Ian Wienand <[email protected]> writes: > On 06/19/2014 01:18 AM, James E. Blair wrote: >> (This requires tracking a bit more state across allocation runs). > > This seems to be the crux of the matter; once we have some state all > sorts of things become possible. > > I've made a proposal in which seems like the most simple place to > start; track if a label requested an allocation last time but didn't > get any, and give them preference the next time around. > > This should ensure minimal forward progress? > > It could easily be expanded to track the last N times and round-robbin > more fairly, or be expanded to percentage based, or a PID algorithm as > suggested by clarkb, etc etc. But I just wanted to start with the > most simple thing I could think of. > > If the general idea is OK, I can work on it some more. Particularly > interested in where people are thinking the state should be stored. > > Thanks, > > -i > > [1] https://review.openstack.org/#/c/101110/
Thanks, I've left further comments in the review. -Jim _______________________________________________ OpenStack-Infra mailing list [email protected] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-infra
