We are also experimenting with hard/soft limits. Some suggestions: - most of you cluster nodes are busy so jobs might be memory-bound & not processor-bound; try MAXPE instead of MAXPROC - try using hard/soft limits for USERCFG instead of GROUPCFG (ie. create a testuser) - can you paste the output of "diagnose -g" & "diagnose -u" ?
Good Luck! Chris Hunter Yale HPC [email protected] > ---------------------------------------------------------------------- > > Message: 1 > Date: Mon, 28 Jun 2010 17:29:08 +0200 > From: Arnau Bria <[email protected]> > Subject: [Mauiusers] soft and hard limits question > To: [email protected] > Message-ID: <[email protected]> > Content-Type: text/plain; charset=US-ASCII > > Hi all, > > We'd like to implement hard/soft limit in our cluster. > > Our farm is quite empty and some user from a minor experiment sent 300 > jobs but his group had a limit of 50 PROCS. > > Now, we'd like to let that user use 300 PROCS on this lightly farm load, > so I set something like: > > GROUPCFG[group] FSTARGET=10 QDEF=qos1 MAXPROC=50,300 > > *no MAXPROC Limits for qdef. > > Now I have this scenario: > > # qstat -u nemo004|grep -c R > 121 > # qstat -u nemo004|grep -c Q > 136 > > but farm is still at 50%: > > 905 Active Jobs 905 of 1788 Processors Active (50.62%) > 221 of 224 Nodes Active (98.66%) > > That not what I've undestood from > http://www.clusterresources.com/products/maui/docs/6.2throttlingpolicies.shtml#limits > so, could anyone explain me why not all the jobs are in R state? > > Also, at checkjob's output: > > job cannot run in partition DEFAULT. (job 10930203 violates active SOFT > MAXPROC limit of 50 for group group (R: 1, U: 121) > > what's the meaning of R: 1 U:121? > *U:121 are runnign jobs by user? > > TIA, > Arnau _______________________________________________ mauiusers mailing list [email protected] http://www.supercluster.org/mailman/listinfo/mauiusers
