[Moabusers] Job reservations changing when Moab restarted ?

[Gareth.Williams at csiro.au]

Hi Chris,

I may have been wrong about resetting queue time.  I just restarted moab
and I think the major priority change actually came from resetting the
bypass count - it was clearly not preserved across restarts.

-- Gareth

> -----Original Message-----
> From: Williams, Gareth (HPSC, Melbourne - HPSC)
> Sent: Tuesday, 29 January 2008 11:26 AM
> To: csamuel at vpac.org; moabusers at supercluster.org
> Subject: RE: [Moabusers] Job reservations changing when Moab restarted
?
> 
> Hi Chris,
> 
> I have a related observation/problem. We have RESERVATIONPOLICY
> CURRENTHIGHEST so don't have your specific issue, but we do have a
USRCFG
> MAXIJOB limit to prevent/limit large numbers of jobs from increasing
in
> priority due to queue time or expansion factor.  When moab is
restarted,
> it seems to lose information about jobs having been ineligible to run
and
> so can raise the priority of a number of jobs in an undesirable
manner.
> 
> Both issues are a pain because otherwise restarting moab has minimal
> impact and is all but required for significant config changes and
> obviously for updates.
> 
> -- Gareth
> 
> -----Original Message-----
> From:	Chris Samuel [mailto:csamuel at vpac.org]
> Sent:	Fri 25/01/2008 9:33 AM
> To:	Moab Users
> Cc:
> Subject:	Re: [Moabusers] Job reservations changing when Moab
restarted
> ?
> 
> 
> ----- "Chris Samuel" <csamuel at vpac.org> wrote:
> 
> > We've had some complaints about jobs start times
> > getting pushed back (according to showstart) even
> > though we have RESERVATIONPOLICY HIGHEST which should
> > (if my reading of the manual is correct) mean once
> > they've got a reservation then that's it, it shouldn't
> > change..
> 
> Case in point - a 128 CPU job (submitted on 17th) said on the 18th it
> would start in 2 days but a week later it is still Idle and now says
it
> won't start for almost 6 days. :-(
> 
> [root at tango-m jobs]# showstart -e all 78607
> job 78607 requires 128 procs for 20:00:00
> 
> Estimated Rsv based start in              5:19:28:11 on Thu Jan 31
> 04:58:52
> Estimated Rsv based completion in         6:15:28:11 on Fri Feb  1
> 00:58:52
> 
> Estimated Priority based start in         5:19:28:11 on Thu Jan 31
> 04:58:52
> Estimated Priority based completion in    6:15:28:11 on Fri Feb  1
> 00:58:52
> 
> Estimated Historical based start in       5:19:28:11 on Thu Jan 31
> 04:58:52
> Estimated Historical based completion in  6:15:28:11 on Fri Feb  1
> 00:58:52
> 
> Best Partition: testing
> 
> This is *not* good..  any clues please ?
> 
> --
> 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
> 
> 
> 
> 
>