14:04:21 <jokke_> #startmeeting glance
14:04:21 <openstack> Meeting started Thu Apr 26 14:04:21 2018 UTC and is due to finish in 60 minutes.  The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:04:22 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:04:24 <openstack> The meeting name has been set to 'glance'
14:04:31 <jokke_> #topic roll-call
14:04:34 <abhishekk> o/
14:05:12 <rosmaita> o/
14:05:34 <jokke_> So today's agenda does not exist in the etherpad ... I have few items I'd like to discuss and lets try to get some decent OpenDiscussion time for us
14:05:41 <jokke_> Agenda:
14:05:58 <jokke_> 1) Belmiro's Image hiding spec
14:06:28 <jokke_> 2) Abhishek's multiple store back-ends
14:06:54 <jokke_> 3) policy refactoring spec
14:07:00 <jokke_> 4) Open Discussion
14:07:17 <jokke_> #topic Image hiding spec
14:07:50 <rosmaita> i asked belmiro to add something about why discoverability is important, since that came up at the ptg
14:08:02 <jokke_> #link https://review.openstack.org/#/c/545397/
14:08:16 <rosmaita> he says it is for his use case, it is also needed in our cloud
14:08:22 <jokke_> I'm started to get worried about this spec
14:08:37 <rosmaita> in what way
14:10:08 <jokke_> first of all there does not seem to be traction to work on it. Second I'm starting to be less and less convinced that the use cases actually match on the proposed purpose
14:10:21 * smcginnis sneaks in late
14:10:30 <jokke_> smcginnis: welcome \o
14:11:42 <jokke_> So hiding the images from the image list was discussed to be an option when you want to retire the image so people won't boot new instances from it but you need to have the image available for nova to do migrations and deal with snapshots
14:12:13 <rosmaita> and also for specific customer uses
14:12:31 <rosmaita> i mean, image needs to be available for some specific customer use cases
14:12:41 <jokke_> now it looks like it's just turning to opposite flag for something like "latest": True filter by default if "latest" == True
14:13:04 <jokke_> rosmaita: then it should not be retired
14:13:25 <rosmaita> jokke_ exactly, that is why it is "hidden"
14:13:31 <jokke_> if there is specific reason why image should be around and discoverable, you should not hide it
14:13:38 <rosmaita> jokke_ no
14:14:03 <rosmaita> the idea is that the public images in the default image list are the suggested images that end users should use
14:14:21 <rosmaita> but there may be specific reasons for using other public images that have not been pulled from circulation
14:16:37 <abhishekk> (due to network issue, I am getting late updates :( )
14:17:33 <jokke_> it's also not hidden And that's exactly what I mean by saying that it seems we're not talking about the same feature anymore ... so now we need visibility = Public-Hidden because Private, Shared, Community, Public is not good enough anymore
14:18:12 <jokke_> maybe you can just instead of hiding the public image change the visibility to community and you would achieve the same end result without any changes to Glance
14:18:54 <rosmaita> well that was a workaround, but people reacted with horror to it
14:18:58 <rosmaita> this sounds like a forum discussion
14:19:19 <rosmaita> what did we wind up submitting for forum topics?
14:19:36 <rosmaita> (i am still catching up after being away all last week(
14:19:57 <smcginnis> There are a few open rooms. If there's nothing for this, we might be able to get some time.
14:20:08 <smcginnis> https://docs.google.com/spreadsheets/d/15hkU0FLJ7yqougCiCQgprwTy867CGnmUvAvKaBjlDAo/edit#gid=1998591996
14:21:44 <rosmaita> smcginnis: thanks! room 3 seems to have some availability
14:22:01 <jokke_> 11:50 at Thu in room 2 Image Lifecycle Management
14:22:18 <smcginnis> I think the main plan is those can be used for discussions that overflow their allocated slots, but I'm sure new topics can get added if needed.
14:22:35 <smcginnis> Oh perfect.
