14:01:45 <abhishekk> #startmeeting glance
14:01:45 <openstack> Meeting started Thu Aug 13 14:01:45 2020 UTC and is due to finish in 60 minutes.  The chair is abhishekk. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:01:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:48 <openstack> The meeting name has been set to 'glance'
14:01:52 <abhishekk> #topic roll call
14:01:56 <dansmith> o/
14:01:58 <abhishekk> #link https://etherpad.openstack.org/p/glance-team-meeting-agenda
14:02:01 <abhishekk> o/
14:02:05 <Steap> o/
14:02:13 <jokke> o/
14:02:18 <abhishekk> lets wait couple of minutes for others
14:03:45 <abhishekk> lets start, others will join soon
14:03:50 <abhishekk> #topic Updates
14:04:07 <abhishekk> Date for next PTG are out
14:04:26 <abhishekk> it will be held after one week of W summit
14:04:36 <abhishekk> between October 26th to October 30th, 2020
14:05:03 <abhishekk> registrations are open for PTG and summit and as it is virtual it is free
14:05:09 <abhishekk> Summit registration - https://openinfrasummit2020.eventbrite.com
14:05:16 <abhishekk> PTG registration - https://october2020ptg.eventbrite.com
14:05:34 <abhishekk> moving ahead
14:05:43 <abhishekk> #topic release/periodic job updtes
14:05:54 <abhishekk> this is a release week for us
14:06:09 <rosmaita> o/
14:06:15 <abhishekk> we have released python-glanceclient for master and stable/ussuri
14:06:45 <abhishekk> we have also released stable/train and stable/ussuri for glance with some important bug fixes
14:06:58 <abhishekk> glance_store 2.2.0 release patch is in review
14:07:11 <abhishekk> #link  https://review.opendev.org/745796
14:07:11 <patchbot> patch 745796 - releases - Release glance_store 2.2.0 - 1 patch set
14:07:16 <abhishekk> smcginnis, ^^
14:07:49 <abhishekk> We are approaching towards V3 milestones, which is just 4 weeks away
14:08:12 <abhishekk> and for non-client release we have 3 weeks
14:08:24 <jokke> Also note, there was some issue in the announce-release job but the release of 3.2.1 itself went through just mail of it wasn't sent out
14:08:44 <abhishekk> jokke, ++, thank you
14:08:46 <jokke> that's for python-glanceclint
14:09:18 <abhishekk> glance sparse image upload, cinder multiple stores support we are expecting this to be completed on time
14:09:50 <abhishekk> For glanceclient we don't have any major addition this time, so we are good on that front
14:10:28 <abhishekk> Also for glance apart from cinder multiple stores support related changes and few bug fixes we are going to shift most of the work to next cycle
14:10:45 <jokke> And I think the final client release is later anyways if we need to get something released from the work that is still to be done
14:10:50 <abhishekk> Periodic job, mostly yellow due to requirement constrainsts issue, couple of failures due to timeout and flaky test copy_image_revert_lifecycle
14:10:59 <abhishekk> jokke, agree
14:11:51 <abhishekk> dansmith, has found out the cause of timeout issue we were hitting and submitted one patch to reduce its possibility
14:12:15 <dansmith> well, that patch should prevent it always,
14:12:22 <abhishekk> cool
14:12:27 <dansmith> but it just fixes a bug that causes us to deadlock, not the actual problem
14:12:50 <dansmith> debugging of the real issue here: https://bugs.launchpad.net/glance/+bug/1891190
14:12:51 <openstack> Launchpad bug 1891190 in Glance "test_reload() functional test causes hang and jobs TIMED_OUT" [Undecided,New]
14:12:54 <dansmith> but not to the bottom of it yet
14:13:19 <jokke> So I'm not exactly sure how that is the case (I assume we're talking about the bug 1891352)
14:13:20 <openstack> bug 1891352 in Glance "Failed import of one store will remain in progress forever if all_stores_must_succeed=True" [Undecided,In progress] https://launchpad.net/bugs/1891352
