Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2021-01-21
11:21:35 sean-k-mooney ok ill do it in a patch on top
11:26:00 openstackgerrit sean mooney proposed openstack/nova-specs master: [FUP] remove final reference to trait in vdpa spec https://review.opendev.org/c/openstack/nova-specs/+/771796
11:26:37 stephenfin lyarwood, gibi: when live migrating on API microversion >= 2.25, is there any reason to set block_migration to any value *except* auto?
11:26:46 stephenfin i.e. to explicitly request block or shared migration
11:27:07 sean-k-mooney stephenfin: not that i no of
11:27:47 stephenfin okay, so would it make sense to simply ignore the '--block-migration' and '--shared-migration' flags of 'openstack server migrate --live-migration' on suitable microversions
11:27:51 stephenfin with a log, of course
11:38:17 lyarwood stephenfin: I guess as a workaround when auto doesn't detect that things either are or are not shared?
11:38:40 stephenfin Does that happen?
11:38:58 lyarwood I've never seen it, just playing devils advocate
11:39:17 stephenfin fair. I've gone with an overrideable default so we should be good
11:39:31 lyarwood I guess that's why we allow three values instead of just dropping the param and always going auto under the covers
11:39:46 lyarwood which FWIW I'd be fine with
11:39:51 lyarwood and if that fails raise a bug
11:40:03 lyarwood but that's another microversion
11:43:02 sean-k-mooney i think we just did not remove it because we need the code for the lower microverions so did not see the point
11:43:18 lyarwood yeah that's fair
11:44:00 sean-k-mooney stephenfin: you should not eed to pass --block-migration or --shared-migration however when using --live-migration i think
11:44:31 sean-k-mooney e.g. if osc requires either today we should dorp that requirement with microversion >=2.25
12:30:13 gibi stephenfin: defaulting to auto works for me
12:30:41 stephenfin Cool. That's here https://review.opendev.org/c/openstack/python-openstackclient/+/771801
12:31:14 stephenfin Along with what I hope to be the remainder of the feature gap closing patches for OSC. My doc of novaclient gaps is now empty save for things I don't think we need
12:37:29 gibi stephenfin: ack, will try to get to it
12:37:42 gibi sean-k-mooney: I approved the vdpa spec and the fup. but I have a question
12:38:03 gibi sean-k-mooney: do you have some link to a ticket that tracks the live migration support for vdpa in QEMU?
12:38:31 gibi (suddenly my downstream product mgr got really interested in all the vdpa stuff)
12:40:04 sean-k-mooney ah good question. there are mail threads on that topic let me see if we have something in bugzilla. i get periodic updates form our qemu folks but not directly tracking it myself
12:40:21 sean-k-mooney also thanks :)
12:41:28 sean-k-mooney gibi: by the way vdpa will support block devices at some point to
12:41:43 sean-k-mooney so that will evenruatly become and rfe
12:42:04 sean-k-mooney it provides a way to offload thinks like nvme over rdma to the nic and pass it though to the guest
12:42:17 gibi sounds interesting
12:42:33 sean-k-mooney but also provides a way to take a local ssd and use sriov/vdpa to devide it up
12:43:16 sean-k-mooney that the qemu support for that https://bugzilla.redhat.com/show_bug.cgi?id=1886123 still looking for the migration
12:43:17 openstack bugzilla.redhat.com bug 1886123 in qemu-kvm "Supporting vDPA block in QEMU" [Medium,New] - Assigned to sgarzare
12:44:09 sean-k-mooney gibi: thre will be dpdk support for vdpa eventually too
12:44:56 gibi yeah, I found a slideset about that ^^
12:45:59 openstackgerrit Merged openstack/nova-specs master: libvirt: support vdpa kernel networking https://review.opendev.org/c/openstack/nova-specs/+/764999
12:46:03 openstackgerrit Merged openstack/nova-specs master: [FUP] remove final reference to trait in vdpa spec https://review.opendev.org/c/openstack/nova-specs/+/771796
12:46:15 sean-k-mooney they are currenlty reworking the original vdpa support that was protype wiht dpdk to work with what landed in the kernel
12:47:59 sean-k-mooney gibi: im not imidetly seeing the live migration tracking bz so ill ask our pm where its being tracked and ill let you know
12:48:07 gibi sean-k-mooney: thanks
13:00:23 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Load and cache volume drivers on-demand https://review.opendev.org/c/openstack/nova/+/741545
13:00:24 openstackgerrit Lee Yarwood proposed openstack/nova master: tests: Posion nova.objects.fields.Architecture.from_host https://review.opendev.org/c/openstack/nova/+/769920
13:00:24 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Stop NVMe and ScaleIO unit tests from calling os-brick https://review.opendev.org/c/openstack/nova/+/771806
13:01:22 lyarwood gibi: ^ went down a rabbit hole for a while with the ppc64 stuff but I think https://review.opendev.org/c/openstack/nova/+/741545 is required now.
