On Wed, Mar 21, 2007 at 11:16:27AM +0100, Thomas Dargel wrote: > On Wed, Mar 21, 2007 at 01:21:44PM +1100, Chris Samuel wrote: > > On Tue, 20 Mar 2007, Thomas Dargel wrote: > > > > > There is no walltime limit set, neither at torque/maui nor in the job > > > script. > > > > It certainly used to be the case that if you restarted Maui with jobs that > > didn't have wall times set then it would kill them all because they had > > exceeded their default of 0 hours, 0 minutes and 0 seconds. > > > > Could this be what happened to you here ? > > > > cheers, > > Chris > > -- > > Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager > > Victorian Partnership for Advanced Computing http://www.vpac.org/ > > Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia > > > > Hi Chris, > > exactly this issue was posted by me in november 2005 > (http://www.clusterresources.com/pipermail/mauiusers/2005-November/001825.html). > Garrick provided a patch for this maui version (3.2.6p14-snap1129921819) and > the cancelation of running jobs without a given walltime by maui was solved. > > By the way, this restart/cancel phenomenon is gone with maui revision > 3.2.6p18-snap1164120675. > > At the end of this thread you predicted the dead of the jobs after 100 days. > (http://www.clusterresources.com/pipermail/mauiusers/2005-November/001891.html). > > This is now prooved. There was no restart of maui, there is 'only' the > interpretation of 'infinity' in maui, which means INFINITY=100:00:00:00 . > > Does anybody know where this definition is made in the sources of maui???? >
I think I found the reason in include/msched.h: line 341: #define MDEF_SYSJOBWCLIMIT 8639999 8639999 [s] = 99:23:59:59 I will try to increase this number and recompile the code... ;-P Thomas. > Thank you for your attempt to help > best regards > > Thomas Dargel. > _______________________________________________ mauiusers mailing list [email protected] http://www.supercluster.org/mailman/listinfo/mauiusers
