| *** sdague has quit IRC | 00:01 | |
| *** david-lyle has quit IRC | 00:03 | |
| *** dklyle has joined #openstack-meeting-cp | 00:03 | |
| *** david-lyle has joined #openstack-meeting-cp | 00:07 | |
| *** dklyle has quit IRC | 00:08 | |
| *** edmondsw has quit IRC | 00:15 | |
| *** dklyle has joined #openstack-meeting-cp | 00:29 | |
| *** david-lyle has quit IRC | 00:30 | |
| *** david-lyle has joined #openstack-meeting-cp | 00:55 | |
| *** dklyle has quit IRC | 00:55 | |
| *** david-lyle has quit IRC | 01:01 | |
| *** edmondsw has joined #openstack-meeting-cp | 01:24 | |
| *** edmondsw has quit IRC | 01:29 | |
| *** yamahata has quit IRC | 02:54 | |
| *** iyamahat has quit IRC | 02:54 | |
| *** diablo_rojo has quit IRC | 03:13 | |
| *** edmondsw has joined #openstack-meeting-cp | 03:13 | |
| *** edmondsw has quit IRC | 03:17 | |
| *** aselius has quit IRC | 03:21 | |
| *** iyamahat has joined #openstack-meeting-cp | 04:10 | |
| *** nhelgeson has quit IRC | 04:11 | |
| *** coolsvap has joined #openstack-meeting-cp | 04:19 | |
| *** yamahata has joined #openstack-meeting-cp | 04:26 | |
| *** edmondsw has joined #openstack-meeting-cp | 06:49 | |
| *** edmondsw has quit IRC | 06:54 | |
| *** david-lyle has joined #openstack-meeting-cp | 07:19 | |
| *** gouthamr has quit IRC | 07:20 | |
| *** markvoelker has quit IRC | 07:37 | |
| *** yamahata has quit IRC | 07:38 | |
| *** edmondsw has joined #openstack-meeting-cp | 08:38 | |
| *** markvoelker has joined #openstack-meeting-cp | 08:38 | |
| *** edmondsw has quit IRC | 08:42 | |
| *** MarkBaker has quit IRC | 08:44 | |
| *** iyamahat has quit IRC | 08:52 | |
| *** _pewp_ has quit IRC | 09:39 | |
| *** _pewp_ has joined #openstack-meeting-cp | 09:39 | |
| *** Guest18637 has quit IRC | 09:41 | |
| *** kbyrne has quit IRC | 09:43 | |
| *** kbyrne has joined #openstack-meeting-cp | 09:44 | |
| *** gnarld_ has joined #openstack-meeting-cp | 09:44 | |
| *** david-lyle has quit IRC | 09:46 | |
| *** coolsvap has quit IRC | 09:59 | |
| *** tommylikehu has quit IRC | 10:09 | |
| *** tommylikehu has joined #openstack-meeting-cp | 10:15 | |
| *** tommylikehu has quit IRC | 10:15 | |
| *** tommylikehu has joined #openstack-meeting-cp | 10:15 | |
| *** MarkBaker has joined #openstack-meeting-cp | 10:47 | |
| *** sdague has joined #openstack-meeting-cp | 11:02 | |
| *** MarkBaker has quit IRC | 11:53 | |
| *** tang has joined #openstack-meeting-cp | 12:07 | |
| tang | hello? | 12:07 |
|---|---|---|
| *** coolsvap has joined #openstack-meeting-cp | 12:10 | |
| *** edmondsw has joined #openstack-meeting-cp | 12:14 | |
| tang | 你好啊 | 12:14 |
| *** tang has left #openstack-meeting-cp | 12:17 | |
| *** edmondsw has quit IRC | 12:20 | |
| *** edmondsw has joined #openstack-meeting-cp | 13:18 | |
| *** markvoelker has quit IRC | 13:23 | |
| *** markvoelker has joined #openstack-meeting-cp | 13:24 | |
| *** margaret has joined #openstack-meeting-cp | 14:03 | |
| *** coolsvap has quit IRC | 14:19 | |
| *** iyamahat has joined #openstack-meeting-cp | 15:00 | |
| *** yamahata has joined #openstack-meeting-cp | 15:00 | |
| *** zhipeng has joined #openstack-meeting-cp | 15:19 | |
| *** MarkBaker has joined #openstack-meeting-cp | 15:21 | |
| *** zhipeng has quit IRC | 15:22 | |
| *** gouthamr has joined #openstack-meeting-cp | 15:27 | |
| *** nikhil has joined #openstack-meeting-cp | 15:42 | |
| *** mriedem has joined #openstack-meeting-cp | 15:48 | |
| ildikov | #startmeeting cinder-nova-api-changes | 16:00 |
| openstack | Meeting started Thu Nov 30 16:00:15 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 |
| openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 16:00 |
