16:00:22 #startmeeting Cinder 16:00:23 Meeting started Wed Mar 21 16:00:22 2018 UTC and is due to finish in 60 minutes. The chair is jungleboyj. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:24 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:27 The meeting name has been set to 'cinder' 16:00:29 Hello! 16:00:43 Courtesy ping: jungleboyj DuncanT diablo_rojo, diablo_rojo_phon, rajinir tbarron xyang xyang1 e0ne gouthamr thingee erlontpsilva patrickeast tommylikehu eharney geguileo smcginnis lhx_ lhx__ aspiers jgriffith moshele hwalsh felipemonteiro lpetrut 16:00:49 hi! o/ 16:00:52 hi 16:00:54 hi 16:00:57 .o/ 16:01:01 hi 16:01:07 @! 16:01:09 hi 16:01:44 Hello everyone. 16:01:49 hi 16:02:05 Don't have a large agenda today so I am sure we will run out of time. 16:02:23 #link https://etherpad.openstack.org/p/cinder-rocky-meeting-agendas 16:02:44 e0ne: Thank you. 16:02:48 o/ 16:03:08 smcginnis: Is here. Now I can start. ;-) 16:03:41 :P 16:03:44 #topic announcements 16:04:04 So, first announcement is a reminder that we have an etherpad for Forum Topic planning: 16:04:04 o/ 16:04:18 #link https://etherpad.openstack.org/p/YVR-cinder-brainstorming 16:04:33 So far the only ideas in there are the ones I have put. Anyone have proposals? 16:04:59 * jungleboyj hears crickets 16:05:26 I will need to be proposing things soon, so if you have topics, please add them. 16:06:12 So, I also wanted to note that I have seen an uptick in reviews from some people from the PTG. 16:06:16 Thank you for the help! 16:06:36 It is appreciated. We have been able to get a new driver merged and make progress on our specs. 16:06:43 Think it is a good start for Rocky. 16:07:10 So we are not off to a Rocky start? *chortle* 16:07:34 *sad_trombone.wav* 16:07:51 :) 16:07:55 * jgriffith throws a mule at Swanson 16:08:02 jgriffith: ++ 16:08:07 followed by a bag of snakes 16:08:29 So, also, just a reminder to register for the Vancouver Summit if you are able to go. Hope to see a number of you there. 16:08:57 :-) 16:09:28 Ok, moving on before jgriffith Gets really violent 16:09:41 #topic Rocky Priorities review ... 16:09:58 #link https://etherpad.openstack.org/p/cinder-spec-review-tracking 16:10:06 Not a lot of activity in here this week. 16:10:20 e0ne: Any luck getting the spec for generic backup updated? 16:10:24 Sorry if I missed it. 16:10:49 jungleboyj: I'm working on it. hope to finish it in next two days 16:11:32 e0ne: Sounds good. I don't think I have seen reviews come through for other patches that need to be added to the list. 16:11:45 e0ne: eharney smcginnis Appreciate help getting stuff added in there if you see it. 16:13:02 Ok, don't have much more to dig into there then. 16:13:20 #topic HA Check-in 16:13:31 geguileo: Do you have HA all working and tested? 16:13:34 * jungleboyj giggles 16:14:16 XD 16:14:31 geguileo: Any progress to report? 16:14:57 I've just started looking at an HA A-A replication issue that Jon Bernard brought to my attention 16:15:14 I'll start working on the document after I fix those issues 16:15:46 geguileo: That makes sense. 16:16:23 It is good we are finding bugs and getting those fixed before more people start using it. 16:17:05 geguileo: Thank you for continuing to push that. 16:17:20 #topic New Driver Support Matrix ... 16:17:48 So, I am working on putting together the new feature-classification based version of the Support Matrix 16:18:07 This is a mind numbing task but I think the result is going to be great. 16:18:45 I am finding that the old matrix on the Wiki is very out of date with drivers listed that don't exist anymore. 16:18:58 I am also find that many drivers are completely undocumented. 16:19:00 jungleboyj: will it be autogenerated from sources? 16:19:06 I think that has been very poorly maintained for some time. 16:19:12 * jungleboyj glares at the Huawei people 16:19:32 e0ne: No, but it will be generated from an INI file that is easily maintained. 16:19:50 When a driver changes we just need to make sure the change also updates the matrix. 