Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-08-24
15:10:18 mriedem rabel: you can yell at sean-k-mooney about that
15:14:27 dtantsur dansmith: see vdrok's comment above ^^^. is it going away with resource classes? especially wrt "claim happens on compute"?
15:14:41 sean-k-mooney haha well its not run by our team but what was the questing releating to the pci ci
15:15:17 rabel the question was just if i can ignore if it fails. ;)
15:15:17 dansmith dtantsur: yep
15:15:33 dtantsur good to know
15:15:54 sean-k-mooney ah well yes and no. if you are not touching the pci code probably if you are you might want to take a look at it
15:16:12 sean-k-mooney the pci ci is testing a very limited scope
15:18:17 openstackgerrit Eric Fried proposed openstack/nova master: docs: Document the scheduler workflow https://review.openstack.org/475810
15:18:19 sean-k-mooney we had planned to combine it with the nfv ci and expand the scope to introduce livemigration and other testing but we lost 2 headcount including our ci lead so that did not go ahead
15:18:59 dtantsur dansmith: why aren't compute_group_opts here https://github.com/openstack/nova/blob/master/nova/conf/compute.py#L1122-L1130 ? I think it causes this option to not appear in the sample conf
15:19:39 dansmith dtantsur: I dunno what sort of black magic is involved in all this.. maybe ask mriedem or sfinucan or edleafe
15:20:07 dtantsur ack
15:22:14 dtantsur I also see compute_driver option twice - in [DEFAULT] and in [compute], wut?
15:23:50 mriedem yeah that looks like a bug, the former i mean
15:24:51 dtantsur mriedem: well, actually the option is in [DEFAULT]. lemme bake a quick patch, you'll see what I mean
15:25:17 efried mriedem dtantsur L1141 should be compute_group_opts, not compute_opts.
15:25:32 dtantsur this is what I'm fixing right now, yes :)
15:25:41 efried Cool
15:26:25 efried Looks like Intel PCI CI is out of disk space or something. All runs failing, log links 404.
15:26:39 openstackgerrit Dmitry Tantsur proposed openstack/nova master: Fix sample configuration generation for compute-related options https://review.openstack.org/497460
15:26:42 dtantsur efried, mriedem ^^^
15:27:18 efried dtantsur +1, good catch.
15:27:55 openstackgerrit Eric Fried proposed openstack/nova master: Monkey patch the blockdiag extension https://review.openstack.org/476159
15:30:34 openstackgerrit Eric Fried proposed openstack/nova master: Add formatting to scheduling activity diagram https://review.openstack.org/476204
15:30:50 efried edleafe Found another error in https://review.openstack.org/#/c/475810/ , fixed it, and rebased ^^ for cumulative edits.
15:30:59 efried sfinucan ^ cc
15:31:08 openstackgerrit Matt Riedemann proposed openstack/nova master: Fix sample configuration generation for compute-related options https://review.openstack.org/497460
15:33:29 efried gr, stephenfin (split personality these days?)
15:34:05 stephenfin efried: I had an away message set of sfinucan, but it seems to have been unset
15:34:21 stephenfin I'll just switch it off altogether. stephenfin has been in use for long enough now
15:35:25 stephenfin ...and done
15:36:14 mriedem he died in a tent in greece somewhere
15:36:23 mriedem from a splitting hangover
15:36:55 efried With that kind of hangover, death is a mercy.
15:38:49 efried dtantsur You got the pike backport?
15:41:16 mriedem efried: i just posted it
15:41:24 mriedem well, 5 minutes ago
15:41:51 efried hm, no bot?
15:41:55 mriedem not for stable
15:41:59 mriedem we could add that
15:42:06 mriedem it's somewhere in project-config
15:42:09 efried And no cherry-pick comments in the original?
