup before
>> exiting. I don't seem to be able to catch the signal, making me assume that
>> it's SIGKILL that is being sent. Am I right in this assumption, and, if so,
>> is there any way to change it to SIGTERM?
>> 
>> 
>> Thanks
>> 
>> 
>> Hugh Macdonald
>> nvizible – VISUAL EFFECTS
>> 
>> hugh.macdonald at nvizible.com
>> +44(0) 20 3167 3860
>> +44(0) 7773 764 708
>> 
>> www.nvizible.com
>> 
>> 
>> _______________________________________________
>> users mailing list
>> users at gridengine.org
>> https://gridengine.org/mailman/listinfo/users
>> 
> 
> _______________________________________________
> users mailing list
> users at gridengine.org
> https://gridengine.org/mailman/listinfo/users
>