TargetingNodes

From UF HPC Wiki

Introduction

This document describes how to target nodes with batch jobs in Torque.

Single Nodes

Normally, job scripts will have resource requests like #PBS -l nodes=1:ppn=2 or somesuch.

Sometimes it is useful to target N processors on a particular node (for example, reboot jobs). The following example will requests 4 procs on the node r4a-s42.ufhpc:

-bash-3.00$ qsub -l nodes=r4a-s42.ufhpc:ppn=4 testjob.job
4541814.torx.ufhpc
-bash-3.00$ qstat -u prescott -n

torx.ufhpc: 
                                                                   Req'd  Req'd   Elap
Job ID               Username Queue    Jobname    SessID NDS   TSK Memory Time  S Time
-------------------- -------- -------- ---------- ------ ----- --- ------ ----- - -----
4541814.torx.ufhpc   prescott investor testjob     22490     1   1    --  00:01 R   -- 
   r4a-s42+r4a-s42+r4a-s42+r4a-s42

Multiple Nodes

Targeting multiple nodes with a single job is an extension of the single node syntax. This is useful to do on accasion, as well (benchmarking, for example). This example requests four procs on r4a-s42.ufhpc and another two on r3a-s18.ufhpc:

-bash-3.00$ qsub -l nodes=r4a-s42.ufhpc:ppn=4+r3a-s18.ufhpc:ppn=2 testjob.job
4541816.torx.ufhpc
-bash-3.00$ qstat -u prescott -n

torx.ufhpc: 
                                                                   Req'd  Req'd   Elap
Job ID               Username Queue    Jobname    SessID NDS   TSK Memory Time  S Time
-------------------- -------- -------- ---------- ------ ----- --- ------ ----- - -----
4541819.torx.ufhpc   prescott investor testjob       --      2   1    --  00:01 R   -- 
   r4a-s42+r4a-s42+r4a-s42+r4a-s42+r3a-s18+r3a-s18