| *** openstack changes topic to " (Meeting topic: cinder-nova-api-changes)" | 16:00 | |
| openstack | The meeting name has been set to 'cinder_nova_api_changes' | 16:00 |
| ildikov | johnthetubaguy jaypipes e0ne jgriffith hemna mriedem patrickeast smcginnis diablo_rojo xyang1 raj_singh lyarwood jungleboyj stvnoyes | 16:00 |
| jungleboyj | @! | 16:00 |
| _pewp_ | jungleboyj ( ^_^)／ | 16:00 |
| * johnthetubaguy hides at back of the room | 16:01 | |
| * ildikov sees johnthetubaguy :) | 16:01 | |
| * jungleboyj pulls johnthetubaguy to the front | 16:01 | |
| jungleboyj | Nice spot for you here buddy. | 16:01 |
| mriedem | o/ | 16:01 |
| * johnthetubaguy doh! | 16:01 | |
| ildikov | ok, let's start | 16:02 |
| * smcginnis hides behind johnthetubaguy | 16:02 | |
| ildikov | so we have the last bit of the new flow enablement in Nova: https://review.openstack.org/#/c/330285/ | 16:03 |
| ildikov | I was begging for review on the Nova meeting today | 16:03 |
| mriedem | need to re-run my tempest patch against that to make sure we're failing if trying to attach the same volume to the same instance | 16:04 |
| mriedem | twice | 16:04 |
| mriedem | https://review.openstack.org/#/c/515426/ | 16:04 |
| ildikov | although I think I need to rely on mriedem and johnthetubaguy with that or maybe I can ask jaypipes too | 16:04 |
| ildikov | mriedem: right, thanks for the reminder | 16:05 |
| johnthetubaguy | ildikov: don't think there were many open questions now, trying to remember the hour or so I spent going through the patch | 16:05 |
| ildikov | johnthetubaguy: we must be close now, I did some polishing on the functional tests too by now | 16:06 |
| johnthetubaguy | I was worried about left over attachments in a few places, but they didn't look like new bugs (missing unreserve call too) | 16:07 |
| johnthetubaguy | otherwise felt close I think | 16:07 |
| ildikov | yep, I remember that one, we can fix it in a follow up patch | 16:07 |
| *** MarkBaker has quit IRC | 16:08 | |
| ildikov | I would not want to increase the size of this one if we can avoid that | 16:08 |
| ildikov | we can figure out the remaining nits on the review | 16:09 |
| ildikov | I have the libvirt patch for multi-attach on top of this one, we can get that landed too if we agree on a few nits as it doesn't do any harm but will help enabling multi-attach | 16:10 |
| smcginnis | ildikov: Looks like the patch at least needs a rebase. Just waiting to see if there are any other changes? | 16:10 |
| smcginnis | The flow patch, not the libvirt one. | 16:11 |
| ildikov | smcginnis: we were talking about doing one more version bump once the shared_targets changes land | 16:11 |
| mriedem | ildikov: have you run my tempest test for multiattach on your nova multiattach patch? i'm assuming that tempest test is in need of cleaning up now though. | 16:11 |
| ildikov | smcginnis: this one: https://review.openstack.org/#/c/520676/ | 16:11 |
| ildikov | and jgriffith promised a follow-up patch with the API changes and the microversion bump | 16:12 |
| smcginnis | ildikov: Ah, OK. Maybe jungleboyj can take a look at that (hint, hint) | 16:12 |
| * jungleboyj will look. | 16:12 | |
| ildikov | mriedem: nope, the multi-attach changes need some update now to match the new flow | 16:12 |
| smcginnis | Although a couple good questions jgriffith should probably respond to first. | 16:12 |
| jgriffith | ildikov: if/when the shared_targets implementation lands yes | 16:13 |
| ildikov | smcginnis: yeah, that kept me back from begging for review again :) | 16:13 |
| ildikov | jgriffith: plz check the two minor comments on that one :) | 16:13 |