14:23:28 <rosmaita> excellent ... i see the image/edge computing meetup was scheduled too
14:23:40 <jokke_> yes and it's double slot
14:24:00 <jokke_> so we should have good time to understand the needs and wishes
14:24:15 <jokke_> ok, moving on
14:24:36 <jokke_> #topic Multiple Store Back-ends spec
14:25:03 <jokke_> #link https://review.openstack.org/#/c/562467/
14:25:24 <jokke_> I put quite a few comments there to kind of open the discussion around that
14:25:44 <abhishekk> Submitted in last week, mainly based on cinder's multiple backends approach
14:26:06 <jokke_> abhishekk: thanks for very exhaustive spec. You clearly documented well what you have been thinking around that
14:26:22 <abhishekk> Main concern is as it's a library we need to maintain backward compatibility
14:26:33 <rosmaita> ++
14:26:36 <abhishekk> jokke_, thank you for your suggestions as well
14:26:43 <rosmaita> (i havent' read the spec yet)
14:27:00 <abhishekk> (I have a POC ready)
14:27:41 <rosmaita> cool
14:27:47 <abhishekk> (where I am able to maintain backward compatibility as well)
14:28:15 <abhishekk> But as jokke_ said, I am open for improvements and better ideas
14:28:53 <jokke_> abhishekk: and that's why I suggested earlier that we define new public API on that lib that will handle this, deprecate the current mess currently there everyone seems to be hating and handle the multiple same stores on that new API
14:29:17 <jokke_> that way we can keep the backwards compatibility during the deprecation period and then drop it
14:29:52 <jokke_> it will become even more horrific mess if we try to put hooks in the current python api of that lib to do all this
14:30:16 <abhishekk> jokke_, I will think on this approach
14:30:38 <jokke_> the store classes should not need huge refactoring, it's really just the mechanism to retrieve the store objects
14:31:00 <abhishekk> what you mean to say is don't modify old glance_store api's instead write the new one
14:31:06 <jokke_> I think the biggest refactoring, as we discussed, would be handling the configs
14:31:43 <abhishekk> I am able to manage that with current structure
14:32:03 <jokke_> so the store objects have their API (read, write, delete, etc.) The major pain point is how we retrieve those store objects
14:32:47 <jokke_> like in the tasks stuff where we are using those same store objects, we need to do quite nasty hacks to get the objects to interface with the backend
14:33:38 <jokke_> and I think that is the part that should be rewritten and simplified into a form where we can get just store object out of glance_store with specific config and use that as we wish
14:34:19 <abhishekk> jokke_, I will try to accommodate  this suggestion, will ping you if has any doubts
14:35:46 <jokke_> cool ... Is there any major concerns, objections etc. around this? rosmaita smcginnis ?
14:36:08 <rosmaita> i havent read the spec, but the general approach you are discussing sounds good
14:36:08 <smcginnis> I need to read through, but not so far.
14:36:59 <jokke_> so the other major point we need to find consensus and good solution is the REST API glance offers to utilize those multiple stores
14:37:14 <abhishekk> jokke_, so we need to keep old way of registering config options as it is for backward compatibility and add a new way for multi store support, right?
14:37:50 <jokke_> abhishekk: unfortunately yes ... the old way needs to stick around for couple of cycles so we can deprecate it
14:38:19 <abhishekk> jokke_, got it
14:38:29 <rosmaita> jokke_ maybe that's something we can ask for room 3 for ... get some feedback from operators
14:39:06 <jokke_> and in the glance-api end it's not that big of a deal once we have the new store api ready and in use, to just translate the old configs to the new way of interfacing with glance_store
14:39:16 <rosmaita> i can see that some ops may want the multiple backends but put restrictions on who can access them
14:39:28 <jokke_> it's the lib that needs to maintain backwards compatibility, not how glance uses it :D
14:40:18 <abhishekk> :D
14:40:22 <jokke_> rosmaita: yeah, I did not want to go there with the initial plans, I think that's something people can request and start working on if they need it
14:40:42 <rosmaita> ok
14:42:30 <jokke_> rosmaita: we saw what happened with Image Import Refactoring when we tried to take every usecase into account ... 3 years of planning and when implemented, not even all usecases were valid anymore :P
14:42:58 <rosmaita> good point
14:43:18 <jokke_> ok lets hop to the next topic
14:43:33 <jokke_> #topic Policy Refactoring Spec
14:43:50 <rosmaita> so that spec was really vague about details
14:43:55 <jokke_> #link https://review.openstack.org/#/c/528021/
14:44:54 <rosmaita> i need to read wxy's comments
14:45:17 <rosmaita> and i guess people are in fact on board with the idea that we should move policies to the api layer (more or less)
14:46:00 <jokke_> that is one thing as it's really complex beast to tackle. it just seems to have lost traction. The branch is there, but haven't heard from dharini since and the spec has not got too much love either
14:46:49 <abhishekk> jokke_, dharini?