14:13:41 <abhishekk> ok, thank you for debugging it
14:13:42 <dansmith> jokke: that's why I wrote the functional reproducer, so it's clear
14:13:57 <dansmith> (for 1891352)
14:15:44 <jokke> dansmith: oh sorry ... actually looking at the code I do see why that is happening ... I read that condition wrong way around
14:15:54 <abhishekk> the scenario here is when all_stores_must_succeed is True and import failed to one of the store then that store was never added to failed list, and the test which I have written for it sometimes proceed before removing the location of previously imported stores
14:16:24 <jokke> for some reason my brain flipped that boolean around when reading the bug
14:16:50 <abhishekk> :D, it happened with me as well
14:17:31 <abhishekk> Ok, moving ahead
14:17:42 <abhishekk> #topic Work moved to next cycle
14:17:42 <jokke> I was wondering "What in earth fails the task when we're catching all exceptions" ;)
14:18:35 <abhishekk> We are moving most of the work to next cycle due to time crunch and ferry of bug fixes occurred this cycle
14:18:53 <abhishekk> Below is the list, which we will shift to next cycle
14:18:54 <abhishekk> Image encryption - Will wait one more week to here from Luzi
14:18:54 <abhishekk> Optimize Ceph store network usage - https://review.opendev.org/#/c/740980/
14:18:54 <abhishekk> Update proposal for duplication image download - https://review.opendev.org/734683
14:18:54 <abhishekk> Cache API - https://review.opendev.org/#/c/665258
14:18:54 <abhishekk> Cluster awareness -
14:18:55 <patchbot> patch 740980 - glance-specs - Optimize Ceph store network usage - 2 patch sets
14:18:55 <abhishekk> Remove single store configuration
14:18:56 <patchbot> patch 734683 - glance-specs - Update proposal for duplication image download - 3 patch sets
14:18:57 <patchbot> patch 665258 - glance-specs - Spec for Glance cache API - 4 patch sets
14:19:18 <abhishekk> Any suggestion/objection?
14:20:09 <abhishekk> alistarle, would you like to share the progress of sparse image upload?
14:20:36 <alistarle> Sure
14:21:11 <jokke> abhishekk: taken the sparse image upload is available, I'll see what I can do about some of the rbd things next week. Don't pull trigger on that just yet
14:21:25 <abhishekk> jokke, ack
14:21:29 <alistarle> We decided to split the two optimization in two commit, first the write optim, which is quite easy, and then the second one, more difficult because it touch some old glance code
14:22:34 <abhishekk> alistarle, sounds good, but we should expect it in this cycle, right?
14:22:38 <alistarle> I will submit the first commit this week, it is fully functionnal and tested in production
14:22:42 <alistarle> sure
14:22:57 <jokke> alistarle: that sounds like a good approach at this point of the cycle
14:23:22 <abhishekk> ++
14:23:24 <abhishekk> and please let us know if you need any help in understanding glance code
14:24:29 <abhishekk> thank you alistarle for updates
14:24:38 <abhishekk> moving to next topic
14:25:02 <abhishekk> #topic doc updates in python-glanceclient
14:25:06 <abhishekk> python-glanceclient still shows create-image-via-import as experimental
14:25:13 <abhishekk> https://docs.openstack.org/python-glanceclient/latest/cli/details.html#glance-image-create-via-import
14:25:49 <abhishekk> So this section shows glance-image-create-via-import as experimental and also states that it might be removed in future
14:26:18 <abhishekk> IMO there is a need to correct it and also backport it to some stable branches
14:26:41 <jokke> yeah ... so the original plan was to change image-crete using the Import workflow and get rid of the mai long image-create-via-import once the code was stable enough to do so
14:27:04 <rosmaita> that's my recollection too
14:27:18 <abhishekk> but we are not moving it, right?
