Am 20.09.2011 um 16:54 schrieb Vic: > >> unfortunately there is only one policy per GridEngine installation. > > Bah. > > Well, at least I'm not being a total idiot this time :-)
To cite myself: it's often hard to have different policies for different parts of the cluster. This would mean in the essence to have one scheduler running for each part of the cluster. For now this is not foreseen. http://comments.gmane.org/gmane.comp.clustering.gridengine.users/20831 >> How do you submit the jobs to the queuing system? You request the queues >> explicitly or request resources? > > Yes. Queues are set up by function (and owner, as it happens). Jobs of > certain types are committed to certain queues via the various tools we've > got running on top of GE. What you can do: run more than one qmaster and more than one execd per machine, you just need additonal ports for the qmaster and execd. As long as each type of user stays inside his "world" all can be fine (they have to source different setttings.sh with the adequate ports). Well, you have to adjust the overall slot count per machine to avoid oversubscription in case you have more than one queue per machine, maybe even dynamically by a script adjusting an RQS in each cluster instance via a cron job. -- Reuti >> All users are using all queues? > > Not yet - but there's a certain amount of mobility coming in. > > Vic. > > _______________________________________________ > users mailing list > [email protected] > https://gridengine.org/mailman/listinfo/users _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
