[torqueusers] Time use 0 using qstat command

Justin Finnerty justin.finnerty at uni-oldenburg.de
Fri Nov 2 09:46:36 MDT 2007


On Fri, 2007-11-02 at 01:13 +0100, jafranco at ono.com wrote:
> Hello.
> 
> I have installed torque 2.1.8 in a single linux computer, but I have a
> problem. 
> 
> I can submit jobs using qsub but when I want to know the cpu time,
> using qstat command,
> qstat indicates that the Time Use flag indicates 0 value, in spite of
> the jobs are running for a long time.

I see 0 cpu time when a user's job is running "outside" of the queue
system's direct control.  For example when users of LAM MPI do not use
the torque communication protocol then only the boot application is
timed not the subprocesses.  For LAM MPI the solution is to recompile
the MPI to use libtorque and then get users to use "mpiexec -lamboot
EXE" instead of "mpiexec -machinefile $PBS_NODEFILE -n ?? EXE".  Also
programs that use rsh to start sub-jobs will only show time for the
starting application.  I don't know if using pbsdsh instead of rsh would
get the timings for these subjobs and I have found that pbsdsh is not
sufficiently similar to rsh to act as a direct replacement.

Cheers
	Justin
-- 
Dr Justin Finnerty
Rm W3-1-218         Ph 49 (441) 798 3726
Carl von Ossietzky Universität Oldenburg



More information about the torqueusers mailing list