14:01:37 <abhishekk> #startmeeting glance
14:01:38 <openstack> Meeting started Thu Aug 20 14:01:37 2020 UTC and is due to finish in 60 minutes.  The chair is abhishekk. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:01:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:42 <openstack> The meeting name has been set to 'glance'
14:01:47 <abhishekk> #topic roll call
14:01:51 <dansmith> o/
14:01:54 <abhishekk> #link https://etherpad.openstack.org/p/glance-team-meeting-agenda
14:01:58 <abhishekk> o/
14:02:12 <alistarle> O/
14:02:13 <jokke> o/
14:02:25 <abhishekk> cool, everyone is here, so lets start
14:02:45 <abhishekk> we have short agenda today
14:02:55 <abhishekk> #topic Updates
14:03:03 <abhishekk> October 2020 PTG slot booking is open
14:03:28 <abhishekk> I am proposing to book timeslots between 1400 UTC 1700 UTC
14:03:44 <jokke> ++
14:03:46 <abhishekk> It might be conflict for dansmith and rosmaita
14:03:54 <alistarle> ++ too
14:04:11 <abhishekk> If you guys have any plans then I can keep one day for later hours
14:04:34 <dansmith> I'm okay with 1400 as far as it being early, but I don't know what else I will have on those days
14:04:50 <abhishekk> dansmith, ack
14:04:52 <dansmith> definitely don't schedule it around me
14:05:05 <abhishekk> ack
14:05:27 <abhishekk> I will talk with rosmaita as well and then will book the slots
14:05:37 <abhishekk> #topic release/periodic job updates
14:05:46 <abhishekk> glance_store 2.2.0 released
14:06:02 <rosmaita> the cinder slots are on the ethercalc thing already
14:06:14 <abhishekk> rosmaita, ack
14:06:19 <rosmaita> i'll try to attend anything you schedule other than those times
14:06:44 <abhishekk> V3 milestones 3 weeks away
14:06:50 <abhishekk> Final release for non clients libraries - 3 weeks from now
14:07:17 <abhishekk> So we need to reviews glance_store related patches this and next week on priority basis
14:07:28 <abhishekk> I will highlight those patches in next topic
14:07:36 <abhishekk> Periodic job all green, TIME_OUT, frequent failure of py36 is resolved
14:08:13 <abhishekk> thanks to dansmith we have finally got rid of TIME_OUT and frequent failure of function-py36/py38 issues
14:08:26 <dansmith> \/o
14:08:30 <dansmith> er, heh
14:08:38 <abhishekk> \o/
14:08:48 <abhishekk> Moving ahead
14:08:56 <abhishekk> #topic Important reviews
14:09:08 <abhishekk> Copy image race condition - https://review.opendev.org/743597
14:09:08 <abhishekk> Sparse image upload - https://review.opendev.org/#/c/744282
14:09:08 <abhishekk> Cinder multiple stores support - https://review.opendev.org/746556
14:09:08 <abhishekk> Rampup rbd resize - https://review.opendev.org/746579
14:09:08 <abhishekk> Fixing flaky copy-image-revert-lifecycle test - https://review.opendev.org/#/c/745937/
14:09:27 <abhishekk> Last patch is merged, so ignore that
14:09:32 <dansmith> I just reviewed the sparse file one
14:09:41 <alistarle> Thanks for your review
14:09:50 <alistarle> I didn't have the time to check it
14:09:51 <jokke> yeah same
14:09:58 <jokke> just minutes before the meeting
14:10:11 <abhishekk> ack, today I am conducting meeting from mobile, so not able to look at those
14:11:20 <abhishekk> These are some important reviews which we need to be focus on this week and next week
14:11:21 <alistarle> Thanks for that guys, I will implement the fix before I go to holidays for 3 week tomorrow, then yebinama will take the lead to get that merged asap
14:11:39 <abhishekk> alistarle, ack, thank you
14:11:46 <dansmith> alistarle: all minor stuff I think, should be quick
14:12:08 <alistarle> good new then
14:12:22 <alistarle> Quick question
14:12:30 <abhishekk> alistarle, there is still some partial work pending or this is it?
14:12:34 <alistarle> should I add more functionnal tests ? Or test for RBD ?
14:13:12 <alistarle> yes this first patch implement only the write optimisation of the spec, not the read one, which require much more cleaning work on our side
14:13:28 <abhishekk> more meaningful tests are always good
14:13:41 <abhishekk> alistarle, so the commit message should say partial implements?
14:13:52 <alistarle> I will update that then
14:14:01 <abhishekk> cool
14:14:02 <jokke> alistarle: I'd definitely would love if we had tempest test for that (I think it's only env where we have actual rbd store available) to make sure that th checksum matches if the thin provision is on.
14:14:41 <dansmith> we can float a patch against glance that depends-on this and will run the rbd job configured thusly
14:14:48 <alistarle> about meaninfull test, I see that functionnal test should be the the same for all stores, but thin provisionning is not possible for all stores
14:14:51 <jokke> alistarle: I have no idea how much work it would need, but would be very nice peace of mind
14:15:01 <dansmith> perhaps we should just always enable thin on our jobs after this?
