[torquedev] qsub related question

Lennart Karlsson Lennart.Karlsson at nsc.liu.se
Wed Jun 27 09:30:23 MDT 2007


Sarvin wrote:
> I like to make the nodes ID specified in my script file. I know for a case of one node I simply add:
>  
> #PBS -l nodes=node1 to the script file.
>  
> How about if I want to specify node=1 upto node16.
>  
> I goggled it and found:
>  
> #PBS -l nodes-node1:ppn=1:+node2:ppn=2+node3:ppn=2....
>  
> It is not working, and the job goes to the first node (here node1) and it doesn't look for an empty node. (e..g if node3 is empty)

Your syntax is wrong. Rewrite it as:

#PBS -l nodes=node1:ppn=1+node2:ppn=1+node3:ppn=1....

The job allocates all specified nodes and does not choose among them
as you suggest.

If you want to use a single node of a number of nodes, you may
mark the nodes in your cluster-wide server_priv/nodes configuration file,
e.g. with the name pool1:

node1 np=1 pool1
node2 np=1 pool1
     .
     .
     .
node16 np=1 pool1
node17 np=1 pool2
node18 np=1 pool2
     .
     .
     .

and then specify

#PBS -l nodes=1:pool1

You mention looking for empty nodes. Your job scheduler, e.g. pbs_sched or
Maui or Moab, and its configuration determines where your job goes. It
decides if the job runs on empty or partially filled nodes.

Cheers,
-- Lennart Karlsson <Lennart.Karlsson at nsc.liu.se>
   National Supercomputer Centre in Linkoping, Sweden
   http://www.nsc.liu.se




More information about the torquedev mailing list