[torqueusers] reading the "Time Use" in qstat

[David Golden]

On Wednesday 13 September 2006 10:47, bill wrote:

> the CPU consumption is stable, around 90% all the time. So why cput
> doesn't grow up? The job is running on two nodes.
>
> Thanks for any help

Okay, another possibility (really a FAQ): if you're running a parallel job, 
what launcher are you using?  You need to launch your parallel processes
using a program that speaks to the PBS TM api. This includes (non-exhaustive
list, probably): pbsdsh included with torque, OSC mpiexec for use with mpich 
and derivatives (n.b. usable for non-mpi parallel launching too), 
openrte/mpiexec from Open MPI, LAM  mpiexec. 

But NOT: mpirun from mpich and derivatives, mpiexec from mpich2 and 
derivatives, ssh.  They effectively bypass the batch system, and the
batch system cannot therefore track their CPU usage. 

You might want to scan through the wiki for torque, e.g.:
http://www.clusterresources.com/wiki/doku.php?id=torque:7.1_mpi_support