[torqueusers] General queue possibilities..Questions...
csamuel at vpac.org
Wed Sep 7 20:26:15 MDT 2005
On Wed, 31 Aug 2005 02:23 am, Dan Roberts wrote:
Here's some information that may help!
> Can anyone point me to some general examples of how queues are created?
These are created with qmgr, here is an example from one of our clusters.
# Create and define queue run_12_hours
create queue run_12_hours
set queue run_12_hours queue_type = Execution
set queue run_12_hours resources_max.walltime = 12:00:00
set queue run_12_hours resources_default.walltime = 12:00:00
set queue run_12_hours enabled = True
set queue run_12_hours started = True
That gives you a queue which will accept jobs up to 12 hours and if you don't
specify a walltime you will get 12 hours.
We use this originally for a surface chemistry commerical package that almost,
but not quite integrates with Torque and doesn't give the users a way to
specify how long their jobs will last, but does give them a list of queues on
> Especially I am interested in queues with different priorities and how
> to use execution and routing queues in concert with each other.
We haven't touched routing queues, so I can't comment there. However, we do
have a queue for short testing jobs called sque that we give priority to, but
this is done in Moab rather than in PBS, so the config there is:
SRCFG[sque] STARTTIME=08:00:00 ENDTIME=20:00:00
SRCFG[sque] PERIOD=DAY DAYS=MON,TUE,WED,THU,FRI
SRCFG[sque] PROCLIMIT<=4 DEPTH=7
SRCFG[sque] TASKCOUNT=2 FLAGS=SPACEFLEX,DEDICATEDRESOURCE
The first part creates a standing reservation of two nodes (4 cpus) 8am-8pm
weekdays and only admits jobs of 4 CPUs or less that will finish in under 15
minutes (the queue defaults to 10 mins).
The second part gives a big boost to jobs submitted in that queue so that
other priority settings shouldn't (hopefully!) block them from running.
> Finally is there a way to automatically change any given queue based on
> server time. In other words if there are now jobs running at 5 pm I open up
> the large queue to all available resources so that large jobs could run over
I don't know about precisely what you ask for, but at another institution I've
configured a node that is usually used for development work during the day
such that it can run jobs overnight using standing reservations. From memory
(I've no network access to the system) I did it by putting a daytime standing
reservation on it restricting access to the root user, who of course cannot
queue jobs in PBS. :-)
Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager
Victorian Partnership for Advanced Computing http://www.vpac.org/
Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20050908/2160dd22/attachment.bin
More information about the torqueusers