| ildikov | jgriffith: so I can annoy people to land it then | 16:13 |
| jungleboyj | Ah, the shared targets one. Cool. | 16:13 |
| ildikov | mriedem: the libvirt change only sets the 'shareable' flag, so it doesn't enable multi-attach right away | 16:14 |
| ildikov | mriedem: johnthetubaguy: as for multi-attach, jgriffith has a spec on the Cinder side for that: https://review.openstack.org/#/c/523608/ | 16:15 |
| ildikov | mriedem: johnthetubaguy: plz give it a quick read so we don't miss anything critical | 16:16 |
| mriedem | so simply creating a volume with multiattach=True isn't good enough? | 16:17 |
| ildikov | jungleboyj: smcginnis: quick question, who could help us with the policy changes in Cinder? | 16:17 |
| jungleboyj | The spec looked pretty good yesterday. I will look through it again now that it is updated. | 16:17 |
| jgriffith | mriedem: I'd prefer not to do that if we can avoid it | 16:17 |
| smcginnis | ildikov: Policy as far as adding a policy check for this? That's just done in the implementation now. Is that what you're asking? | 16:17 |
| jgriffith | mriedem: using the type and capabilities filter makes that a bit more solid | 16:18 |
| jungleboyj | ildikov: Good question. Need a volunteer there. | 16:18 |
| jgriffith | mriedem: and it allows you to update/modify a volume after the fact | 16:18 |
| * jungleboyj smiles at ildikov and jgriffith | 16:18 | |
| jungleboyj | jgriffith: ++ | 16:18 |
| ildikov | smcginnis: I think so, I'm one of the least knowledgable people around as far as policy goes :( | 16:18 |
| jgriffith | mriedem: there's pros/cons to either; but using types fits a bit more with our existing workflows and requriements so I thought it was a good fit | 16:19 |
| smcginnis | jgriffith: Helps with scheduler ability to place it too I think. | 16:19 |
| jungleboyj | smcginnis: Right | 16:19 |
| mriedem | that's fine, it's basically a flavor extra spec | 16:19 |
| jungleboyj | You could then have multi-attach capable and incapable backends. | 16:19 |
| smcginnis | ildikov: Yeah, no worries. With the policy-in-code changes, we just need to add the policy checking in the code for those operations. | 16:19 |
| jgriffith | smcginnis: yes for sure; you can make the flag set the capability but that's another story :) | 16:19 |
| jungleboyj | The scheduler would get the right onw. | 16:19 |
| ildikov | smcginnis: oh, ok, that does not sound too bloody, but we'll see I guess :) | 16:20 |
| jgriffith | smcginnis: ildikov I had just assumed that's part of the code, not sure what's being asked WRT to help on that | 16:21 |
| jgriffith | ^^ what I mean is part of the job for whomever implements the spec/changes | 16:21 |
| jungleboyj | Right. | 16:22 |
| *** MarkBaker has joined #openstack-meeting-cp | 16:22 | |
| ildikov | jgriffith: if that person is me, then I'll need at least someone to ask dumb questions and still get sane answers :) | 16:22 |
| smcginnis | ildikov: Yeah, should be enough existing examples that it won't be too bad. | 16:23 |
| ildikov | smcginnis: sounds good | 16:23 |
| jungleboyj | ildikov: Right. There are lots of policy examples. | 16:23 |
| *** MarkBaker has quit IRC | 16:24 | |
| johnthetubaguy | jgriffith: do you allow changing of the volume while it is attached? | 16:24 |
| *** MarkBaker has joined #openstack-meeting-cp | 16:24 | |
| jgriffith | johnthetubaguy: so that's an interesting one..... retype does allow that sort of thing if it does not include a migration | 16:25 |
| johnthetubaguy | so that would be a problem on the Nova side | 16:25 |