15:42:20 mriedem that's only if you cherry pick from gerrit,
15:42:34 mriedem which if you that before it's merged, you don't get the cherry picked from line in the commit in the cherry pick
15:42:37 mriedem it's a whole big deal
15:42:40 efried geez
15:42:58 mriedem dansmith: let me know when you'd like to process and discuss https://review.openstack.org/#/c/496995/
15:43:05 dansmith yeah sorry on a call
15:43:12 mriedem efried: just click the change id
15:43:15 mriedem it'll be the same
15:43:21 efried or that
15:52:32 openstackgerrit Merged openstack/os-vif master: Updated from global requirements https://review.openstack.org/488086
16:05:46 openstackgerrit Dan Smith proposed openstack/nova master: Add placeholder migrations for Pike backports https://review.openstack.org/496932
16:05:46 openstackgerrit Dan Smith proposed openstack/nova master: Add uuid to migration object and migrate-on-load https://review.openstack.org/496934
16:05:47 openstackgerrit Dan Smith proposed openstack/nova master: Add uuid to migration table https://review.openstack.org/496933
16:05:58 cdent \o/
16:38:58 openstackgerrit Matt Riedemann proposed openstack/nova master: Remove allocation when booting instance rescheduled or aborted https://review.openstack.org/496995
17:16:09 mriedem blarg we only show faults if the instance is in ERROR or DELETED status
17:16:55 mriedem and when a resize/migrate fails on the dest during _prep_resize we don't set the instance to ERROR state
17:17:02 mriedem which i guess makes sense since it's fine on the source
17:17:17 mriedem but you'd wonder why the thing didn't resize or move
17:17:36 dansmith mriedem: I'm returning to that review for like the fifth time
17:17:45 dansmith need to turn off my email and chat or something
17:17:46 mriedem this time it's personal
17:17:51 dansmith hah
17:18:04 dansmith Review The Patch 5: Do or Die
17:20:52 dansmith mriedem: hmm, yeah, I guess right now if we reschedule the scheduler would actually create a doubled move allocation huh?
17:21:06 mriedem tripled
17:21:21 mriedem i think
17:21:37 mriedem the point of the move / double stuff in the scheduler is to not overwrite the existing allocations
17:21:37 dansmith well, N'd for N retries right?
17:21:40 mriedem yup
17:21:43 dansmith yeah
17:21:48 mriedem quintupled?
17:21:59 dansmith we're awesome
17:22:30 mriedem yeah so i was mostly worried that alex's patch was doing things in a too generic way, but i've since talked myself out of that
17:22:37 mriedem but left the detailed comments in there
17:22:41 dansmith so, this functional test has a mock in it, which is kinda contrary to the goal here right?
17:23:12 mriedem the mock is just to force the build failure on the source host and trigger a reschedule - i suppose you could do that by saying, boot on host A which we know won't fit a flavor or something so the claim fails
17:23:25 dansmith or just not make it a functional test
17:23:30 mriedem but you need to get by the scheduler and to the compute
17:23:45 mriedem we have a similar functional regression test that tests reschedule similar to this
17:23:49 mriedem similarly
17:24:00 dansmith meaning it uses a mock?
17:24:10 mriedem a fixture
17:24:11 mriedem but yeah
17:24:25 mriedem https://github.com/openstack/nova/blob/master/nova/tests/functional/regressions/test_bug_1671648.py#L90
17:24:40 dansmith okay I guess I can't keep track of our intentions here
17:24:59 mriedem given there are 4 services involved with a reschedule i like having the functional test
17:25:48 dansmith yeah, just having a fake driver that fails the first spawn or something seems like less unit-test-esque interaction, but whatever
17:26:13 mriedem could do that too - would be simpler
17:26:18 mriedem later cleanup i suppose
17:26:31 mriedem or in this one, whatever
17:28:00 mriedem things get complicated with a special compute driver i think b/c you have to set that in config before starting the compute service you're using, which is done in setUp, unless you make it a standalone test class, or start a new 3rd compute and boot directly to that host
17:28:02 dansmith mriedem: and we're saying this belongs in ServerMovingTests because why? it's a reschedule?
17:28:22 dansmith sure, but it's more cleanerer I think
17:28:28 mriedem i assume he threw it there because there are other tests checking allocation stuff in there
17:28:34 mriedem but it's not a move, yeah
17:28:45 mriedem didn't think abou that

Earlier   Later