Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-09-27
14:07:29 efried dansmith: this https://review.opendev.org/#/c/685181/
14:08:03 dansmith is this a thing we want to do?
14:08:15 efried tldr version discovery endpoints shouldn't require auth, but I don't know how to make /v2.1 use a different paste composite from /v2.1/$everything_else without enumerating $everything_else
14:08:25 dansmith like, should I be able to probe vexxhost to figure out what version it's running with no credentials?
14:08:31 efried absolutely
14:08:49 efried the root discovery documents (everywhere) are already unauth'd
14:09:22 efried but at least nova, cinder, and neutron are broken in this same way, requiring auth for the versioned discovery docs (which are simply subsets of the root one)
14:09:35 dansmith sure, but the microversion available indicates what level of code is running
14:10:31 efried The version discovery process that ksa uses (currently) assumes you can probe both the root and the versioned discovery URIs without auth.
14:11:27 dansmith ack, it just seems like getting into actual details of the system should require auth, but I'm not an expert obviously
14:11:37 efried yeah, it's not really details of the system
14:11:55 efried In theory I agree it would seem as though that process could just rely on the root discovery doc, since it's just a superset of the versioned ones.
14:12:08 efried but I think perhaps the problem is that the service catalog is set up with the versioned endpoint
14:12:08 dansmith available microversion is totally details of the system
14:12:37 efried okay, point is we've already decided that's okay to expose, because it's in the root discovery doc (I think, /me confirms...)
14:13:18 efried yeah:
14:13:18 efried curl http://192.168.218.28/compute/
14:13:18 efried <snip> "version": "2.79" <snip>
14:13:56 efried "perhaps the problem is that the service catalog is set up with the versioned endpoint" shouldn't actually be a problem, which is the point.
14:14:45 efried mordred is working on a hacked up workaround for this in ksa and/or sdk to bridge the gap, but we (nova and the other services running afoul of this) really need to fix it in our APIs.
14:27:51 openstackgerrit Stephen Finucane proposed openstack/nova master: ec2: Move ec2utils functions to their callers https://review.opendev.org/662505
14:27:52 openstackgerrit Stephen Finucane proposed openstack/nova master: api: Remove 'Debug' middleware https://review.opendev.org/662506
14:38:31 efried kashyap: thank you for the thorough response on the ML ++
14:39:54 kashyap efried: No problem. I'll let Manuel file the bug
15:10:18 openstackgerrit Eric Fried proposed openstack/nova master: test cleanup: Remove skipIf test decorator https://review.opendev.org/685375
15:33:50 efried dansmith: are you aware of a reason we don't use oslotest.base.BaseTestCase in nova?
15:34:18 efried or stephenfin ^
15:34:29 stephenfin I'm not sure
15:34:36 stephenfin I'd guess just legacy
15:34:44 stephenfin nova is olllld
15:34:55 dansmith nova pre-dates oslo, so yeah something like that
15:36:45 efried cool, Ima go clean that up.
15:37:02 dansmith is there a reason?
15:37:49 efried tech debt reduction?
15:38:04 dansmith what's the tech debt here?
15:38:26 efried We have copies of half a dozen fixtures and are duplicating the setup steps for them.
15:38:38 efried Output capture, logging, timeout, tempdir...
15:38:55 dansmith we have copies of those in our fixtures?
15:38:58 efried yes
15:39:36 dansmith okay, the actual oslotest base doesn't seem to have much in it, but copies fo the fixtures make sense to squash
15:41:52 mriedem dansmith: this is a fix for a race in the numa live migration series (that came up in the ML), probably good to get in for an rc2 if we have one https://review.opendev.org/#/c/684409/
15:43:21 dansmith mriedem: hrm, do we call that method from anywhere else other than the periodic?
15:43:54 dansmith "accepted" could be considered to be non-terminal and thus "in progress"
15:44:19 dansmith oh, I see it's removing accepted
15:44:28 mriedem right, the ~ is confusing,
15:44:31 mriedem it's only used by the RT,
15:44:36 mriedem there is another method below that is used by the API
15:44:44 mriedem migration_get_in_progress_by_instance
15:44:58 mriedem that will still filter out 'accepted' state migrations
15:44:59 dansmith okay, so this is querying for finished migrations we need to clean up
15:45:20 mriedem which?
15:45:35 mriedem migration_get_in_progress_by_instance was written for the 2.23 change in the api to list in-progress live migrations that can be cancelled for force-completed
15:45:35 dansmith the one we're changing
15:46:00 dansmith yeah, I mean the host-n-node one here
15:46:00 mriedem the one we're changing is used by the RT to track incoming/outgoing migrations so we claim/don't drop resource claims on the source/dest host
15:46:06 dansmith right
15:47:42 dansmith so we need to backport that to stable/train to be ready for rc2?
