[torqueusers] Newly online nodes / queued jobs

[Garrick Staples]

On Thu, Jan 04, 2007 at 11:42:32AM -0600, Chris Evert alleged:
> Torque Users,
> 
> I fixed a node and put it online.  There were some 200 jobs in the Q 
> state, but none are going onto the node.
> 
> The situation didn't change after 5 minutes, which I believe is the 
> sleep time for my queue server and scheduler.
> 
> Thinking that jostling the job mix would relieve the logjam, I submitted 
> new jobs and they went right onto the newly available node.  When those 
> jobs finished, nothing old went on the node.  One of the running jobs 
> finished and one of the queued jobs took its place, but the now online 
> node remains idle.
> 
> qrun successfully started a couple of jobs on that node.
> 
> I am using torque-2.1.6 and maui-3.2.6p14
> 
> Why aren't jobs that are already chomping at the bit to run jumping onto 
> newly onlined nodes?  More importantly, how can I avoid this behavior 
> (aside from not having jobs in the Q state :-)?

Does maui have those jobs held?  What does 'checkjob' say about those
jobs?