| jgriffith | johnthetubaguy: that's probably worth detailing and clearing up in the spec | 16:25 |
| jgriffith | johnthetubaguy: how so? | 16:25 |
| jgriffith | johnthetubaguy: I mean being a problem on the Nova side? | 16:26 |
| johnthetubaguy | we need to tell libvirt to stop its caching when its multi-attach | 16:26 |
| johnthetubaguy | we can't do that if its already attached | 16:26 |
| jgriffith | johnthetubaguy: ahh, right... on *all* of the attachments right | 16:26 |
| ildikov | jgriffith: yes | 16:26 |
| smcginnis | johnthetubaguy: No way to change that setting while attached? | 16:27 |
| jgriffith | johnthetubaguy: I'll look into that; we may end up back at using a flag on create again then :( | 16:27 |
| jgriffith | johnthetubaguy: where did things on the extend attached volume end up on the Nova side? | 16:27 |
| johnthetubaguy | its fine if you can update the flag only when there are no attachment records I guess | 16:27 |
| smcginnis | Just, from the Cinder side, it's very simple to just retype to enable/disable multiattach this way. | 16:28 |
| jgriffith | johnthetubaguy: we could just use the same mechanism there maybe? | 16:28 |
| mriedem | "where did things on the extend attached volume end up on the Nova side?" not sure what that means | 16:28 |
| johnthetubaguy | so detach can fail, when the volume is in use, so its not easy on the Nova side | 16:28 |
| johnthetubaguy | mriedem: I think the answer is we never merged that code? | 16:28 |
| jgriffith | mriedem: there was a patch a while back to allow extending an attached volume | 16:28 |
| mriedem | it was merged in pike | 16:29 |
| jgriffith | mriedem: said operation required a disconnect/reconnect on the Nova side IIRC | 16:29 |
| *** MarkBaker has quit IRC | 16:29 | |
| mriedem | https://review.openstack.org/#/c/454322/ | 16:29 |
| jgriffith | mriedem: thanks | 16:29 |
| *** aselius has joined #openstack-meeting-cp | 16:30 | |
| johnthetubaguy | ah... | 16:30 |
| ildikov | hmm, so are we saying now that a volume retype in Cinder would trigger hacking the libvirt domain xml to set the 'shareable' flag while the volume is attached? | 16:31 |
| johnthetubaguy | would that work? | 16:32 |
| jgriffith | ildikov: it's an option... johnthetubaguy ildikov I don't think it's worth doing at the onset though | 16:33 |
| jgriffith | another option is to put login in Cinder to check that condition and disallow it at the API layer | 16:33 |
| ildikov | johnthetubaguy: I don't know, I never checked whether or not it is possible | 16:33 |
| jgriffith | grrr | 16:33 |
| jgriffith | s/login/logic/ | 16:34 |
| johnthetubaguy | seems a good starting point | 16:34 |
| jgriffith | and make that a policy as well so if we change it some day we have that ability | 16:34 |
| smcginnis | So we need to block retype if a) it's attached, and 2) the retype include a multiattach capability change? | 16:34 |
| jgriffith | smcginnis: correct | 16:34 |
| jgriffith | smcginnis: that's what I'm thinking for now at least | 16:34 |
| ildikov | jgriffith: sounds like a good start to me | 16:34 |
| smcginnis | Bleh. OK, that works. | 16:34 |
| jgriffith | smcginnis: it's pretty easy to infer that on the Cinder side | 16:34 |
| jgriffith | smcginnis: yeah... not great, but not the worst either | 16:35 |
| smcginnis | Yeah, at least it won't involve a too much hackery. | 16:35 |
| jungleboyj | It seems like a reasonable start. | 16:35 |
| jungleboyj | Would rather limit the function than hack up Nova. | 16:35 |
| jungleboyj | It still gives us better all-around function. | 16:36 |
| ildikov | I wouldn't be comfortable with the hackery on the Nova side now, so if these restrictions would work in Cinder that made me happy for now | 16:36 |
