Hi, > Am 18.05.2017 um 20:35 schrieb Carlos Madrid-Aliste <[email protected]>: > > I have a computer having 2 CPU and 56 cores in total. A month ago I installed > son of gridengine 8.1.9 and configured one queue (all.q). I was able to > submit and run 56 concurrent jobs in all.q with no problem. > > Yesterday, I defined a PE (smp) with 16 slots and I was able to run > multi-threads programs on it. > > After the PE was configured and assigned to all.q I noticed a weird > behavior; every time that I submit regular jobs to all.q I can run just 40 > jobs simultaneously instead of 56. ( like used to be). Why is this happening? > It looks like 16 cores are exclusively reserved for the PE and nobody is able > to run jobs on them, unless qsub -pe smp 16 shell.sh is explicitly specified? > > How can I run 56 concurrent jobs when the PE is not being used? Is there any > way to do that?
This shouldn't be (unless you are using features like advance reservation or backfilling) – but not in the default installation. The slot setting in the PE has indeed the task to limit the slots, which can be used in this PE across all machines and queues. Hence you could alos submit up to two time 8 cores. Can you please post the output of `qstat -f` and `qalter -w v <JOB_ID>` of the first waiting job. -- Reuti _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
