[Mauiusers] Suspend/resume jobs

Angel de Vicente angelv at iac.es
Wed Aug 31 08:42:13 MDT 2005


Hi,

we use Maui/Torque 3.2.6/1.2.0, and I'm trying to get the suspend/resume stuff
working, but I have a small problem...

OK, I have applied the patch by Bernward Platz
http://www.supercluster.org/pipermail/mauiusers/2004-July/001284.html

and it looks like suspend/resume is more or less OK (the jobs gets suspended, as
far as Maui and Torque are concerned, but they actually continue using CPU, but
I could live with that).

But apparently it will only suspend jobs if the number of processors for the
PREEMPTOR job is equal to that of the PREEMPTEE, which is not great. If I have a
8 CPU job running and then I submit a 1 CPU job in the high QOS, I would like to
be able to suspend whatever it takes to get this high priority job running (I
thought I could accomplish this with the RESTARTPREEMPT flag). 

Anyone got something like this working? Any tips?

Thanks,
Angel de Vicente

------------------------------------------------
The relevant parts of the maui.cfg look like this:


#Job policy for IAC

QUEUETIMEWEIGHT        1
XFACTORWEIGHT          100

RESWEIGHT              200
PROCWEIGHT             10

FSWEIGHT               100
FSGROUPWEIGHT          10
GROUPCFG[high]         FSTARGET=37.5+

FSPOLICY              DEDICATEDPS
FSDEPTH               7
FSINTERVAL            86400
FSDECAY               0.50

USERCFG[DEFAULT]      MAXPS=2419200
USERCFG[DEFAULT]      QLIST=high

BACKFILLPOLICY        FIRSTFIT
RESERVATIONPOLICY     CURRENTHIGHEST
RESERVATIONDEPTH      8

NODEALLOCATIONPOLICY  MINRESOURCE

JOBNODEMATCHPOLICY    EXACTNODE

PREEMPTPOLICY  SUSPEND

QOSCFG[high] QFLAGS=PREEMPTOR:RESTARTPREEMPT
QOSCFG[high]  MAXWC=600

QOSCFG[DEFAULT]  QFLAGS=PREEMPTEE


-- 
----------------------------------
http://www.iac.es/galeria/angelv/

PostDoc Software Support
Instituto de Astrofisica de Canarias



More information about the mauiusers mailing list