16:00:19 <ildikov> #startmeeting cinder-nova-api-changes 16:00:24 <openstack> Meeting started Thu Aug 24 16:00:19 2017 UTC and is due to finish in 60 minutes. The chair is ildikov. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:25 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:27 <openstack> The meeting name has been set to 'cinder_nova_api_changes' 16:00:32 <ildikov> johnthetubaguy jaypipes e0ne jgriffith hemna mriedem patrickeast smcginnis diablo_rojo xyang1 raj_singh lyarwood jungleboyj stvnoyes 16:00:41 <hemna> hi 16:00:44 <jungleboyj> @! 16:00:44 <mriedem> o/ 16:00:44 <_pewp_> jungleboyj ヽ(´・ω・`)、 16:00:53 <smcginnis> o/ 16:00:54 <jungleboyj> Though also in another meeting too. :-) 16:01:22 <ildikov> jungleboyj: thanks for keeping an eye on this one too :) 16:01:32 <jungleboyj> ildikov: I do what I can. :-) 16:01:39 <ildikov> jungleboyj: :) 16:01:39 <stvnoyes> o/ 16:02:14 <ildikov> ok, let's start 16:02:30 <ildikov> current open reviews: https://review.openstack.org/#/q/topic:bp/cinder-new-attach-apis 16:02:56 <smcginnis> https://review.openstack.org/#/q/topic:bp/cinder-new-attach-apis+status:open 16:02:59 <smcginnis> ;) 16:03:19 <ildikov> smcginnis: my bad :) 16:03:37 <smcginnis> Just a more concise view. :) 16:03:45 <ildikov> mriedem: the two on the top of the list are small changes to add an attachment_complete call and to tweak yet once more on the attachment_ref, mainly the connection_info part 16:04:01 <ildikov> mriedem: those can be reviewed already 16:04:32 <ildikov> for the attach patch we still need a client release to get a clean test run on the gate 16:04:40 <ildikov> locally things seem to work 16:05:05 <ildikov> except one thing that stvnoyes has just found with 'cinder show volume' 16:05:43 <smcginnis> ildikov: Latest on a new client release - looks like that will need to be next week. 16:05:53 <ildikov> with this patch we seem to use volume_id as opposed to attachment_id: https://github.com/openstack/python-cinderclient/commit/63ac82a55489d55246da939b5ae60b8a65fb9ec7 16:06:36 <ildikov> smcginnis: is that still issues with stable? 16:06:56 <ildikov> smcginnis: we might need to fix one more thing in the client, so from that perspective it sounds ok 16:07:04 <ildikov> smcginnis: I just wonder what else we have on the table 16:07:06 <smcginnis> ildikov: Yeah, the lib freeze will need to go until everything is complete for pike. 16:07:49 <ildikov> smcginnis: I guess we should be confident with end of next week then? 16:08:26 <smcginnis> ildikov: Better be, otherwise something went really wrong. ;) 16:08:48 <mriedem> i'd like more details in the commit message for https://review.openstack.org/#/c/493324/ 16:08:52 <ildikov> smcginnis: my thinking as well, just wanted to be sure :) 16:09:03 <mriedem> ildikov: also, can you re-propose the spec for queens and make any necessary updates? https://review.openstack.org/#/c/373203/ 16:09:12 <ildikov> mriedem: sure, will do 16:09:24 <mriedem> i guess i never got https://review.openstack.org/#/c/459134/ in 16:09:33 <ildikov> mriedem: yeah, I planned to, wanted to double check with you 16:10:10 <ildikov> mriedem: I will add that and I should add attachment_complete to it too 16:10:49 <mriedem> i've lost context on what my amendment is even about anymore 16:11:16 <mriedem> it was basically trying to capture discussion for something we'd need to consider when supporting multi-attach i think 16:12:17 <ildikov> it was about keeping begin_detaching with the new flow too, so we have the call in the API to put the volume into 'detaching' state as I remember 16:12:45 <ildikov> which is pretty much needed for multi-attach as we don't want two detaches going at the same time 16:13:45 <jgriffith> ildikov why not? 16:14:23 <jgriffith> We've again kinda lost sight of the whole point of discreet and independent attachment objects I think 16:14:42 <jgriffith> heck, at this point I'm not even sure I understand it any more :) 16:15:13 <ildikov> jgriffith: well, if we don't change the volume state that can be attach and etach as well that goes in parallel and I'm not