[GE users] problem of qdel and parallel running in SGE

reuti reuti at staff.uni-marburg.de
Sat Aug 21 19:05:47 BST 2010


Am 21.08.2010 um 06:34 schrieb mrostaee:

> Thanks for your reply.

please quote always the post you are referring to, snipping out only already answered parts. There is a button "quote" in the web interface.

> mentioned problem of qdel and PE was for Fluent programs.
> defined PE for fluent program, is using a stop script for "stop-args" parameter. in this script call for a cleanup script is done.this clenup script kill all of PID processes of a SGE fluent program, then at the end of line remove itself.

At time the stop_args script runs, the tasks of your application should have been killed by OGE already beforehand. So it seems the application is not tightly integrated into OGE.

> after a qdel command of a SGE fluent program, this cleanup is remeain and this means that qdel didnot done completely and some processes of that program is still running , but qstat hasn't any result for that program (qdel has deleted it , but not completely).
> what can i do for this problem?

Hard to say from remote. I don't use this application, but maybe someone else on this mailing list got a working setup. What parallel library does Fluent use?

I found these:


When it's just MPI (which particular MPI?), you can first check with:

ps -e f

(f w/o -) whether all Fluent tasks are kids of the shepherd. If yes, and you still have processe left, setting:

execd_params                 ENABLE_ADDGRP_KILL=TRUE

in OGE's setup might help.

-- Reuti

> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=275753
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].


To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list