On Tue, Aug 09, 2016 at 05:33:04AM +0000, [email protected] wrote: > > > > > Hi all, > > > > > > I have an issue regarding slots configuration in SGE queue. > > We have a 2 node cluster with each node having 8 procs. > > We have 4 queues in this cluster and each with differenrt slots > configuration. > > Total slots altogether for all queue is 27 and I am submitting 27 jobs and > not mentioning the queue. > > SGE is itself taking care of this and scheduling jobs and 27 jobs are > going to r state. > > > > My question is How can 27 jobs go to r state although we have total 16 > procs only.
The number of jobs you can run on a grid engine cluster host has nothing to do with the number of processors. The number of jobs you can have running is controlled by slots alone. If you want to stop the scheduler from oversubscribing a node (that is scheduling more jobs/slots to a node than it has processors) then you probably want to configure a slots recource in the complex_values of each exechost. So in this case add slots=8 to the complex_values of each exec host. William
signature.asc
Description: Digital signature
_______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