sure that's a good idea 16:15:38 <ildikov> jgriffith: and the volume state machine doesn't support it either 16:15:46 <jgriffith> probably not a discussion for right now, but that whole thing needs to go away/stop 16:15:50 <ildikov> so for now it's better not to create a mess :) 16:15:56 <jgriffith> the attachment needs to be the source of truth 16:16:17 <ildikov> it sounds good for the long run 16:16:30 <ildikov> but we can't start those changes in Nova 16:16:38 <jgriffith> understood 16:16:54 <jgriffith> but don't try and cheat and force in a half assed multi-attach solution either :) 16:17:10 <mriedem> ildikov: how about we just re-propose the spec and leave out https://review.openstack.org/#/c/459134/ 16:17:16 <mriedem> we can roll that into the queens spec later if we want 16:17:21 <jgriffith> anyway, not sure it's really relevant quite yet... sorry 16:17:28 <jgriffith> don't want to derail 16:17:30 <mriedem> right let's avoid the noise 16:17:34 <mriedem> and just ignore https://review.openstack.org/#/c/459134/1 16:17:38 <ildikov> jgriffith: we will talk about the multi-attach bits on the PTG, I hope :) 16:17:55 <ildikov> mriedem: works for me 16:18:04 <ildikov> mriedem: thanks for the hint 16:18:17 <ildikov> jgriffith: BTW, did you see my above comment on cinder show volume not returning the correct attachment_id? 16:18:31 <jgriffith> No 16:18:44 <ildikov> jgriffith: stvnoyes has just found out as he uses that for one of his tempest tests 16:18:48 <smcginnis> mriedem: Do we have a good time for a joint PTG session? 16:18:55 <smcginnis> jungleboyj: ^ 16:18:56 <mriedem> for a good time... 16:19:00 <mriedem> like on a bathroom wall? 16:19:16 <mriedem> smcginnis: anytime wed-fri is fine 16:19:32 <mriedem> we just have an etherpad with proposed stuff right now, nothing in order 16:19:40 <ildikov> jgriffith: so this is apparently the volume_id for some reason: https://github.com/openstack/python-cinderclient/commit/63ac82a55489d55246da939b5ae60b8a65fb9ec7#diff-799a44cd44f14a12e4f1d2a8b56ab63aR39 16:19:48 <smcginnis> mriedem: Same for us. 16:19:57 <jungleboyj> :-) mriedem We should coordinate that. 16:20:07 <mriedem> i hope we're not in the same wedding ballrooms as last time 16:20:09 <ildikov> smcginnis: thanks for bringing it up, so I don't forget :) 16:20:28 <smcginnis> ildikov: I've been meaning to for a while, so I figured I better while we're all here. :) 16:20:29 <jungleboyj> mriedem: Do you have a preference. Maybe Thursday some time? 16:20:34 <mriedem> jungleboyj: that works 16:20:37 <jgriffith> wedding? who's getting hitched? Free food and drink.. YAY 16:20:37 <mriedem> thursday morning 16:20:40 <smcginnis> ++ 16:20:45 <jungleboyj> jgriffith: ++ 16:20:50 <smcginnis> Was just going to suggest morning so there's follow on time. 16:20:50 <mriedem> i just want a room with a ceiling that's <20 feet 16:20:57 <jungleboyj> mriedem: That sounds good. 16:21:03 <ildikov> sounds good 16:21:26 <jungleboyj> jgriffith: ildikov You guys will get everyone up to speed on where the API rework and impacts during that time? 16:21:58 <jungleboyj> mriedem: Is there interest on the Nova side there? I know that the Cinder team needs to be updated. 16:22:16 * smcginnis has to drop off for a bit for solo parent duties 16:22:32 <ildikov> jungleboyj: that's an option, but it would be more high level if we have both teams there 16:22:54 <mriedem> i'd be happy if more nova people were involved 16:23:05 <ildikov> mriedem: +2 +A 16:23:09 <mriedem> if johnthetubaguy is going to be at the ptg he'd need to be updated 16:23:15 <mriedem> otherwise maybe gibi will get the mantle 16:23:24 <jungleboyj> johnthetubaguy: You going to be there 16:23:35 * jungleboyj whispers "please say yes, please say yes" 16:23:37 <ildikov> I think he has a ton of concerns without update too :) 16:23:50 <mriedem> yeah anyway 16:23:52 <diablo_rojo_phon> He is going to be there. 