16:19:56 jungleboyj: sounds good 16:20:01 nice 16:20:03 Much better than the current process. 16:20:15 Just getting it created is a bit of a nightmare. 16:20:31 I could see this eventually being groundwork for automatically generating that ini source, but we probably have a lot of work to do first before we can get there. 16:20:50 smcginnis: ++ 16:20:53 Baby steps. 16:21:03 So, my question to the team is as follows. 16:21:25 A number of drivers have, in the Wiki, multiple entries for one driver based on model number. 16:21:54 Dell/EMC and Huawei ... Is that something I should be reflecting in what I am putting in this new support matrix? 16:22:27 There may be differences. 16:22:40 jungleboyj: sadly it seems like those are different devices/drivers so you probably will have to 16:22:41 Maybe put them all in for now, then we can see if we can clean it up some once it's there? 16:22:55 I agree with jgriffith 16:23:03 different entries 16:23:19 jgriffith: :-( ok. 16:23:30 I was guessing that was the answer. 16:23:50 smcginnis: I am going to need some help mapping the Huawei entries in the Wiki to drivers. 16:24:07 jungleboyj: I probably will too. 16:24:16 tommylikehu: ^^ 16:24:23 jungleboyj: Maybe we can get some... yeah, tommylikehu. :) 16:24:25 jungleboyj: sorry Charlie 16:24:31 :) 16:25:01 jgriffith: It is ok. This is going to be a great improvement once done. 16:26:18 Ok, so, I will get all the different models represent. Probably going to be another week or so before I have a WIP pushed up for people to look at. 16:26:49 Then will need some help from people to verify it for Vendor drivers. 16:27:43 Any questions or comments there? 16:28:47 Thanks for working on that. 16:28:59 smcginnis: Welcome. It will be good. 16:29:12 #topic Open Discussion 16:29:23 That was all I had for this week. Anything else people want to talk about? 16:30:09 Is tommylikehu Not here? 16:30:18 Working on trying to get cinderclient python-bindings to actually work 16:30:20 He is pushing the backup across AZ issue. 16:30:34 jgriffith: How so? 16:30:36 If anybody else has ever gone down that rabbit hole give me a ping 16:30:47 jungleboyj: it doesn't work currently/still/again 16:31:00 * jungleboyj has obviously not been down that rabbit hole. 16:31:06 can't authenticate 16:31:21 jungleboyj: It's not backup across AZs, if you are referring to the latest ML discussion. 16:31:35 we have an outdated example in the docs still and on the github readme: https://github.com/openstack/python-cinderclient 16:31:36 smcginnis: Shared backups? 16:31:48 jungleboyj: There's some customer desire (or a customer has been told this would be the way to do it) for sharing a backup. 16:32:32 smcginnis: Thank you. Yes, I misspoke there. 16:32:47 jgriffith: Ok. Wasn't aware that was broken. Thanks for taking a look at it. 16:32:55 So it could be across AZs, but it is also across owners. 16:33:24 jungleboyj: it's "my pleasure" 16:33:27 ok.. not really 16:33:28 smcginnis: What are your thoughts there? 16:33:33 jgriffith: What's the issue? It can't authenticate as a library, or even as a CLI? 16:33:35 jgriffith: You aren't selling me there. 16:33:49 smcginnis: as a lib 16:34:21 jungleboyj: My feedback was pretty much what has been stated. That's not what backup was intended for, no matter the similarities to AWS snapshots. Glance images sound like the more correct solution for OpenStack. 16:34:26 something with our client init and newer versions of keystone 16:34:28 jgriffith: That's not good. 16:34:44 keystoneauth changes? 16:34:44 tried using the keystone v1 session method as well, and ran into problems there. 16:35:12 Yeah, most likely; but I think *old* Keystone changes... not sure 16:35:15 smcginnis: ++ Yeah, using Glance makes more sense. 16:35:32 lbragstad: Any input here ^^^ 16:35:41 hopefully I'll figure it out today and I can update our docs 16:35:54 Sounds good. 16:35:56 jungleboyj: Err, I mean, we should definitely share backups. It's what the customer wants. 