[torqueusers] question regarding compile queues

Jacqueline Scoggins jscoggins at lbl.gov
Mon Nov 7 10:50:00 MST 2005


Doug,

I had a similar request and what I did was placed properties that were
different for the nodes in the node file so that the users would have to
request that particular queue for his jobs to run on.  

i.e.

nodes file would look like this:

node0001 np=4 dualcore
node0002 np=4 dualcore
node0003 np=2 shared
node0004 np=2 shared

Users would have to request in the qsub (-l nodes=2:dualcore or -l
nodes=2:shared)  or within their script   (#PBS -l nodes=2:dualcore or
#PBS -l nodes=2:shared).  PBS will schedule the jobs to the appropriate
nodes because of your node file settings.

I hope this help.

Jackie

On Mon, 2005-11-07 at 09:28, Doug Smith wrote:
> We are currently running Torque 1.2.0.p6 on RedHat Linux cluster. 
> 
> We have set up a compile queue for our Pathscale compilers since we only 
> have a one user license on the head node.
> Users can submit a make/compile job to this queue and not have to worry 
> about license contention for the PathScale compilers.  However, we have 
> run into one little snag, now that the head node is defined in the 
> nodefile it is taking part in computational jobs not just compiling.
> 
> Is there a way to configure a queue in Torque so that jobs submitted to 
> a particular queue run on a specific host and that host does not accept 
> jobs from the general execution queues?
> 
> Thanks
> JDS
> 
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers



More information about the torqueusers mailing list