14:27:47 <jokke> Personally I'm not sure if we want to keep that via-import around but instead merge that with image-create with perhaps some flag indicating which way to go
14:28:19 <jokke> it would be cleaner to use, but I'm open for suggestions
14:28:39 <abhishekk> hmm
14:28:41 <jokke> I'd kind of prefer just one image-create command
14:29:07 <abhishekk> One command sounds good to me as well
14:29:29 <rosmaita> right, and since we're talking about CLI here and not changing the rest of the client code, shouldn't impact any services
14:29:43 <dansmith> well,
14:29:46 <jokke> I could do it to utilize --method ... so if import method is provided that flow will be used
14:29:53 <dansmith> people write scripts based on the CLI so I wouldn't say it won't break anyone
14:29:58 <dansmith> not any services, true, but..
14:30:00 <rosmaita> party pooper
14:30:25 <jokke> dansmith: yeah, there is reason why it's flagged experimental with warning it may go away :P
14:30:41 <dansmith> is it flagged that way in the CLI output?
14:31:04 <abhishekk> it is not flagged in CLI output but in the documentation IMO
14:31:12 <jokke> yes
14:31:16 <dansmith> yeah, then nobody noticed :)
14:31:57 <jokke> If you take the help text of the command (which you would need to know what parameters to use) there is like 10 line EXPERIMENTAL: warning on it
14:32:18 <dansmith> so there is in the CLI?
14:32:22 <abhishekk> Also I think, its lot to do in this cycle, so we should do it in next cycle
14:32:27 <jokke> yup
14:32:33 <dansmith> ack, though abhishekk said not
14:32:42 <jokke> `glance help image-create-via-import`
14:33:26 <rosmaita> "EXPERIMENTAL: Create a new image via image import."
14:33:35 <rosmaita> even in all caps
14:34:05 <dansmith> so this was just a shortcut for doing create..stage..import all in one go?
14:34:15 <abhishekk> cool, I never looked at help message as I known all the parameters
14:34:15 <jokke> dansmith: correct
14:34:35 <rosmaita> yes, the idea was that it would be a drop-in replacement for glance image-create
14:35:22 <dansmith> well, just MHO, but I'm sure lots of people interested in import used that instead of the three separate calls...
14:35:26 <rosmaita> (for some definition of "drop-in")
14:35:33 <abhishekk> I am glad that I didn't removed experimental directly and decided to bring it here
14:35:47 <dansmith> I'd not have been in favor of a temporary command in the CLI in the first place, so I'm a little biased, but...
14:36:20 <jokke> dansmith: yeah, that's why I prefer to kind of squash it with image-create so people could keep doing that with very minimal change
14:36:35 <abhishekk> I also think it will be tricky for copy-image support
14:36:44 <dansmith> it's little stuff like this that people hate about upgrading in openstack.. even if the get all the hard ducks in a row, there's still stuff like this to make it painful.. but, you warned them, so I guess you're safe :)
14:36:47 <jokke> as image-create is already doing different stages depending of the parameters it's given
14:37:29 <dansmith> what's the cost of keeping it for compatibility? just weight on our conscience?
14:38:12 <jokke> dansmith: since the initial explosion around v1->v2 we have been pretty darn good flagging stuff we expect might change once the feedback comes in as experimental
14:38:38 <jokke> dansmith: and clutter in the client commands ... the list os already long
14:38:57 <jokke> s/os/is/
14:39:22 <dansmith> it is, and confusingly overlapping
14:39:35 <jokke> yup
14:40:45 <jokke> So I'd like to see i I can get the feature set in image-create before I start my holidays, so we can mark that experimental,deprecated and we could clean it out next cycle. I don't want to not give any transition period for those who are cripting on it
14:41:28 <abhishekk> So to be on safe side, we will keep this command as deprectaed this cycle as well and remove it next cycle
14:41:45 <dansmith> can we reno that it _will_ be remove next cycle?
