[torqueusers] scheduling cycle

Ronny T. Lampert telecaadmin at gmail.com
Wed Aug 8 02:07:25 MDT 2007

> My torque2.1.8 server has scheduling cycle:
> set server scheduler_iteration = 600
> When I submit jobs, they run immediately, none of them 
> waited 10 minutes until the next scheduling cycle.
> Why?

scheduler_iteration is (as far as I know) only a SUGGESTION to the 
scheduler; pbs_sched will use that value, maui/moab will not.

The reason why jobs run immediately is because pbs_sched gets notified 
by pbs_server of a new event == a new job. This will initiate a small 
scheduling cycle.

I used to qsub a small dummy job to get the jobs quickly running after I 
had to restart pbs_server or pbs_sched and there were free nodes, but 
the scheduler was still sleeping until the next iteration.


More information about the torqueusers mailing list