16:00:08 <ildikov> #startmeeting cinder-nova-api-changes
16:00:09 <openstack> Meeting started Thu Apr  6 16:00:08 2017 UTC and is due to finish in 60 minutes.  The chair is ildikov. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:12 <openstack> The meeting name has been set to 'cinder_nova_api_changes'
16:00:24 <ildikov> DuncanT ameade cFouts johnthetubaguy jaypipes takashin alaski e0ne jgriffith tbarron andrearosa hemna erlon mriedem gouthamr ebalduf patrickeast smcginnis diablo_rojo gsilvis  xyang1 raj_singh lyarwood breitz jungleboyj
16:00:35 * johnthetubaguy lurks with intent
16:00:37 <jgriffith> o/
16:00:47 <mriedem> o/
16:01:49 <ildikov> we have a few folks on vacation this week, so we might not get more people
16:02:28 <ildikov> we have the detach refactor patch merged now and have the detach PoC up for review using the new detach flow: https://review.openstack.org/#/c/438750/
16:03:19 <ildikov> lyarwood has one more refactor patch up to centralize the terminate and detach calls: https://review.openstack.org/#/c/452181/
16:03:41 <ildikov> my first question here would be whether that's the direction we want or not?
16:05:02 <mriedem> no
16:05:04 <mriedem> well,
16:05:12 <mriedem> nova.compute.api has never called nova.virt.block_device before
16:05:15 <mriedem> that's a layer violation
16:05:26 <mriedem> nova.virt has always been tied more to nova.compute.!api
16:05:45 <mriedem> even though i see we have nova.virt.hardware in nova.compute.api, gros
16:05:46 <mriedem> *gross
16:06:49 <mriedem> idk, at this point i wouldn't hold up the new detach patch for the refactor
16:07:06 <jgriffith> mriedem that's what I needed to hear ;)
16:07:31 <mriedem> ideally long term it's all centralized
16:07:38 <mriedem> because the copy/paste is not fun
16:07:40 <ildikov> mriedem: fine with me too, at least we don't have to co-maintain a patch chain again which makes me happy at this point :)
16:07:46 <ildikov> mriedem: +1
16:08:54 <mriedem> i feel like we're missing some things in the unit tests on the detach patch
16:09:14 <jgriffith> mriedem in all honesty there's a bunch of things I'll need to add to that
16:09:17 <mriedem> heh, yeah, because the tests are all changed to set attachment_id=None
16:09:19 <jgriffith> unit tests included
16:09:42 <jgriffith> I've just been out of the loop the past couple weeks
16:10:03 <mriedem> that's fine
16:10:05 <ildikov> mriedem: yeah, as a first step we tried to keep the test suite clean and get a general ok to the direction and then finish
16:10:08 <mriedem> i'll comment when i do a full review
16:10:20 <ildikov> mriedem: cool, tnx!
16:11:24 <ildikov> I think we're fine with detach for now unless there's something that seems to be problematic and should be discussed here
16:12:11 <mriedem> yeah i think that's the goal right now, the detach patch,
16:12:14 <mriedem> anything else we need to talk about?
16:12:43 <mriedem> there is something that came up in the spec review,
16:12:46 <mriedem> if you're not tracking it,
16:12:49 <mriedem> which is changes to grenade
16:13:03 <mriedem> to make a volume attachment on the old side and preserve that through to the new side and then detach
16:13:08 <mriedem> that doesn't exist today
16:13:27 <mriedem> so that's up for anyone to pick up if they want
16:14:11 <ildikov> I don't think we went this deep so far on the meetings
16:14:32 <mriedem> i'm happy to end the meeting :)
16:14:46 <ildikov> sorry, I'm just thinking about the grenade case
16:14:56 <mriedem> we're nowhere near needing that yet
16:15:03 <mriedem> it's just something that could be done in parallel
16:15:21 <ildikov> ok, then we're on the same page :)
16:15:58 <ildikov> as until we kind of keep the old attachments and detach them with the old flow we should be ok
16:16:51 <ildikov> another thing around detach is the remove check_detach patch: https://review.openstack.org/#/c/446671/
16:17:04 <ildikov> johnthetubaguy has concerns with that one
16:17:14 <johnthetubaguy> has he
16:17:31 <ildikov> johnthetubaguy: you don't have anymore after finding the BDM checks?