14:42:03 <abhishekk> just for confirmation image-stage and image-import will stay as it is
14:42:31 <jokke> abhishekk: experimental at least, if I get the squash of the features done this cycle, then deprecated. Otherwise I get it done next cycle and we can deprecate it then and remove following
14:42:37 <jokke> abhishekk: correct
14:42:45 <jokke> just like image-upload
14:42:52 <abhishekk> dansmith, the patch jokke will push will have releasenote saying it is deprecated and will be removed in next cycle
14:43:00 <abhishekk> jokke, ACK
14:43:24 <jokke> dansmith: for sure, I've been quite decent with renos too ;)
14:43:40 <jokke> even no-one reads them, no docs
14:43:55 <abhishekk> ok, moving into open discussion
14:44:13 <abhishekk> #topic Open discussion
14:44:26 <abhishekk> we need reviews on copy-image race condition patches
14:44:44 <jokke> oh, the favorite
14:45:01 <abhishekk> its almost in last phase and will be good enough if we have it merged before vacation period starts
14:45:14 <abhishekk> jokke, rosmaita kindly have a look at those patches
14:45:36 <rosmaita> abhishekk: ack ... can you give me a list?
14:45:37 <dansmith> I think the new functional test is pretty easy to read also
14:45:48 <dansmith> so it should be find to start from there to get the idea
14:45:52 <dansmith> *fine
14:45:57 <abhishekk> rosmaita, https://review.opendev.org/743597
14:45:57 <patchbot> patch 743597 - glance - Implement time-limited import locking - 16 patch sets
14:46:05 <jokke> I think there is still that one revert that will more likely break than not before it hits the code that is supposed to not break it
14:46:46 <dansmith> I didn't parse that
14:47:26 <jokke> The _CompleteTask ... I flagged in one of the previous PSs
14:47:28 <rosmaita> dansmith: nice commit message on 743597
14:47:38 <abhishekk> we also need to backport it to stable/ussuri
14:48:20 <dansmith> jokke: I replied to a comment of yours in _CompleteTask but never saw a reply, from PS8
14:48:41 <dansmith> not sure if that's what you're referring to or not
14:49:05 <dansmith> but if you think something is broken, kindly highlight it again and I'll try to cover that concern with tests
14:50:05 <jokke> dansmith: likely yes. How about I reply to it in the PS8 so no need to hop back and forth trying to follow the convo. IIRC nothing in that part changed between
14:50:27 <dansmith> ack
14:50:38 <jokke> kk will do it after the meeting
14:50:59 <abhishekk> dansmith, how tough it will be to backport to stable/ussuri (considering we need to backport your ImportAction work as well)?
14:51:05 <dansmith> note that we're actively hitting this race in the nova jobs
14:51:10 <dansmith> so it would definitely be good to get this in
14:51:24 <dansmith> abhishekk: I dunno, seems like a big backport
14:51:34 <jokke> yeah we've been bikeshedding around this for what almost two months now
14:51:37 <abhishekk> yeah, I suspect that
14:51:50 <dansmith> the biggest problem is with copy-image.. when did that become a thing? first in ussuri?
14:52:03 <abhishekk> yes, in ussuri
14:52:17 <dansmith> yeah, so that's as far back as I'd want to take it, but still.. eesh, it'd be a big backport
14:52:44 <abhishekk> right
14:52:54 <jokke> mhm ... also the whole race got introduced with that copy-image, it did not exist before
14:53:07 <abhishekk> :D
14:53:14 <jokke> well not on the obvious scale
14:54:03 <abhishekk> Just for FYI tomorrow I will not be around
14:54:08 <dansmith> same
14:54:13 <jokke> same for all of us
14:54:15 <abhishekk> (I guess most of us will not be)
14:54:17 <jokke> pretty much
14:54:46 <abhishekk> That's it from me for today
14:55:01 <abhishekk> we have 5 minutes left before closing
14:55:08 <jokke> I don't think I had anything else either
14:55:32 <jokke> thanks abhishekk for reminding about the -via-import I had kind of forgotten that whole thing already
14:56:01 <abhishekk> I found one small bug in it
14:56:11 <abhishekk> and that's when I noticed it
14:56:37 <abhishekk> there is '-' missing in 'create-image-via import'
14:57:18 <abhishekk> actually its not me but my teammate rajat found it :P
14:57:50 <abhishekk> Lets wrap up for today
14:58:01 <abhishekk> have a nice long weekend guys
14:58:03 <jokke> Thanks all
14:58:11 <jokke> indeed!
14:58:12 <abhishekk> thank you all
14:58:31 <abhishekk> #endmeeting