14:15:15 <dansmith> jokke: trivial just to run the job with this enabled
14:16:02 <abhishekk> I guess that is good idea, to always run job with thin provisioning enabled
14:16:52 <dansmith> I'll put a patch up and alistarle can use it to recheck while waiting for the glance_store patch to merge
14:16:52 <alistarle> Ok, should I update the functionnal test of glance_store in this commit then ?
14:17:07 <alistarle> oh thanks
14:17:09 <jokke> well that just flips the situation around when we never test the fat provisioned. Thus I was thinking rather just specific test than full tempest run
14:17:41 <jokke> flip, create an image, download it, see the hash matches
14:18:12 <dansmith> that's not how tempest jobs work
14:18:13 <dansmith> everyone else would test the thick case though
14:18:28 <abhishekk> I am not sure it is possible to flip the glance config parameter in tempest
14:18:43 <dansmith> correct
14:18:54 <jokke> oh, k
14:19:31 <jokke> so every config variation needs it's own test run?
14:20:24 <abhishekk> AFAIK, you can specify your configs in zuul.yaml while defining job
14:20:31 <dansmith> pretty much, although there are some shortcut ways around that but that's not employed very often
14:21:20 <abhishekk> Ok, alistarle thank you for your work
14:21:43 <jokke> fair enough ... so yeah, lets flip it on by default and I think we have periodic ceph job as well, don't we? Maybe having that fat and we have both ways covered on something we monitor
14:22:04 <alistarle> Not for glance_store, but for glance I think
14:22:22 <jokke> yeah, need to test it in glance runs
14:22:24 <alistarle> functionnal test are very light in glance store, only a simple crud for filesystem en swift
14:22:28 <alistarle> *and
14:22:48 <jokke> yup
14:22:55 <abhishekk> we have periodic job for swift and cinder for glance_store
14:23:19 <jokke> was something couple of people started/were supposed to do and both of them moved to other things before much progress was made
14:23:43 <abhishekk> that was ages ago :D
14:23:44 <alistarle> I will not have the time to check that before my holiday, but I can put yebinama on that glance test work if he has the time
14:24:13 <abhishekk> alistarle, ack
14:24:53 <abhishekk> apart from this we have Rampup rbd resizing and cinder multiple store support for reviews
14:25:19 <abhishekk> Ok, moving ahead
14:25:25 <jokke> alistarle: if it's just definition of new job/changing a config option in existing one instead of figuring out how to reconfigure and respin the service, I don't see it being requirement for that patch to merge, just something we should have in place before release
14:25:52 <abhishekk> or I can also take care of that
14:26:13 <dansmith> I'm working on the job right now
14:26:18 <abhishekk> cool
14:27:21 <abhishekk> #topic Open discussion
14:27:49 <abhishekk> So we have some tempest tests for multiple stores
14:27:54 <abhishekk> #link https://review.opendev.org/745712
14:29:11 <dansmith> nice
14:29:52 <jokke> good stuff
14:30:50 <dansmith> that reminds me,
14:30:53 <abhishekk> It is difficult to trigger failure from tempest so adding negative tests might need more time
14:31:03 <dansmith> I should enable my nova-and-glance-rbd-multistore job on glance as well
14:31:43 <abhishekk> dansmith, it is possible to restart the g-api service on cleanup?
14:31:49 <abhishekk> gmann, ^^ in tempest
14:32:10 <dansmith> cleanup of what? the job yes, a test no
14:32:29 <abhishekk> ohh, so in a job can we have single test running?
14:32:45 <abhishekk> but that will be too much of overhead
14:32:48 <dansmith> in nova we have a couple tests that require really weird stuff for live migration, so we actually run tempest restricted to a couple tests multiple times with config switching in between
14:33:09 <dansmith> it's not very common, and a bit tricky, but it's doable
14:33:40 <abhishekk> I will have a look at those tests to get some ideas, could you please point it ?
14:33:46 <dansmith> but tempest tests that confirm invalid configs or something are probably not likely to merge
14:34:02 <gmann> yeah, what dansmith said.
14:34:02 <dansmith> abhishekk: I'll have to go find it, will point you at them in a bit
14:34:28 <gmann> is it changing the config in between of test?
14:34:52 <abhishekk> I want to try revert logic on all_stores_must_succeed is True, by deleting the staging_store or one of the file store in between
14:35:24 <alistarle> can't we use mutating config for that ?
14:35:40 <alistarle> in order to change glance live config without running a new job
14:35:53 <dansmith> abhishekk: if you mean one test that stages, and then break staging, and another test that tries import, that's not a legit thing to land in tempest I think
14:35:57 <abhishekk> gmann, it is about if I delete the staging store file directory during test then that will cause failure for other tests as well if that staging dir is not created again
14:36:13 <dansmith> abhishekk: that's out of scope for tempest I think
14:36:22 <gmann> yeah
14:36:33 <gmann> you can write functional tests for that
14:36:47 <abhishekk> hmm, so may be we need to depend on functional test which we have
14:36:50 <dansmith> abhishekk: we could do that with some shell scripting in post or something, but yeah.. that's what functional tests are for
14:36:56 <tosky> also tempest-like tests in a plugin, if you want to reuse the infrastructure, I guess
14:37:41 <abhishekk> dansmith, ack
14:37:45 <gmann> tosky: it would not be in scope of how tempest tests work.
