[Moabusers] queue to node mapping

[Douglas Wightman]

Because this is the moabusers mailing list I'm going to suggest a moab
based solution.

Take a look at REMAPCLASS:

http://www.clusterresources.com/products/mwm/docs/3.5credoverview.shtml#remapclass

It appears as though the routing queue in TORQUE is not set up
correctly.  You could get help on that using the torqueusers mailing
list or use the REMAPCLASS instead of the TORQUE routing queue.

Once the REMAPCLASS is setup the CLASSCFG[] HOSTLIST= will function as
you expect.

- Douglas

On Fri, 2007-09-28 at 11:42 -0700, Steven DuChene wrote:
> Nope, did not work. Submitted job specifiying ia64 still went to 
> batchx86 queue instead of the batchia64 queue.
> 
> I added those lines to the moab.cfg file and then restarted moab.
> As my test user I did:
> 
> qsub -l nodes=1:ia64 myjob_ia64.pbs
> 
> where myjob_ia64.pbs looks like:
> 
> #PBS -S /bin/bash
> #PBS -m ae
> 
> #! Name of job
> #PBS -N Test_Serial_ia64.1
> 
> ## --------------------  PREAMBLE -----------------------------------
> export RUN_HOST=`hostname`
> n=`wc -l < $PBS_NODEFILE`
> cd $PBS_O_WORKDIR
> ## --------------------  END PREAMBLE --------------------------------
> 
> echo 
> "================================================================================"
> echo "."
> echo ".            hybrid computing single processor BATCH JOB"
> echo ".            -------------------------------------------"
> echo "."
> echo ". Job Running on queue: $PBS_QUEUE"
> echo "."
> echo ". Running on HOST:    $RUN_HOST"
> echo ". Allocated processor: $n"
> echo ". Working Dir:  `pwd`"
> echo "."
> echo 
> "================================================================================"
> echo .
> 
> 
> Wightman wrote:
> > You could do something like:
> >
> > CLASSCFG[batchx86] HOSTLIST=oscarnode[1-4]
> > CLASSCFG[ia64] HOSTLIST=<ia64_nodes>
> >
> > You can also use the REMAPCLASS feature in moab to route jobs.
> >
> > Standing reservations won't help if the job is routed to a queue with no
> > ia64 nodes.
> >
> > - Douglas
> >
> >   
> >> I have a mixed architecture cluster and I want to have queues that are
> >> constrained to run jobs
> >> on a particular type of machine architecture. In torque I have two
> >> execution queues and a
> >> routing queue. Each of the execution queues has nodes of a particular
> >> architecture tied to it
> >> using the acl_hosts directives:
> >>
> >> set queue batchx86 queue_type = Execution
> >> set queue batchx86 acl_host_enable = False
> >> set queue batchx86 acl_hosts = oscarnode4
> >> set queue batchx86 acl_hosts += oscarnode3
> >> set queue batchx86 acl_hosts += oscarnode2
> >> set queue batchx86 acl_hosts += oscarnode1
> >>
> >> and then I tried sending jobs to the routing queue with a node
> >> architecture defined like:
> >>
> >> qsub -q route -l nodes=1:x86_64 myjob.pbs
> >>
> >> This seems to work ok for the x86_64 architecture but when I submit a
> >> ia64 job it still ends up
> >> in the x86_64 queue and then stalls because there are no ia64 nodes
> >> available in that queue.
> >>
> >> I think I did something like this a few years ago with standing
> >> reservations but I do not recall
> >> the exact details. I have read through section 12.1.3.1 where it talks
> >> about doing this but
> >> maybe there is something I am missing.
> >>
> >> Any guidance or hints are appreciated.
> >> --
> >> Steven DuChene
> >>
> >> _______________________________________________
> >> moabusers mailing list
> >> moabusers at supercluster.org
> >> http://www.supercluster.org/mailman/listinfo/moabusers
> >>
> >>     
> >
> >   
>