[Moabusers] Running a moab simulation
Jared Mecham
jmecham at clusterresources.com
Mon Oct 16 12:20:31 MDT 2006
Corby,
Try adding to the moab.cfg file the following line:
SIMAUTOSHUTDOWN FALSE
If there is a problem with either the resource or workload trace file
then Moab will automatically shut down. This parameter prevents Moab
from shutting down. Once Moab no longer shuts down you should be able
to figure out if the resource or workload trace file is causing the
problem (using showq and checkjob).
Thanks,
-Jared
On Sat, 2006-10-14 at 14:50 -0700, Corby wrote:
> I've been messing around with this for awhile now, using the example
> for the workload that you gave, but Moab is still doing what it's
> always been doing, i.e. exiting right away. The logs and stats don't
> show it did anything ( e.g ".10/14 12:24:51 INFO: scheduling
> complete (queues are empty) on iteration 0"). This is the same
> whether I set the simulation to run now, or set the time to be in the
> past (I've been using around 12:30 AM Jan 1, 2000) and update the
> times in the workload trace accordingly.
>
> I feel like I've tried almost every variable to this problem, and it's
> always the same. Why would Moab just keep exiting right away when
> doing a simulation?
>
> -Corby Ziesman
>
> On 10/9/06, wightman <wightman at clusterresources.com> wrote:
> In simulation Moab only looks for JOBEND events from the
> workload trace
> file. You should populate the workload trace with JOBEND
> events only.
>
> - Douglas
>
> On Mon, 2006-10-09 at 12:40 -0700, Corby wrote:
> > So when I want to run a simulation, do I just need a
> JOBCREATE and
> > then Moab will do a JOBSTART and JOBEND as it would if it
> were
> > actually running?
> >
> >
> >
> >
> > On 10/9/06, wightman < wightman at clusterresources.com> wrote:
> > Workload traces are generated automatically by Moab
> during
> > normal
> > execution. If you take a look in the "stats"
> directory you
> > will see
> > "event" files. Here is what a valid entry looks
> like:
> >
> > 10:33:39 1160411619 job 754 JOBEND
> 1 1
> > wightman
> > wightman 600 Completed [batch:1] 1160411014
> 1160411014
> > 1160411014
> > 1160411619 - - - >= 0M
> > >= 0M -
> > 1160411014 1 0 -:- [RESTARTABLE] -
> > - - 0
> > 163.56 pbs 1 0M 0M 0M
> 0
> > 2140000000
> > maka pbs - - [DEFAULT] - - 0.00 - - - - -
> >
> > (all one 1 line)
> >
> > - Douglas
> >
> > On Mon, 2006-10-09 at 12:35 -0700, Corby wrote:
> > > I was trying to follow the format from
> > >
> >
> http://www.clusterresources.com/products/mwm/docs/16.4.0simulations.shtml but realized quickly something was wrong. The problem is that I couldn't figure out what exactly. There are so many fields so there's a lot of variables when trying to figure out what part of it was wrong.
> > >
> > > I wasn't sure how the workload trace should
> work. In
> > particular I
> > > wasn't sure if I needed to do jobcreate and then
> jobstart in
> > order to
> > > make moab actually simulate the execution of a
> program.
> > >
> > > Because of the setup right now that I'm working
> with, I
> > wasn't able to
> > > simply have Moab output some accounting records.
> > >
> > > -Corby Ziesman
> > >
> > >
> > >
> > > On 10/9/06, wightman <
> wightman at clusterresources.com> wrote:
> > > The workload trace entries are
> incorrect. Where did
> > you
> > > obtain these
> > > records? Did Moab output them?
> > >
> > > - Douglas
> > >
> > > On Wed, 2006-10-04 at 14:52 -0700, Corby
> wrote:
> > > > I'm trying to run a simulation on moab,
> but every
> > time I run
> > > moab it
> > > > just exits right away and the job I
> scheduled in
> > my workload
> > > trace
> > > > doesn't seem to have run.
> > > >
> > > > I'm not quite sure how to set up the
> resource and
> > workload
> > > traces, and
> > > > what fields are essential to getting a
> simulation
> > to
> > > actually run at
> > > > all (I can then worry about making it
> reflect the
> > hardware
> > > and
> > > > everything I want to simulate later, but
> first I
> > just need
> > > it to
> > > > work).
> > > >
> > > > My workload trace just has these two
> entries:
> > > >
> > > > 00:00:20_01/01/00 946710020 job
> 1234
> > jobcreate 1 1
> > > root root
> > > > 86400 \
> > > > Completed [batch:1] 946710010 0
> 0 ethernet
> > x86_64
> > > Linux >= 0
> > > > >= 0 \
> > > > - 0 1 -1 - -
> > - /opt/moab/tracefiles/testprogram - -1
> > > 0
> > > > [DEFAULT] \
> > > > 1 0 0 0 0 0 - - - - - - - 0.0 -
> - - - - -
> > > > 00:00:30_01/01/00 946710030 job
> 1234
> > jobstart 1 1
> > > root root
> > > > 86400 \
> > > > Completed [batch:1] 946710010 0
> 0 ethernet
> > x86_64
> > > Linux >= 0
> > > > >= 0 \
> > > > - 0 1 -1 - -
> > - /opt/moab/tracefiles/testprogram - -1
> > > 0
> > > > [DEFAULT] \
> > > > 1 0 0 0 0 0 - - - - - - - 0.0 -
> - - - - -
> > > >
> > > > and my resource trace just has these:
> > > >
> > > > COMPUTENODE AVAILABLE 1 Node001
> PBS 27580
> > 6442
> > > 100000 2 -1 -1
> > > > -1 Linux \
> > > > x86_64 [NONE] [batch:1]
> [ethernet] 1.0
> > [NONE] [NONE]
> > > [NONE]
> > > > COMPUTENODE AVAILABLE 1 Node002
> PBS 27580
> > 6442
> > > 100000 2 -1 -1
> > > > -1 Linux \
> > > > x86_64 [NONE] [batch:1]
> [ethernet] 1.0
> > [NONE] [NONE]
> > > [NONE]
> > > >
> > > > and my moab.cfg:
> > > >
> > > >
> > SCHEDCFG[Moab] SERVER=xxxxxxxxxxxxxxx:xx
> > > > MODE=SIMULATION
> > > >
> ADMINCFG[1] USERS=root,root
> > > >
> > > >
> > SIMRESOURCETRACEFILE tracefiles/resource.trace
> > > >
> > SIMWORKLOADTRACEFILE tracefiles/workload.trace
> > > > SIMSTOPITERATION 0
> > > > SIMSTARTTIME 00:00:00_01/01/00
> > > > SIMSTOPTIME 12:00:00_01/01/00
> > > >
> > > > RMCFG[base] TYPE=PBS
> > > >
> > > >
> > > > Are there any obvious mistakes I'm
> making?
> > > >
> > > > Thank you for any time you can spare to
> help
> > someone new to
> > > Moab,
> > > > -Corby Ziesman
> > > >
> _______________________________________________
> > > > moabusers mailing list
> > > > moabusers at supercluster.org
> > > >
> >
> http://www.supercluster.org/mailman/listinfo/moabusers
> > >
> > >
> >
> >
>
>
> _______________________________________________
> moabusers mailing list
> moabusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/moabusers
More information about the moabusers
mailing list