15:48:52 mriedem yeah, i haven't checked yet if we have a stable/train set of patches up, but we probably do
15:49:00 dansmith the branch is there
15:49:28 mriedem yar https://review.opendev.org/#/q/project:openstack/nova+status:open+branch:stable/train
15:50:20 openstackgerrit Dan Smith proposed openstack/nova stable/train: Stop filtering out 'accepted' for in-progress migrations https://review.opendev.org/685387
15:50:31 dansmith mriedem: ^
15:52:04 dansmith biab
15:56:07 openstackgerrit Eric Fried proposed openstack/nova master: test cleanup: Use oslotest's Timeout fixture https://review.opendev.org/685389
15:56:19 efried dansmith: there's one for e.g. ^
15:58:20 dansmith ack
15:58:29 mriedem huh, interestingly enough, we don't set a migration.status during _post_live_migration before just marking the migration 'completed', meaning a user could attempt to abort/force-complete a live migration after the guest is transferred and we're in post-processing
15:58:41 mriedem it will probably fail, but we shouldn't even make that an option
15:58:51 artom dansmith, thanks for the train backport (sorry, laptop was rebooting/installing updates)
16:00:21 openstackgerrit Merged openstack/python-novaclient master: Stop silently ignoring invalid 'nova boot --hint' options https://review.opendev.org/684762
16:01:19 dansmith artom: np
16:01:21 mriedem efried: to make launchpad sane, i've updated https://launchpad.net/nova/train to be in pre-release freeze status and ussuri is active development
16:01:32 efried thanks
16:01:44 mriedem e.g. technically https://launchpad.net/bugs/1845146 will be fixed in ussuri but not yet fixed in train
16:01:44 openstack Launchpad bug 1845146 in OpenStack Compute (nova) train "NUMA aware live migration failed when vCPU pin set" [High,In progress] - Assigned to Artom Lifshitz (notartom)
16:04:12 mriedem artom: when you're back you might sanity check my comments on https://review.opendev.org/#/c/684409/1/nova/db/sqlalchemy/api.py to see if we have a (latent) bug to follow up on
16:05:29 openstackgerrit Matthew Booth proposed openstack/nova master: libvirt: Ignore DiskNotFound during update_available_resource https://review.opendev.org/685391
16:06:43 openstackgerrit Matthew Booth proposed openstack/nova master: libvirt: Ignore DiskNotFound during update_available_resource https://review.opendev.org/685391
16:07:39 openstackgerrit Eric Fried proposed openstack/nova master: test cleanup: Use oslotest's CaptureOutput fixture https://review.opendev.org/685392
16:17:42 openstackgerrit Matt Riedemann proposed openstack/nova stable/train: Stop filtering out 'accepted' for in-progress migrations https://review.opendev.org/685387
16:18:04 openstackgerrit Merged openstack/nova master: ec2: Remove ec2.CloudController https://review.opendev.org/662503
16:22:08 openstackgerrit Merged openstack/nova master: ec2: Pre-move cleanup of utils https://review.opendev.org/662504
16:33:58 openstackgerrit Eric Fried proposed openstack/nova master: test cleanup: Make base TestCase subclass oslotest https://review.opendev.org/685399
16:34:13 efried dansmith, stephenfin: I think that's the last one ^
16:34:42 efried stephenfin: note that this conflicts (both git-wise and spiritually) with your patch to remove those output-capturing env vars.
17:48:32 openstackgerrit Merged openstack/nova stable/rocky: Fix rebuild of baremetal instance when vm_state is ERROR https://review.opendev.org/680871
17:48:38 openstackgerrit Merged openstack/nova stable/rocky: Fix 'has_calls' method calls in unit tests https://review.opendev.org/677330
17:48:45 openstackgerrit Merged openstack/nova stable/rocky: Fix wrong assertions in unit tests https://review.opendev.org/677389
18:08:05 openstackgerrit Merged openstack/nova stable/rocky: Retrun 400 if invalid query parameters are specified https://review.opendev.org/676027
18:08:10 openstackgerrit Merged openstack/nova stable/rocky: Fix non-existent method of Mock https://review.opendev.org/676840
18:08:16 openstackgerrit Merged openstack/nova master: Update master for stable/train https://review.opendev.org/685258
18:08:24 openstackgerrit Merged openstack/nova master: Add TODO note for mox removal https://review.opendev.org/576758
19:11:00 openstackgerrit Merged openstack/nova master: Add functional tests for virtual persistent memory https://review.opendev.org/678470
20:07:40 openstackgerrit Eric Fried proposed openstack/nova master: Allow versioned discovery unauthenticated https://review.opendev.org/685181
20:07:59 efried I think I did it dansmith stephenfin ^
20:45:34 openstackgerrit Merged openstack/nova master: nova-net: Migrate some API sample tests off of nova-net https://review.opendev.org/684325
20:45:41 openstackgerrit Merged openstack/nova stable/train: Update .gitreview for stable/train https://review.opendev.org/685256
20:48:13 openstackgerrit Merged openstack/nova stable/train: Update TOX/UPPER_CONSTRAINTS_FILE for stable/train https://review.opendev.org/685257
20:55:34 mriedem BLARG
20:55:37 mriedem melwitt: fun fact,

Earlier   Later