On 1/11/2019 9:44 AM, Chris Dent wrote:
I've been saying for a few weeks that "progress continues on gpu-reshaping for libvirt and xen" but it looks like the work at:
* <https://review.openstack.org/#/q/topic:bp/reshape-provider-tree+status:open>
is actually stalled. Anyone have some insight on the status of that work?
This came up in yesterdays placement/scheduler meeting as well. Sylvain was pinged about the libvirt change being stalled waiting for updates from him (he might have lost this change since it's owned by me in gerrit, but I'm not working on it - I'd suggest starring it to keep tabs on it in gerrit). As for the xenapi changes, the original author is no longer working on this and the replacement hasn't picked it up, so also stalled. Without a major push on the reshaper stuff (feature freeze is March 7) those are likely going to be deferred including an (arguably optional for now) hook to do reshapes during fast-forward upgrades (which would unfortunately have to be run on every compute host as the mdev information is on the host, not the DB). For the extraction work, I think the main thing we're looking for (and Sylvain signed up for in Berlin [1]) is to have a functional test which creates a server in a flat provider tree, reshapes the tree to move inventory and allocations to a child provider, and then schedules another server to the same resource class and make sure that all works properly. We have existing functional tests that do the reshape stuff with a fake virt driver, just not tests that hook in the scheduler aspects I don't think. [1] https://etherpad.openstack.org/p/BER-placement-extract -- Thanks, Matt