Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-11-13
15:36:27 mdbooth lyarwood: Re https://review.opendev.org/#/c/694033/ What did you think about combining these 2 traits?
15:37:00 mdbooth IIUC their window of usefulness is limited to the period between stable disks and BFV rescue landing, right?
15:37:24 mdbooth COMPUTE_RESCUE_BFV implies COMPUTE_RESCUE_STABLE_DEVICES
15:39:02 lyarwood mdbooth: for the libvirt driver implementation yeah
15:40:48 sean-k-mooney i think having two traits is fine
15:41:01 mdbooth lyarwood: As I said in the spec, though, if this discussion gets in the way of getting this done, it's not worth it.
15:41:01 sean-k-mooney glad to see you put them under compute :)
15:41:32 mdbooth But I think 1 trait would be better.
15:42:07 sean-k-mooney its only better if we support rescue for BFV in the same step
15:42:21 sean-k-mooney or release
15:42:44 mdbooth sean-k-mooney: Right, but the change from stable disks to rescue BFV is trivial.
15:42:58 mdbooth IIRC it's just removing a check in the api which prevents it.
15:43:23 sean-k-mooney sure although one point. you cant assume that just using a usb device will mean it wont reorder the disk in all cases
15:43:37 sean-k-mooney its going to be true 99% of the time
15:43:43 mdbooth sean-k-mooney: That's unrelated.
15:44:00 mdbooth (True, but unrelated)
15:44:02 sean-k-mooney well its part of the premisis of the stable_device resuce spec
15:44:06 sean-k-mooney ya
15:44:39 lyarwood it's not going to reorder the physical layout, that's all we can guarantee
15:44:50 sean-k-mooney just said i would mention it since you can use hw_disk_bus to usb already https://github.com/openstack/glance/blob/master/etc/metadefs/compute-libvirt-image.json#L39
15:44:54 lyarwood everything else within the guestOS is out of our control
15:45:00 mdbooth Right. It's definitey way better than what we do now in all cases.
15:45:16 mdbooth lyarwood: Don't suppose you still have a link to the old patches kicking about, do you?
15:45:28 lyarwood yeah I'm working through a rebase now
15:45:54 lyarwood mdbooth: https://review.opendev.org/#/q/topic:bp/virt-rescue-stable-disk-devices
15:46:07 lyarwood mdbooth: hope to have it posted later this evening
15:47:02 mdbooth lyarwood: Cool. Did we discuss switching it on unconditionally in a new microversion, btw?
15:47:22 mdbooth IIRC there's a new microversion involved anyway.
15:48:31 lyarwood mdbooth: that's what I'm suggesting in the follow up spec at the moment
15:48:47 lyarwood mdbooth: well, with the trait
15:48:54 lyarwood mdbooth: so it's not unconditional
15:53:49 openstackgerrit Matt Riedemann proposed openstack/nova master: api-ref: re-work migrate action post-conditions https://review.opendev.org/694103
15:54:54 openstackgerrit Stephen Finucane proposed openstack/nova master: Fix review link. https://review.opendev.org/689612
15:59:05 openstackgerrit Elod Illes proposed openstack/nova stable/pike: Add functional test for resize crash compute restart revert https://review.opendev.org/687913
16:04:07 openstackgerrit Merged openstack/nova master: ItemsMatcher: mock call list arg in any order https://review.opendev.org/689487
16:04:11 artom sean-k-mooney, I know you're respinning https://review.opendev.org/#/c/674072/4 soon, but there are a couple more typos I found
16:04:57 mriedem melwitt: dansmith: this might have come up before and i'm just not remembering, but an instance that fails initial scheduling and is buried in cell0 doesn't have a 'create' instance action because we apparently don't create that in cell0
16:05:19 dansmith mriedem: not sure it has come up before
16:05:30 dansmith mriedem: did it have one prior to the bury-in-cell0 behavior?
16:05:46 melwitt I don't recall talking about this before either
16:07:53 mriedem looking back at mitaka, the api would create the 'create' action in the 'nova' cell db https://github.com/openstack/nova/blob/mitaka-eol/nova/compute/api.py#L1180 before casting off to conductor
16:08:53 mriedem i don't see anything in conductor which would "complete" that action with a fail even if scheduling failed https://github.com/openstack/nova/blob/mitaka-eol/nova/conductor/manager.py#L374
16:08:56 mriedem but at least the action would exist
16:09:05 dansmith ack
16:09:10 dansmith well, makes sense
16:11:22 openstackgerrit Elod Illes proposed openstack/nova stable/pike: Prevent init_host test to interfere with other tests https://review.opendev.org/687916
16:13:04 mriedem makes sense that we wouldn't create the instance action in cell0?
16:13:53 sean-k-mooney artom: cool i ran a spell check on it since i figured out how to do that with emacs
16:14:21 sean-k-mooney artom: im just running the tests currently
16:14:37 artom sean-k-mooney, it was just capitalization and "use case" into 2 words
16:15:24 openstackgerrit Merged openstack/nova master: Fix ItemMatcher to avoid false positives https://review.opendev.org/689690
16:15:59 mriedem sean-k-mooney: are you planning on writing functional tests for https://review.opendev.org/#/c/674072/ ?
