[GE users] qdel not remove all instances of jobs

Lloyd Cha lccha+sgeusers at immerbox.com
Wed Mar 28 01:20:23 BST 2007


I have a similar problem with a different tool.  The common
characteristic is that I am also dealing with a wrapper script that
spawns process with a new process group id.  I have attempted to work
around this by putting intelligence into a higher level wrapper that
attempts to walk the tree and kill all the child processes, but this
does not always work properly since I can't walk and kill all the
processes atomically.

I noticed that all the jobs that are spawned by the shepherd seem to
share the same session ID, which is the PID of the shepherd's
immediate child process.  Would it be possible for the shepherd to
kill all processes with the session id of the child process before
exiting?  Since this might not be the cleanest kill, it might be
suitable only for a "last ditch" kill -9 after all the usual kill
methods have failed.

Note:  I'm using RedHat Enterprise Linux 3.

-L

> On Mar 22, 2007, at 6:18 PM, Reuti wrote:
...
> >this is clear now: the processgroup 12003 will be killed by SGE but  
> >not more. Unfortunately the "jmonitor" is already creating a new  
> >processgroup 12192, and also the "sh -c /scr/..." the 12222 one.

---------------------------------------------------------------------
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