Jeremy, good day. Thu, Jun 26, 2008 at 09:26:15AM -0500, Jeremy Mann wrote: > What we want is to load up our 3 clusters with low priority, niced 19 jobs > and let them run to completion. These jobs take anywhere from a few days > to a week to complete. The problem is that I have an over demanding boss > (and users) who want their jobs to run now, regardless of what is already > queued and running. They can not and will not wait a few days (or weeks) > for their jobs to complete. > > So my solution was the 2nd installation of Torque/Maui. If there is a way > to trick Torque/Maui into running multiple jobs at the same, I would like > to hear it.
May be you should try to add 'np=X' to the end of the node description in <PBS>/server_priv/nodes, telling Torque that this particular node has X CPUs. Next thing to do is to play with $ideal_load and $max_load in the MOM's configuration file to tell Torque what uptime should be treated as the "Busy" ones. May be you'll need to adjust Torque's server and queue variable resources_available.nodect to make Torque believe that you have large number of CPUs. But this will be needed if users will submit jobs for more CPUs that the number of nodes in your cluster. If your users run serial jobs, this shouldn't be needed. -- Eygene Ryabinkin, Russian Research Centre "Kurchatov Institute" _______________________________________________ mauiusers mailing list [email protected] http://www.supercluster.org/mailman/listinfo/mauiusers
