[GE users] Dynamic queue -- sge schedule policy

Reuti reuti at staff.uni-marburg.de
Fri Aug 25 11:09:12 BST 2006


Hi,

Am 25.08.2006 um 03:54 schrieb Eric Zhang:

>
> Hi, Reuti:
>
>     I have read your suggestions, thanks. But I still have problems  
> with
> this:
>
> ////////////////////////////////////////////////////////////////////// 
> ////////////////////////////////////
> For the jobs with 16G request, use a second queue but attach only the
> @mem16, and subordinate the first queue. As only jobs on the same
> host are suspended, it should be sufficient to simply subordinate the
> compete other queue.
> ////////////////////////////////////////////////////////////////////// 
> ////////////////////////////////////
>
> Does this mean when I submit a 16G job, the 4G jobs will be suspended
> and trigger a reschedule/restart event? How does the sge reach this?
> Just define 16G queue as a subordinate queue of the first queue?

In the 16G queue, the queue definition should contain:

subordinate_list      my_4G_queue

This way, the jobs will be suspended (by sending a sigstop signal to  
the processgroup) - that's all.

If you want them to be rescheduled, you must use a checkpointing  
interface in addition and define the triggering event for the  
migration: "when x" like explained in "man checkpoint".

> In more details, I don't know what this sentence means:
>
> "As only jobs on the same host are suspended, it should be  
> sufficient to
> simply subordinate the compete other queue."

Just ignore it - it was a wrong idea - sorry.

-- 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