16:18:17 <johnthetubaguy> I am less worried
16:18:25 <johnthetubaguy> just want to make sure we don't change the API behaviour
16:18:26 <ildikov> but you're still worried :)
16:18:30 <johnthetubaguy> around error handling
16:19:57 <ildikov> I have some extra raise InvalidVolume lines in the patch for that one
16:20:25 <ildikov> like how we needed those for removing check_attach due to the error translations
16:21:38 <ildikov> but I can extend the commit message with info if that helps
16:23:39 <ildikov> johnthetubaguy: ^ ?
16:26:41 <johnthetubaguy> sorry, distracted
16:26:58 <johnthetubaguy> I think more context in the commit message would be good
16:27:52 <ildikov> ok, I will add more info there
16:28:39 <ildikov> another one that just hangs is the Cinder v2 API deprecation: https://review.openstack.org/#/c/444253/
16:28:46 <ildikov> johnthetubaguy: mriedem: ^
16:29:07 <ildikov> marginal, but at least small and trivial
16:29:07 <johnthetubaguy> is that all in the etherpad now?
16:29:22 <ildikov> johnthetubaguy: I think I added them, but will check
16:29:26 <mriedem> hmm, i thought we had some plan around when we would deprecate cinder v2 usage
16:29:35 <mriedem> but i'd have to check my notes
16:30:10 <mriedem> http://lists.openstack.org/pipermail/openstack-dev/2017-February/113063.html
16:30:17 <mriedem> "Outside of  the session, Sean Dague suggested that Nova actually deprecate it's  usage of Cinder v2 starting in Pike too, and make Cinder v3 required for  Nova in Queens."
16:30:51 <mriedem> however,
16:30:56 <mriedem> that side steps the issue of upgrades
16:30:57 <ildikov> johnthetubaguy: added the deprecation patch to the etherpad now, check_detach was there already
16:30:58 <mriedem> and old attachments
16:31:08 <mriedem> so if we deprecate cinder v2 usage in pike,
16:31:11 <mriedem> and drop it in queens,
16:31:31 <mriedem> that means we have to have an upgrade plan to migrate old v2-created attachments to the v3 style
16:31:38 <mriedem> which we punted on at the ptg
16:31:43 <mriedem> and said we'd tackle that in queens
16:31:54 <ildikov> mriedem: but v3 is the same as v2, dropping v2 does not mean we go up to 3.27 immediately
16:32:03 <ildikov> or I'm missing something
16:32:04 <mriedem> that's true i guess
16:32:13 <mriedem> we can still call detach/terminate_connection with v3
16:32:17 <mriedem> on the v3.0 microversion yes?
16:32:29 <ildikov> so until we don't add the microversion detection Nova just uses base v3, which is basically v2
16:32:38 <ildikov> mriedem: yes
16:32:46 <mriedem> alright i feel better about that then
16:32:49 <ildikov> mriedem: Nova is already configured to use v3
16:32:53 <mriedem> i know
16:32:58 <mriedem> i just worry
16:33:03 <mriedem> always
16:33:04 <mriedem> worried
16:33:12 <ildikov> the catalog_info patch merged and it's the old flow everywhere the old way
16:33:38 <ildikov> mriedem: you can be relaxed about this one, piece of cake really
16:34:38 <mriedem> ok i'll take a look
16:34:46 <ildikov> so don't make me worry about you're worrying too much :)
16:34:51 <ildikov> mriedem: cool, tnx
16:35:37 <ildikov> this is mainly it from my side for today
16:36:24 <mriedem> let's end it
16:36:45 <ildikov> ok :)
16:37:10 <ildikov> thanks everyone, let's look into the detach PoC until the next meeting and try to get that done
16:37:34 <ildikov> I will be in Korea next week, no idea what time the meeting is there, but will try my best
16:38:06 <ildikov> thank you all!
16:38:12 <ildikov> have a good day!
16:39:03 <ildikov> #endmeeting