[torquedev] reordering of routing queues

David Beer dbeer at adaptivecomputing.com
Thu Feb 24 10:29:42 MST 2011



----- Original Message -----
> ----- Original Message -----
> > On 02/23/2011 03:23 PM, Martin Siegert wrote:
> > > Hi,
> > >
> > > it appears that qmgr reorders route_destinations:
> > > E.g.,
> > >
> > > Qmgr: p s
> > > ...
> > > #
> > > # Create and define queue default
> > > #
> > > create queue default
> > > set queue default queue_type = Route
> > > set queue default route_destinations = ql
> > > set queue default route_destinations += qs
> > > set queue default route_destinations += q1
> > > set queue default enabled = True
> > > set queue default started = True
> > > ...
> > > Qmgr: set queue default route_destinations = q1
> > > Qmgr: set queue default route_destinations += qs
> > > Qmgr: set queue default route_destinations += ql
> > > Qmgr: p s
> > > ...
> > > #
> > > # Create and define queue default
> > > #
> > > create queue default
> > > set queue default queue_type = Route
> > > set queue default route_destinations = ql
> > > set queue default route_destinations += qs
> > > set queue default route_destinations += q1
> > > set queue default enabled = True
> > > set queue default started = True
> > > ...
> > >
> > > I have no idea when this reordering was introduced (under
> > > torque-2.4.8
> > > this does appear to work correctly), but this looks like a bug to
> > > me.
> > >
> > > The order of the route_destinations is of crucial importance since
> > > torque
> > > tries the queues in order and the job gets routed to the first
> > > queue
> > > that
> > > fits. See:
> > >
> > > http://www.adaptivecomputing.com/resources/docs/torque/4.1queueconfig.php#routing
> > > "The ordering of the route_destinations is important. In a routing
> > > queue, a job is assigned to the first possible destination queue
> > > based
> > > on the resources_max, resources_min, acl_users, and acl_groups
> > > attributes."
> > >
> > > Cheers,
> > > Martin
> > >
> 
> I think I know what is going on here. This is a conflict due to some
> code we put in to allow the environment to be overwritten without
> duplicating variables. This was causing crashes because it would
> create job environments that were like 33k characters. Obviously, this
> was an unintended byproduct that will be fixed. I will let you know
> once I figure out a way to make the two work.
> 

Actually, it appears that change doesn't cause it but either way I have reproduced the error with 2.5 and I can look into it.

-- 
David Beer 
Direct Line: 801-717-3386 | Fax: 801-717-3738
     Adaptive Computing
     1656 S. East Bay Blvd. Suite #300
     Provo, UT 84606



More information about the torquedev mailing list