13:02:56 gibi lyarwood: will look at it after my call
13:03:04 lyarwood gibi: ack thanks
13:03:12 gibi I thank you
13:03:43 lyarwood stephenfin: https://review.opendev.org/c/openstack/nova/+/771711 - not sure if you saw my ping about this last night btw, finally copied some notes from Matt into our docs
13:16:57 brinzhang sean-k-mooney, gibi: https://review.opendev.org/c/openstack/tempest/+/769120 this patch based on https://review.opendev.org/c/openstack/tempest/+/769857, but why the fist patch still run the stable branch zuul job?
13:17:05 brinzhang do you have time to check this?
13:18:51 brinzhang jobs tempest-full-py3 , tempest-full-victoria-py3, tempest-full-ussuri-py3, and tempest-full-train-py3 already failed, because of https://review.opendev.org/c/openstack/nova/+/768852
13:19:48 brinzhang and another issue is the nova patch and tempest patch depends-on each other, I have no idea to deal them :(
13:21:45 brinzhang if you have time to help, please left comment or said in irc, I will back home now, thanks
13:26:50 sean-k-mooney why are you skiping the usage tests?
13:28:17 sean-k-mooney tempest is ment to be branchless right and work with any version of openstack
13:30:14 sean-k-mooney i think what we need to do is update the tempest tests such that they use the old microversion when doing usage checks
13:30:32 sean-k-mooney not skip
13:30:36 gibi yepp ^^ commented the same thing
13:31:56 sean-k-mooney brinzhang: that will break the cylic depency as the tempest change is not depent on the nova one then
14:09:44 elod lyarwood: as I added some info to Nova Meeting's stable section (feel free to extend it :)) I realized, that rocky is still blocked. could you have a look at this easy patch some time to unblock it? >>> https://review.opendev.org/c/openstack/nova/+/766492
14:10:09 lyarwood elod: ack will do
14:10:34 elod lyarwood: thx in advance :]
14:14:50 stephenfin lyarwood: I did but hadn't got to it yet. Looking
14:15:14 lyarwood stephenfin: ack sorry didn't mean to spam
14:15:24 stephenfin nope, all good
14:40:15 openstackgerrit Dan Smith proposed openstack/nova master: Try harder to exclude encrypted volumes https://review.opendev.org/c/openstack/nova/+/771822
14:40:40 dansmith lyarwood: we're also suddenly running cinder encrypted tests in that multistore job despite the blacklist regex, which are coming from the cinder tempest plugin
14:40:46 dansmith lyarwood: ^ this tries to hit them too
14:43:04 sean-k-mooney hehe i like the title
14:43:25 dansmith I dunno why that has started failing all the sudden,
14:43:32 dansmith like if a name changed and we started running those or something
14:43:46 dansmith but they end up putting an image int he wrong glance repo, which doesn't work
14:44:46 dansmith oh, heh
14:44:57 sean-k-mooney https://github.com/openstack/cinder-tempest-plugin/commit/5ab3fef76283ca91b92fa130ba2685886bd8af68
14:44:58 dansmith looks like there's already something in gate
14:45:00 sean-k-mooney maybe that
14:45:10 dansmith sean-k-mooney: yuuup
14:46:08 sean-k-mooney so ya just new test added
14:47:38 dansmith lyarwood: sorry, just saw yours from 1am.. I had half figured this out last night but ran out of time, thanks for grabbing that
14:47:52 dansmith that makes three distinct and legit non-nova, non-glance issues with that job in a week :)
14:53:12 sean-k-mooney well at least there is now testing for this so its still progress
14:53:26 sean-k-mooney even if there are teething issues
14:59:12 dansmith lyarwood: I assume that snuck in because we run the whole tempest-cinder-plugin on that job, but they don't gate on it.. so my first thought was just to not load their tempest plugin in there to prevent this from happening in the future.. thoughts on that?
14:59:44 dansmith sean-k-mooney: yes, especially good when they're actual problems and not just flaky or unreliable tests
15:03:19 lyarwood dansmith: morning, sorry was afk
15:03:51 lyarwood dansmith: yeah correct, I was going to see if we could add this job into their gate tbh
15:03:56 dansmith lyarwood: np, it took a night of sleep to realize that test was from a tempest plugin, thanks for nailing that whilst I slept
15:04:02 dansmith lyarwood: that also works
15:04:50 lyarwood so what's the deal with these project specific tempest plugins
15:05:01 lyarwood from what I've seen most of this should live in core tempest anyway
15:05:03 dansmith I dunno, asking for trouble :)
15:05:21 openstackgerrit Balazs Gibizer proposed openstack/nova master: Replace blind retry with libvirt event waiting in detach https://review.opendev.org/c/openstack/nova/+/770246
15:05:29 lyarwood hehe
15:05:52 lyarwood I just don't get the need for loads of different test repos all doing the same time tbh, I'll go away and read up
15:05:54 dansmith without co-gating it's just an attack vector
15:12:56 lyarwood dansmith: multistore attempts to cover some edge usecases right?
15:21:06 dansmith lyarwood: the whole arrangement in devstack is edgy yeah
15:21:49 lyarwood dansmith: cool, so looking at the actual test that borked here this might be something we can fix on their side

Earlier   Later