16:24:00 <jungleboyj> Ok, that is good news. 16:24:06 <mriedem> what next 16:24:12 * mriedem still has rc2 bugs to fix 16:24:33 <ildikov> mriedem: yeah, I gave gibi an idea on how annoying I can be with review requests in the second line after 'congrats'... :) 16:24:52 <ildikov> mriedem: not much more on the Nova side until we don't have the cinderclient fixed and released 16:25:05 <mriedem> ok 16:25:26 <ildikov> jgriffith: so as for volume show, it seems to be an easy client side fix 16:25:30 <mriedem> maybe i'll throw https://review.openstack.org/#/c/463987/ in next week 16:25:44 <mriedem> stvnoyes: is ^ still happy with the new flow? 16:25:58 <mriedem> or i guess we can't test that in upstream CI until cinderclient fixes are released 16:26:05 <stvnoyes> working thru that cinder client issue atm 16:26:18 <mriedem> ok, because the stack of api changes at https://review.openstack.org/#/c/330285/ depends on the cinderclient fixes 16:26:24 <mriedem> so i guess i'll hold off 16:27:03 <ildikov> jgriffith: it's weird why we have the volume_id as the value of 'id' under 'attachments' in the volume though, at least for my taste 16:27:15 <ildikov> jgriffith: maybe hemna has the history of that 16:27:36 <ildikov> mriedem: well, things mostly look good locally, but I guess it sounds less convincing anyway 16:28:33 <jgriffith> ildikov there's a lot of garbage in there frankly 16:28:59 <jgriffith> like storing attach-status in 3 different places etc; I'll look at the volume-show stuff you mentioned and see what's up 16:29:00 <mriedem> the attachment_id in the compute API for attached volumes has always been the volume_id i think 16:29:35 <stvnoyes> jgriffith- in cinderclient.v2.shell.py._translate_attachments, the attachment_id is replaced with attachment['id'] which happens to be the volume id 16:29:45 <jgriffith> mriedem I believe you're correct on that 16:29:49 <mriedem> https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/volumes.py#L226 16:29:53 <ildikov> jgriffith: I think for now we can just use the attachment_id in the translate function in the client 16:29:55 <mriedem> super legacy 16:29:58 <stvnoyes> the correct attachment id is in 'attachment_id' 16:30:11 <stvnoyes> not id 16:30:19 <jgriffith> stvnoyes got ya 16:30:22 <mriedem> right, my point is it's probably there fore super old legacy reason 16:30:24 <mriedem> like https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/volumes.py#L226 16:30:32 <mriedem> when cinder wasn't a thing, it was nova-volume 16:30:47 <jgriffith> mriedem +1 you're correct 16:30:50 <jgriffith> n-vol thing 16:30:58 <ildikov> mriedem: ok, then we need to fix that too 16:31:22 <stvnoyes> so this show issue may have been introduced with that translate_attachments change... 16:31:43 <stvnoyes> it was working ok earlier. 16:32:32 <jgriffith> stvnoyes yeah, I made some changes to cinderclient that might be culprit too though. But yes it worked beautifully for a while :) 16:33:04 <jgriffith> anyway... i think this is a Cinder show thing that needs to be fixed and I don't want to pile on things regarding changes to legacy stuff until we finish what's in flight 16:33:28 <jgriffith> there are too many things up in the air right now IMO 16:33:57 <ildikov> jgriffith: we still need the correct data in the volume object too 16:34:36 <ildikov> jgriffith: so we can double check the client, we have a bit more time till the lib freeze is finally over 16:35:56 <ildikov> anything else to the topic? 16:36:54 <mriedem> nope 16:37:11 <ildikov> ok, then I think that was it for today 16:37:38 <ildikov> unless someone has further questions/comments to discuss 16:38:26 <ildikov> ok, we can figure out topics for the joint session on the PTG offline and keep fixing up the cinderclient 16:38:31 <ildikov> thanks everyone! 16:38:41 <ildikov> C U here next week 16:38:43 <hemna> ildikov, that volume_id thing I believe was a placeholder that has been there from the n-vol days, I just left it there to not break anything. 16:39:51 <ildikov> hemna: sounds good, we have the attachment_id explicitly there, so we can fall back to that and do cleanups later 16:40:15 <ildikov> would be my thinking re not breaking anything 16:41:36 <ildikov> alright, have a good rest of the day everyone :) 16:42:09 <ildikov> #endmeeting