On Oct 31 22:23 +0000, Dave Love wrote: > Steve Schmerler <[email protected]> writes: > > OK yes, I saw that it always reserves the requested number of slots, free > > or not. Makes sense. Does the system switch reservation from used to free > > slots if more slots become free (I guess that's default)? > > I don't understand the question, but it will try to make reservations as > early as possible possibly within other constraints. I'm not sure how > accurately it reflects the implementation, but maybe see > <http://arc.liv.ac.uk/repos/darcs/sge/doc/devel/rfe/resource_reservation.txt>.
Since the scheduler always reserves N slots in each scheduling run and N is usually a mix of free and used slots, the question was if slots are automatically reserved if they become free, thereby abandon used slots, but William Hay's response clarified that. Thanks for the document. I didn't know the "Implementation" section (isn't contained in [1]). > > We have two PEs, which differ only in allocation_rule. The PE with > > allocation_rule $fill_up > > does reservation, while > > allocation_rule 8 > > (to request complete 8-core nodes) doesn't. Why may that be? > > If that means reservations are only made in the first PE, we can only > guess, but all the reservations currently in our system are for PEs with > allocation_rule N. We don't have overlaps with $fill_up PEs on the same > hosts, though. Simply nothing happens. No "RESERVING" lines in $SGE_ROOT/$SGE_CELL/common/schedule. The doc [1] doesn't seem to shed much light on that, unfortunately. [1] http://docs.oracle.com/cd/E24901_01/doc.62/e21978/management.htm#autoId26 best, Steve _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