| jgriffith | we used to disallow retype of in-use volumes | 16:38 |
| jgriffith | anyway... I'll look at it and see what we can/can't do | 16:38 |
| jgriffith | if nothing else fall back to using an option on create | 16:38 |
| ildikov | jgriffith: sounds good, thank you | 16:39 |
| johnthetubaguy | option on create that can be updated when there are no attachments? | 16:40 |
| jgriffith | johnthetubaguy: no | 16:40 |
| jungleboyj | jgriffith: Sounds good. | 16:40 |
| jgriffith | johnthetubaguy: if it's an option on create it's static for the life of the volume | 16:40 |
| jgriffith | johnthetubaguy: if you want to change then you have to do something like create-from-volume with that option | 16:40 |
| johnthetubaguy | OK | 16:41 |
| jgriffith | johnthetubaguy: I'm saying that if restricting the retype of an in-use volume looks hacky/ugly then revisit the use of a create flag | 16:41 |
| ildikov | jgriffith: johnthetubaguy: we talked about changing the flag if the volume is not attached, I guess we can iterate on that later if we would fall back to the flag | 16:41 |
| jgriffith | ? | 16:41 |
| smcginnis | That's the nice thing with using volume type. It let's us switch that without forcing the user to do stupid things. | 16:41 |
| *** aselius has quit IRC | 16:42 | |
| jgriffith | ildikov: well if you're going to do the same monkey trick with a create flag then there's zero point in going that route as opposed to a type | 16:42 |
| *** aselius has joined #openstack-meeting-cp | 16:42 | |
| ildikov | jgriffith: the request to change the flag or type or whatever we choose came up earlier | 16:43 |
| jgriffith | ildikov: what? | 16:43 |
| jungleboyj | jgriffith: Retyping versus having to clone a volume though, far less obvious to the end user as far as how to get the desired result. | 16:43 |
| jgriffith | ildikov: if you're talking about using a create flag, and then saying implement something to allow that to be changed then I'm saying that's a horrible idea and I'm back to just doing type/retype | 16:44 |
| ildikov | jgriffith: so I think it'll come up again once we have the feature enabled, so better to figure out now what we are willing or not willing to/can or can't do | 16:44 |
| jgriffith | and won't consider the flag option, because you're proposing putting in the same logic checks that would need to go into retype which is an accepted/general solution to these sorts of things already | 16:44 |
| ildikov | jgriffith: type sounds good if we can block that for attached volumes for now | 16:44 |
| jgriffith | ildikov: yeah | 16:44 |
| ildikov | jgriffith: then we're on the same page? :) | 16:45 |
| jungleboyj | Sounds like it. | 16:45 |
| ildikov | jgriffith: anyway, I don't want to argue, just remember having a proposal in Cinder to update the multiattach flag a while back, so if type/retype will be the solution for that with care, I'm all supportive; if we have other ideas that would be fine with me too | 16:48 |
| ildikov | jgriffith: what fits into Cinder the best | 16:49 |
| ildikov | so to sum it up | 16:49 |
| ildikov | I will rebase the new attach patch in Nova and re-run the tests a couple of times to see whether the blocking to attach the same volume to the same instance works fine | 16:50 |
| ildikov | then get that patch merged finally if everything works well | 16:50 |
| ildikov | in Cinder let's have the shared_targets patch merged and have the microversion bump in place | 16:51 |
| *** MarkBaker has joined #openstack-meeting-cp | 16:51 | |
| ildikov | update the Cinder spec with the type/retype restrictions we've talked about just now | 16:51 |