14:38:12 <abhishekk> tosky, that will create also create new repo, right, or could we maintain it in project repo as well
14:38:47 <abhishekk> gmann, is there any restriction on the size of image we can use for web-download?
14:38:54 <abhishekk> in tempest?
14:39:08 <dansmith> abhishekk: you can control the image in your job,
14:39:13 <gmann> Tempest tests cannot control on when API operation finish and in between verification
14:39:17 <dansmith> so unless it causes it to take longer than the job timeout, should be okay
14:39:24 <tosky> that's why I wrote tempest-like
14:39:38 <abhishekk> dansmith, ack
14:40:03 <gmann> dansmith: yeah but that might not be guaranteed  right
14:40:19 <gmann> i am concern that can end up unstable test
14:40:35 <tosky> if the pattern is: run some test, change the configuration, re-run that test, couldn't that be done with some ansible/zuul glue?
14:40:37 <dansmith> surely, but he asked about specific tempest-based limits, of which there are none, AFAIK
14:40:42 <dansmith> gmann: ^
14:40:46 <gmann> yeah
14:41:21 <gmann> Tempest limits are what API has
14:41:31 <dansmith> abhishekk: btw, really glad to see this kind of team-wide test conversation taking place in glance :)
14:41:46 <tosky> (there is whitebox-tempest-plugin, if you need to something else on the nodes)
14:41:52 <abhishekk> :D
14:42:31 <dansmith> tosky: yeah, do we run that anywhere in upstream CI?
14:42:56 <abhishekk> still we have lot many features to cover with tempest, hopefully that will be covered in this cycle
14:42:57 <gmann> dansmith: on whitebox-tempest-plugin gate itself
14:43:08 <tosky> dansmith: I don't remember on top of my head; the people working on it are mostly working on nova
14:43:13 <dansmith> gmann: okay
14:44:09 <abhishekk> will try to figure out something for negative multiple-stores tempest tests
14:44:38 <abhishekk> that's it from me today
14:44:49 <abhishekk> do you guys have anything else in mind?
14:44:52 <tosky> question before you close: who is responsible for the 'cursive' library? It seems a dependencies, but it's in the x/ namespace
14:45:02 <tosky> I noticed while cleaning some barbican jobs
14:45:22 <tosky> but the usage pointed to glance more than other projects
14:45:47 <gmann> tosky: owner of x/<lib>  repo.
14:46:01 <abhishekk> not sure, I have submitted one patch in cursive (may be 2 years before), there was only one reviewer I guess
14:46:24 <tosky> right, but if it's a glance dependency, just be aware it may break
14:46:30 <rosmaita> cursive was maintained by the johns hopkins applied physics lab
14:46:47 <rosmaita> but they stopped working on openstack maybe two cycles ago
14:46:52 <rosmaita> not sure who owns it now
14:47:06 <abhishekk> we have periodic job for cursive library
14:47:25 <tosky> I mentioned it because we had an x/ repository (devstack-plugin-nfs) which really ended up there by mistake and it was moved back to openstack/ after being officially adopted by cinder & qa
14:47:28 <abhishekk> that will help us to know if something is breaking due to cursive
14:47:36 <rosmaita> yeah, glance, cinder, and nova all use it
14:48:13 <abhishekk> yes
14:48:45 <rosmaita> fwiw, i think it belongs in openstack namespace
14:49:20 <abhishekk> it is now in openstack-archive :o
14:50:02 <tosky> https://opendev.org/x/cursive
14:50:22 <gmann> openstack-archive is because it was retired from opensatck - opendev is what current location is https://opendev.org/x/cursive
14:50:54 <abhishekk> ack
14:51:28 <abhishekk> so we need to find one maintainer to bring it back under openstack?
14:51:51 <abhishekk> as glance, cinder and nova all use it for image signature verification stuff?
14:52:04 <gmann> some project needs to own it or a separate project with maintainer if no project want to own
14:52:33 <abhishekk> ack
14:52:40 <rosmaita> maybe nova, glance, cinder, and barbican can get together at the PTG and work something out
14:52:47 <dansmith> [07:02:45]  <abhishekk>	we have short agenda today
14:52:47 <abhishekk> +1
14:52:50 <dansmith> :P
14:52:52 <abhishekk> :D
14:53:13 <abhishekk> yep, I will add this topic on glance PTG  agenda
14:53:33 <abhishekk> anything else, before closing?
14:53:58 <abhishekk> happy holidays alistarle and jokke
14:54:10 <jokke> o/~
14:54:34 <jokke> I'll be around start of next week, off from Thu latest
14:54:50 <abhishekk> cool
14:54:56 <abhishekk> thank you all
14:55:16 <jokke> thanks everybody
14:55:41 <abhishekk> #endmeeting