16:17:06 dansmith mriedem: no makes sense that we missed doing that when we created bury-in-cell0, and makes sense that we should fix that
16:17:13 mriedem dansmith: ah ok
16:17:25 mriedem i'll open a bug in a bit
16:20:51 sean-k-mooney am i can if you would like them
16:21:23 mriedem ?
16:21:23 mriedem anything involving image meta / flavor extra specs + pci + affinity + scheduling + compute likely means unit tests aren't sufficient, yeah
16:21:29 mriedem maybe that's just me
16:21:48 openstackgerrit Alexandre arents proposed openstack/nova master: Abort live-migration during instance_init https://review.opendev.org/678016
16:23:26 sean-k-mooney mriedem: the code change is just reuseing the existing support for numa policies
16:23:44 sean-k-mooney so im jsut passing that policy via the flavor instead of the alisa
16:24:00 sean-k-mooney that said we may not have existing fucntional test for that so i will look
16:27:57 sean-k-mooney mriedem: i was kind of assuming the existing functional test for the numa polices would be sufficent but i will check.
16:30:06 mriedem i never assume anything relating to numa test coverage is sufficient for new code dealing with numa
16:30:17 sean-k-mooney fair
16:30:22 mriedem but i'm a cranky old troll
16:37:21 openstack bug 1833581 in OpenStack Compute (nova) train "instance stuck in BUILD state if nova-compute is restarted" [Low,Fix committed] https://launchpad.net/bugs/1833581 - Assigned to Balazs Gibizer (balazs-gibizer)
16:37:21 openstackgerrit Elod Illes proposed openstack/nova stable/pike: Functional reproduce for bug 1833581 https://review.opendev.org/687917
16:38:13 mriedem melwitt: can you hit https://review.opendev.org/#/c/693554/ to keep amodi's doc bug fix backports moving?
16:45:02 melwitt mriedem: yup, thanks
16:45:10 stephenfin mriedem: You're not that old
16:45:30 stephenfin 🙃
16:47:45 sean-k-mooney artom: looks like i have already fixed most of your comments locally im going to fix the last few then ill push it up.
16:48:42 sean-k-mooney stephenfin: do you recall if you wrote functional tests when you implemented the orgininal numa policies. if so ill extend them if not i guess ill write tests for both.
16:49:25 stephenfin I very much doubt it. I've only started writing those for the last two cycles or so
16:49:58 artom sean-k-mooney, I know I added https://review.opendev.org/#/c/682941/ recently
16:50:11 sean-k-mooney ok
16:50:22 sean-k-mooney artom: ya i rememebr
16:50:35 sean-k-mooney so those cover the alias based policies i think
16:50:38 sean-k-mooney you added the missing one
16:51:05 sean-k-mooney so i was more or less relying on thos but it hsould be easy to etend those to use the policy form the flavor
16:51:07 artom sean-k-mooney, well yeah, because we only have the alias based policies for now :)
16:52:03 sean-k-mooney ok ill push there version i have now after i adress you comments and ill start working on the func test after
16:54:04 openstackgerrit sean mooney proposed openstack/nova master: support pci numa affinity policies in flavor and image https://review.opendev.org/674072
17:09:10 openstackgerrit Eric Fried proposed openstack/nova master: Consolidate [image_cache] conf options https://review.opendev.org/690723
17:09:11 openstackgerrit Eric Fried proposed openstack/nova master: Add image caching to the support matrix https://review.opendev.org/690748
17:09:56 efried mriedem, stephenfin: Rebased and nit-fixed ^
17:12:15 openstackgerrit Stephen Finucane proposed openstack/nova master: "SUSPENDED" description changed in server_concepts guide and API REF https://review.opendev.org/663590
17:13:00 openstackgerrit Elod Illes proposed openstack/nova stable/pike: Error out interrupted builds https://review.opendev.org/687918
17:22:19 artom stephenfin, oh, friendly reminder that the NUMA LM func test stack is ready for another look: https://review.opendev.org/#/c/687404/
17:22:36 artom (I know it's not so much a learning curve as a learning brick wall)
17:24:10 stephenfin artom: Sure thing, but that's too much work for 5:20pm when jetlagged. I'll grab it in the morning :)
17:24:21 artom stephenfin, ack, thank you!
17:25:40 openstackgerrit Eric Fried proposed openstack/nova master: Remove functional test specific nova code https://review.opendev.org/683609
17:31:18 bauzas mriedem: dumb question but what the user sees from the API when we error out some instance because of NoValidHost ?
17:31:37 bauzas by a nova show I mean
17:32:00 bauzas mriedem: was about to say +1 to your ML thread, but was wondering the API result for the same
17:33:11 mriedem $ openstack server show build-fail1 -f value -c fault {u'message': u'No valid host was found. ', u'code': 500, u'created': u'2019-11-13T15:57:13Z'}
17:33:11 mriedem bauzas: the non-admin user? they see this:

Earlier   Later