| ildikov | and that will cover a weekly work I think easily | 16:51 |
| ildikov | did I miss anything? | 16:51 |
| jungleboyj | ildikov: That sounds good to me. | 16:53 |
| ildikov | jungleboyj: cool :) | 16:54 |
| ildikov | mriedem: johnthetubaguy: jgriffith: plz let me know if I missed any critical item from the list above | 16:54 |
| ildikov | anything else to discuss today? | 16:54 |
| jgriffith | ildikov: sounds like you got it | 16:54 |
| ildikov | jgriffith: cool, thanks | 16:55 |
| ildikov | ok, then thanks everyone for joining today! | 16:57 |
| ildikov | let's get some progress by the meeting next week!! :) | 16:57 |
| ildikov | have a good rest of the day everyone! | 16:57 |
| ildikov | BTW, I know it's KubeCon next week | 16:57 |
| ildikov | will we still have the majority of people available? | 16:58 |
| jungleboyj | I will be around. | 16:58 |
| jungleboyj | Wanted to go but couldn't work it out. | 16:58 |
| ildikov | johnthetubaguy: mriedem: jgriffith: smcginnis: ^^? | 16:58 |
| ildikov | jungleboyj: sadness, next time I guess | 16:58 |
| jungleboyj | ildikov: Indeed. I am sure jgriffith and smcginnis will be good reps. | 16:59 |
| smcginnis | I'll be at Kubecon, so probably won't be able to attend. | 16:59 |
| *** david-lyle has joined #openstack-meeting-cp | 16:59 | |
| ildikov | ok, will chat with the Nova guys then I guess :) | 16:59 |
| ildikov | thanks again | 16:59 |
| ildikov | let's move the further chats to the channels | 17:00 |
| jungleboyj | Sounds good. | 17:00 |
| ildikov | #endmeeting | 17:00 |
| *** openstack changes topic to "OpenStack Meetings || https://wiki.openstack.org/wiki/Meetings" | 17:00 | |
| openstack | Meeting ended Thu Nov 30 17:00:14 2017 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 17:00 |
| openstack | Minutes: http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2017/cinder_nova_api_changes.2017-11-30-16.00.html | 17:00 |
| openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2017/cinder_nova_api_changes.2017-11-30-16.00.txt | 17:00 |
| openstack | Log: http://eavesdrop.openstack.org/meetings/cinder_nova_api_changes/2017/cinder_nova_api_changes.2017-11-30-16.00.log.html | 17:00 |
| *** mriedem has left #openstack-meeting-cp | 17:01 | |
| *** nhelgeson has joined #openstack-meeting-cp | 17:06 | |
| *** iyamahat has quit IRC | 17:07 | |
| *** yamahata has quit IRC | 17:08 | |
| *** kmalloc has quit IRC | 17:11 | |
| *** iyamahat has joined #openstack-meeting-cp | 17:33 | |
| *** MarkBaker has quit IRC | 17:36 | |
| *** yamahata has joined #openstack-meeting-cp | 17:51 | |
| *** nikhil has quit IRC | 17:51 | |
| *** diablo_rojo has joined #openstack-meeting-cp | 18:22 | |
| *** felipemonteiro has joined #openstack-meeting-cp | 19:17 | |
| *** kmalloc has joined #openstack-meeting-cp | 20:07 | |
| *** edmondsw has quit IRC | 22:24 | |
| *** edmondsw has joined #openstack-meeting-cp | 22:24 | |
| *** edmondsw has quit IRC | 22:25 | |
| *** edmondsw_ has joined #openstack-meeting-cp | 22:27 | |
| *** edmondsw_ has quit IRC | 22:31 | |
| *** felipemonteiro has quit IRC | 22:44 | |
| *** felipemonteiro has joined #openstack-meeting-cp | 23:02 | |
| *** iyamahat_ has joined #openstack-meeting-cp | 23:22 | |
| *** iyamahat has quit IRC | 23:22 | |
| *** MarkBaker has joined #openstack-meeting-cp | 23:32 | |
| *** felipemonteiro has quit IRC | 23:38 | |
| *** MarkBaker has quit IRC | 23:54 | |
| *** felipemonteiro has joined #openstack-meeting-cp | 23:56 | |
| *** felipemonteiro_ has joined #openstack-meeting-cp | 23:57 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!