On Thu, 30 Jul 2015 08:25:05 +0000 "Winkler, Ursula (ursula.wink...@uni-graz.at)" <ursula.wink...@uni-graz.at> wrote:
> > $pe_slots restricts you to a single node so I'm guessing the jobs > > that don't start are jobs that need more than one node. > > Yes, that should be right. > > > While we don't use gepetools we do have a JSV that rewrites > > people's requested PE based on the number What you need I think is > > something that routes jobs that request 1 node to PEs with a > > $pe_slots allocation rule while other jobs are routed to nodes with > > an allocation rule equal to the requested ppn. In all cases the > > number of slots to request should be nodes*ppn. > > The problem is that single-node jobs which require just smaller parts > of the available slots (e.g. 2) are always started on completely free > nodes instead of starting on already busy nodes which have the > requested resources too. Jobs which order more than 1 node but also > few cores/node are typically requiring more than the default of other > resources (e.g. memory) so they are not the problem. And apart from > the 1-node-few-slots trouble the scheduling works pretty satisfying. > My comment was meant to be read in the context of: i)Reuti's earlier suggestion that jobs in PEs with $pe_slots allocation_rule will be scheduled according to the queue_sort_method and load_formula like serial jobs ii)Your response that settting the allocation rule to $pe_slots stops (multi-node) jobs from scheduling. The problem is therefore how to take advantage of (i) without causing (ii) My suggestion was to modify your jsv/gepetools to force single node parallel jobs into PEs with $pe_slots allocation rules (which gives you control over where they are scheduled via queue_sort_method and load_formula) while sending the others to PEs with other (appropriate) allocation rules that won't cause (ii). William
pgpwqzyK4ujbh.pgp
Description: OpenPGP digital signature
_______________________________________________ users mailing list users@gridengine.org https://gridengine.org/mailman/listinfo/users
