[torqueusers] RE: scheduling bug?

[Alexander Saydakov]

I have just seen that again. Looks like there is some randomness in
selecting jobs despite of 'sort_by: high_priority_first'. Most of the time
it works as expected, but sometimes picks a wrong job. I don't specify any
requirements like memory or disk space.

 

Note those 2 jobs running in p2 queue and one job in p3 queue. They have
just started despite of the pile of jobs in p1 queue.

 

> qstat -Q

Queue            Max Tot Ena Str Que Run Hld Wat Trn Ext Type

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

test               0   0 yes yes   0   0   0   0   0   0 Execution

p1                 0 102 yes yes 902 121   0   0   0   0 Execution

p2                 0 607 yes yes 605   2   0   0   0   0 Execution

p3                 0 610 yes yes 609   1   0   0   0   0 Execution

 

 

-----Original Message-----
From: Alexander Saydakov [mailto:saydakov at yahoo-inc.com] 
Sent: Tuesday, January 24, 2006 2:24 PM
To: 'torqueusers at supercluster.org'
Subject: scheduling bug?

 

I am running troque-2.0.0p5 with default scheduler. The only thing I changed
in sched_config is:

#sort_by: shortest_job_first    ALL

sort_by: high_priority_first    ALL

 

I have just seen that a few jobs from the test queue (priority 0) started
despite of a bunch of jobs waiting in the p1 queue (priority 100).

All jobs in the test queue have been submitted after jobs in the p1 queue.

 

> qstat -Q

Queue            Max Tot Ena Str Que Run Hld Wat Trn Ext Type

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

test               0 605 yes yes 603   2   0   0   0   0 Execution

p1                 0 320 yes yes 214 106   0   0   0   0 Execution

 

I wonder how it is possible.

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20060126/c0dba633/attachment.html