[GE users] Cleanup on Rescheduling and Deleting

Reuti reuti at staff.uni-marburg.de
Fri Jan 21 22:00:49 GMT 2005


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi,

> I am using the latest 6.0u3 on Fedora Core 1.  My jobs run on multiple
> hosts/cpus and build a temporary directory structure to use while
> running and then delete them when they finish.
> 
> I would like the script to be able to delete the directory structure if
> it is deleted or rescheduled, but can't seem to find a way for them to
> be notified before they are killed.

depending on your SGE setup, there is more than one possibility to do it:

- you can use the created $TMPDIR.
- put the creation/deletion in a queue or PE prolog/epilog.
- modify the termination method to send a special signal to your job, which you 
can catch with `trap` in your script and remove the directories, before you 
send a kill -9 to the whole process group.

> In my testing, I've noticed that when recheduling a job, the old job
> continues running for 5-15 seconds after the new job starts up.  This
> means that the standard and error outputs get mixed with one or the
> other of the jobs winning.  That is, you don't get both, you get some of
> one and some of the other.

How do you initiate the reschedule of a job? - Reuti


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