14:46:54 <rosmaita> dharini stopped working on openstack last year, right about this time
14:47:09 <jokke_> so more or less wanted to bring this spec up, that we pretty much agreed in PTG it needs to be done and it will take time, so the feature branch is there, now we just need reviews on the spec that it captures what was agreed and hands to start experimenting and doing the work
14:47:24 <jokke_> sorry ... wrong person
14:47:34 <abhishekk> jokke_, bhagyashri has plans to work on it
14:47:40 <jokke_> yes, her
14:47:43 <abhishekk> I will try to contact her
14:47:46 <jokke_> total brainfreeze
14:48:17 <rosmaita> dharini did have a patch up to remove show_multiple_image_locations option at one point, though
14:48:59 <jokke_> rosmaita: yeah and lots of other good work came from here ... just totally my bad confusing those two ladies that both have been doing good work with glance :D
14:49:07 <rosmaita> :)
14:49:08 <jokke_> from her
14:49:59 <jokke_> Ok, so lets get these specs reviewed and merged latest around Summit!
14:50:06 <jokke_> #topic Open Discussion
14:50:37 <rosmaita> my item is an awareness thing
14:50:46 <jokke_> FYI, I'll be more or less off next Mon-Wed ... Parents coming over so will try to take some time with them
14:50:56 <rosmaita> have fun!
14:51:09 <rosmaita> hope the sauna is working ok
14:51:16 <jokke_> Telegram/whatsApp will reach if I'm needed, just expect delays on responses
14:51:23 <jokke_> rosmaita: it is indeed :D
14:51:33 <rosmaita> excellent!
14:51:54 <abhishekk> me too, we have holiday on Tuesday, so on monday I will take leave
14:52:06 <rosmaita> ok, so doug did a bunch of work in june 2017 moving our docs around
14:52:29 <rosmaita> which was great, but there are dead urls left around
14:52:36 <smcginnis> I'm travelling next week, but feel free to ping me if you need anything.
14:52:54 <rosmaita> looks like i will be the only one here!
14:52:58 <abhishekk> rosmaita, I will not be available for bug squad meeting on Monday
14:52:59 <jokke_> Yeah, I've seen you and lots of other people poking those when ever they have stumbled upon them
14:53:13 <rosmaita> anyway, i put up a patch to set up redirects from all the old urls to the new urls
14:53:15 <rosmaita> https://review.openstack.org/#/c/564077/
14:53:31 <jokke_> rosmaita: oh that's what it was! GR8!
14:53:41 <rosmaita> there will still be some dead ones probably, but this covers all the stuff moved last year
14:53:43 <jokke_> saw it coming, just haven't looked into it yet
14:53:57 <jokke_> amazing, thanks
14:54:08 <rosmaita> yeah, i had some problems getting it right, think it is good now
14:54:24 <rosmaita> one of those things that i thought would take an hour or so
14:54:37 <jokke_> #link  https://review.openstack.org/#/c/564077/
14:54:40 <rosmaita> but turned out to be a bit bigger
14:54:42 <jokke_> 3 days later? ;D
14:54:45 <rosmaita> yeah
14:54:54 <rosmaita> :)
14:55:00 <rosmaita> abhishekk : and it has tests!
14:55:22 <abhishekk> :D
14:55:28 <jokke_> there is no such thing as "Just quick think I do between afternoon coffee and logging off" in Glance :D
14:55:34 <rosmaita> anyway, my point is awareness so that when a dead link is found, we can set up a redirect as well as fix the text
14:55:42 <jokke_> nice
14:55:58 <jokke_> that way people's possible bookmarks will still work as well
14:56:03 <rosmaita> exactly
14:56:18 <abhishekk> ++
14:56:37 <jokke_> cool, anything else?
14:57:10 <rosmaita> i've noticed some functional test failures, i think it's a timing issue in the image import tests
14:57:16 <abhishekk> no (will ping you if need help in muti-store specs)
14:57:27 <rosmaita> i will put up a patch, not sure if it's a good idea or not
14:57:46 <abhishekk> i guess we can add a wait until image becomes active?
14:58:02 <rosmaita> we can discuss ... but if you see an error like bad checksum for one of the import functional tests, that's what's happenng, i think
14:58:18 <rosmaita> abhishekk : yeah, that was my thought, a time-bounded status check
14:58:31 <abhishekk> that will be good to hve
14:58:35 <abhishekk> have
14:58:54 <rosmaita> i have something sketched out, will put up a patch later today
14:58:58 <rosmaita> "should only take an hour"
14:58:59 <rosmaita> :)
14:59:08 <abhishekk> great, thanks
14:59:21 <jokke_> perhaps just test helper function to poll when the image comes to specific state and handles the timeout for it
14:59:26 <rosmaita> yeah, those are good tests, abhishekk
15:00:00 <jokke_> ok, we're out of time. Thanks all!
15:00:06 <abhishekk> thank you all
15:00:08 <rosmaita> bye!
15:00:15 <jokke_> #endmeeting