16:35:59 :) 16:36:02 * lbragstad reads scrollback 16:36:16 Thanks to everyone for making doc changes and dealing with me harping on that. 16:36:23 lbragstad: Mix discussions. The parts about authentication not working for cinderclient as a lib. 16:36:28 smcginnis: You also are not selling me on that statement. 16:36:36 jungleboyj: :) 16:36:58 * jungleboyj is all for making customers happy but not increasing complexity. 16:37:01 jgriffith: ah - are you using keystoneauth with cinder client? 16:37:42 lbragstad: *trying* to 16:37:59 but in general just trying to use python bindings to init a Client 16:38:08 ok - what does cinderclient use now to do authentication? 16:38:19 keystone 16:38:27 smcginnis: it's shared backups or transferring backups? 16:38:37 geguileo: Shared. 16:38:37 geguileo: Shared. 16:38:38 jgriffith: like - directly? 16:38:52 geguileo: So I agree with your last bit there, but that's not what he's asking for. 16:39:05 lbragstad: there are two options... the first is direct via Client.__init__ 16:39:08 Like snapshots, it may make sense to be able to transfer backups. 16:39:08 smcginnis: we don't allow sharing volumes, right? 16:39:22 geguileo: Right. They have to transfer them. 16:39:27 The other is going out to auth_v1 and creating a session, then trying to feed that in to Client init 16:39:33 smcginnis: then same thing for backups 16:39:41 So really the only "sharing" is glance images, AFAIK. 16:39:58 smcginnis: as it should be 16:40:04 geguileo: I agree. I mean no, I don't. Yes I do. I'm trapped! :) 16:40:11 smcginnis: He he he 16:40:12 lbragstad: kinda like this: https://docs.openstack.org/python-novaclient/pike/reference/api/index.html 16:40:18 * smcginnis hopes his boss doesn't read this 16:40:27 lbragstad: main difference being Nova works, Cinder doesn't :) 16:40:32 oh - sure 16:40:34 smcginnis: Don't worry, Lenovo is hiring. 16:40:41 jungleboyj: LOL 16:41:02 smcginnis: I don't know if there would be a way to share volumes using the upload to glance in cinder that actually creates a cinder volume 16:41:42 jgriffith: so - you are using ksa to get auth, and build a session, but when you pass that to clientclient.client.Client, it blows up? 16:41:50 smcginnis: Does Glance howere solve his problem of having the images available in multiple AZs. 16:42:01 smcginnis: I haven't tested it, but they could use image_upload_use_cinder_backend 16:42:16 smcginnis: They probably do... 16:42:19 lbragstad: correct, or if I try and use the other method and feed OS vars into the init that fails to auth 16:42:27 jgriffith: do you have a trace from cinderclient? 16:42:31 in other words, Cinder Client is just not right 16:42:40 * jungleboyj is loving the split brain meeting here. 16:42:54 I don't but I can ping you later with info if you want to have a look and help out? 16:43:18 jgriffith: sure - getting clients to consume sessions from ksa should be a pretty established pattern 16:43:18 jgriffith: I think my conversation with smcginnis is over now, since I'm ok to do transfers, but not sharing 16:43:32 Me too. 16:44:04 geguileo: smcginnis So we are in agreement that we still say no to that request and then smcginnis will have terrets afterwards. 16:44:06 jgriffith: i know we decoupled that approach keystoneclient a while back when we moved everything to ksa (since keystoneclient used to build it's own sessions) 16:44:20 approach with* 16:44:33 jungleboyj: I don't want shared backups for sure 16:45:19 That would seem to break the security model for Volumes as the backups are of Volumes. 16:47:07 Ok. So, things have quited down. 16:47:14 lbragstad: Thanks for helping out jgriffith 16:47:30 yeah - no problem, let me know if you get a trace and i can take a look 16:47:41 smcginnis: geguileo I think we will stand with the decision from the PTG on shared backups. 16:47:53 jungleboyj: +1 16:47:55 Any other topics for today? 16:49:21 I will take that as a no. 16:49:41 Thanks to everyone for joining the meeting today. 16:49:51 Thanks for your work on Cinder! 16:49:57 Talk to you all next week! 16:50:06 #endmeeting