Qube! Priority isn't working right

[Wanderer]

Hi everyone!

I've been using Qube for 2 months already, so I think I know how clusters and priority works. Still I know I have a lot to learn about it.

For some reason, our farm is acting weird these days.  Priority doesn't seem to work properly. Even thou multiple jobs are in the same cluster and they're all assigned a different priority (let's say that they have been submitted at the same time too), Qube doesn't care. 

It starts subjobs randomly inbetween the first half of the jobs that are most prioritized. At least it doesn't starts the less important ones but it's still annoying. There's no significant pattern in the way it chooses the jobs, but it won't finish the jobs on top before starting new ones (like it's suppose to be).

And of course, there's no

[jburk]

Check and see if you have "stub_optimize" set in your supervisor's supervisor_flags parameter.  If so, try un-setting it, the restart the supervisor service.

This flag enables a dispatch optimization which results in jobs that are submitted within a short time period of each other being evaluated as a single entity.  It's purpose is to speed up dispatch decisions in scenarios where many jobs are submitted quickly, but will (by design) result in those jobs not being dispatched in a strict first-come, first-served order.

[Wanderer]

The guy responsible of our farm says that our supervisor flag is setup this way:

"supervisor_flags=6150 (host_recontact,heartbeat_monitor,running_monitor,retry_busy)"

"stub_optimize" isn't activate at all.