> Try unsetting > > resources_default.nodes = 1 > > resources_max.nodes = 12 > > resources_min.nodes = 1 > > and using > > resources_min.nodect = 1 > resources_max.nodect = 12 > etc instead.
Set those and restarted both pbs_server and maui. Still same problem with a job that uses 16 threads on one node: 12/23 13:32:04 INFO: job 850431 exceeds requested proc limit (15.86 > 1.00) I also tried setting resources_?.nodes to these before you mentioned completely removing those settings: resources_default.nodes = 1:ppn=1 resources_min.nodes = 1:ppn=1 resources_max.nodes = 12:ppn=16 Still same results, and also tried using RESOURCELIMITPOLICY JOBPROC:ALWAYS:CANCEL and although that didn't cancel this job, it also didn't cancel the job when I requested nodes=1:ppn=1 in the PBS script but had the program use 16 threads, i.e. I don't want users to be able to get away with using more processors than they requested in the PBS file. > How many virtual cpu on compute node? > > cat /proc/cpuinfo Each has 16 processors (dual Nehalem X5570 per each node with HT turned on) > and > cat pbs_mom config I don't understand what you're asking for. The nodes configuration file for the pbs_server has np=16 for each of these compute nodes. The pbs_mom config files just have a bunch of $usecp lines for telling the moms to copy output files directly rather than to use scp or rcp. Thanks, Sabuj Pattanayek _______________________________________________ mauiusers mailing list [email protected] http://www.supercluster.org/mailman/listinfo/mauiusers
