[torqueusers] problem with queue

Thomas Dargel td at chemie.hu-berlin.de
Fri Nov 24 00:21:36 MST 2006


Hi Bibo,

ok, I look again into my definitions, please try to change the
resource_max.nodect to 32.

Here is a snip out of my configuration, it works well (3 'cpu-8'-slots).

set queue cpu-8 queue_type = Execution
set queue cpu-8 from_route_only = True
set queue cpu-8 resources_max.nodect = 24
set queue cpu-8 resources_max.nodes = 8:par
set queue cpu-8 resources_min.ncpus = 8
set queue cpu-8 resources_min.nodect = 8
set queue cpu-8 resources_min.nodes = 8:par
set queue cpu-8 resources_default.ncpus = 8
set queue cpu-8 resources_default.neednodes = par
set queue cpu-8 resources_default.nodect = 8
set queue cpu-8 resources_default.nodes = 8:par

(par is a node property defined in <TORQUEDIR>/server_priv/nodes; v. 1.2.0p6)

I hope that's it,
good luck,

Thomas.

On Thu, Nov 23, 2006 at 05:33:44PM -0200, Albino Aveleda wrote:
> Hi Tomas,
> 
> Unhappily it doesn't work. I don't known what it is wrong.
> 
> Best Regards,
> Bibo
> 
> Quoting Thomas Dargel
> 
> >Hi,
> >
> >give this a try, and add
> >
> >set queue b_16cpus resources_available.nodect = 32
> >
> >to your configuration.
> >
> >Good luck,
> >Thomas.
> >
> >On Tue, Nov 21, 2006 at 10:22:21AM -0200, Albino Aveleda wrote:
> >>Hi,
> >>
> >>I have a cluster with 32 nodes with 1 CPU per node. When I have
> >>submitted two jobs with 16 CPUs (16 nodes), only one running. Although
> >>I have 16 CPUs (nodes) free the second job doesn´t start. It was queued.
> >>
> >>The comment in the "qstat -f <job_id>" command is:
> >>comment = Not Running: Not enough cpus available
> >>
> >>What is wrong in my configuration?
> >>
> >>#
> >># Create and define queue b_16cpus
> >>#
> >>create queue b_16cpus
> >>set queue b_16cpus queue_type = Execution
> >>set queue b_16cpus Priority = 120
> >>set queue b_16cpus max_running = 2
> >>set queue b_16cpus resources_max.ncpus = 16
> >>set queue b_16cpus resources_max.nodect = 16
> >>set queue b_16cpus resources_max.nodes = 16
> >>set queue b_16cpus resources_min.ncpus = 9
> >>set queue b_16cpus resources_default.ncpus = 16
> >>set queue b_16cpus resources_default.nodes = 16
> >>set queue b_16cpus enabled = True
> >>set queue b_16cpus started = True
> >>
> >>Best regards,
> >>Bibo
> >>
> 
> 


More information about the torqueusers mailing list