[Moabusers] Regarding resources and walltime suspension

Chris Samuel csamuel at vpac.org
Sat May 17 02:32:02 MDT 2008


----- "Mukta Sharma" <mukkasharma at yahoo.com> wrote:

> I have some doubts regarding queues in Torque. I will
> appreciate if any could answer my doubts:

Hi there,

> Is there anyway that torque will identify the
> resources required by the job and user dont have to
> explicitly specify the resources required. For example
> if there is multithreaded job which requires multiple
> CPU at a time so does user have to explicitly define
> the resources or Is there any other way of dealing
> with this.

I don't think that trying to second guess a user is
a good idea.

You could try and parse their PBS script in the submit
filter and look for calls to mpirun or mpiexec, and then
set resource requirements from then, but I still don't
think that's a good plan.

You're still not going to spot OpenMP tasks for instance..

> My second doubt is when the walltime of the job is
> reached then job is killed by the system or torque so
> is there any way to suspend that job instead of
> killing or terminating it and resume it later.

Never tried doing suspend/resume I'm afraid.

> More specifically any way by which the system suspends the
> job and resume it later when the resources are
> avaiable again and user dont have to do this
> explicitly.

But if it has already reached its walltime, why
would you want it to continue ?

Surely the user should be setting the right walltime
themselves already ?

cheers!
Chris
-- 
Christopher Samuel - (03) 9925 4751 - Systems Manager
 The Victorian Partnership for Advanced Computing
 P.O. Box 201, Carlton South, VIC 3053, Australia
VPAC is a not-for-profit Registered Research Agency


More information about the moabusers mailing list