09:00:19 #startmeeting blazar 09:00:19 Meeting started Tue Dec 6 09:00:19 2016 UTC and is due to finish in 60 minutes. The chair is priteau. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:00:20 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:00:22 The meeting name has been set to 'blazar' 09:00:33 Greetings everyone 09:00:37 Hi 09:00:45 hi 09:00:46 Welcome to our first Blazar meeting in this new series 09:00:47 Hi 09:00:51 Hello 09:00:56 #chair masahito hiro-kobayashi 09:00:57 Current chairs: hiro-kobayashi masahito priteau 09:00:57 \o 09:01:09 #link https://wiki.openstack.org/wiki/Meetings/Blazar#Agenda_for_December_6_2016 09:01:16 This is the agenda for today 09:01:33 #topic Roll call and quick introduction of participants 09:02:16 Because this is the first meeting and the community is still forming, I thought we could introduce each other 09:02:30 nice idea 09:02:39 But quickly (one line) to give time for technical discussion 09:02:47 I'll go ahead 09:03:34 I am Pierre Riteau, working at University of Chicago on Chameleon (testbed for CS research). We have been using Blazar in production since summer 2015 and added a bunch of bug fixes / new features 09:04:11 Who else is here? 09:04:11 #info Gerald Kunzmann from DOCOMO Eurolabs in Munich, Germany. Committer in OPNFV Promise project. Interested in Blazar as it is a good candidate to address the requirements for resource reservation. 09:04:43 I'm Hiroaki Kobayashi from NTT. I'm a contributor of OPNFV Promise project which requires resource reservation. I started to improve Blazar since last month. 09:05:13 I'm Toshiyuki Moriya for NTT, co-worker of masa and hiro-kobayashi. Tnx for ur arrangement. It's my first time to join Openstack IRC. 09:05:21 I'm Masahito Muroi, cloud architect in NTT and a member of core team of OpenStack Congress project. Now I'm looking at blazar project as our NFV infrastracure manangement. 09:05:33 I am Bertrand Souville from DOCOMO Euro-Labs in Munich. Contributor of OPNFV Promise project. 09:05:50 Sylvain Bauza from Red Hat, worked in a former life as a Blazar core and incepted the project, now fully dedicated to Nova as a nova-core 09:06:00 o/ 09:06:27 I am Janki from RedHat. I contribute to Tacker 09:06:37 toshi_: I don't think we've met, good to see more people involved! 09:07:01 Hello janki 09:07:17 hi janki 09:07:22 priteau, masahito hey 09:07:33 janki: Is Tacker part of OPNFV too? 09:08:05 priteau, yes. NFV is based on ETSI MANO architecture 09:08:19 it falls under openstack project though 09:08:34 janki: IIRC, Tacker is one of openstack project. 09:08:50 Excellent. Anyone else? 09:09:19 masahito, yes. it is openstack project but is based on ETSI MANO architecuture and works closely with that community too 09:09:46 janki: got it. thanks :-) 09:09:48 There is also Ryota Mibu from NEC who couldn't join us today because he is on a trip to the US 09:10:10 Usually Japan-based if I understand correctly 09:10:18 He should be able to join us at next meetings 09:10:18 priteau: yes 09:10:40 All right, if that's all for introductions let's move to the serious part 09:11:00 #topic Blazar status 09:11:12 #link https://etherpad.openstack.org/p/Blazar_status_2016 09:11:38 I created this Etherpad and added my own list of known issues to it. Others have been contributing as well 09:12:53 The goal is to develop a plan for getting Blazar back into a good state 09:13:55 I think there's a good place to discuss goal of Ocata. 09:14:57 Yes we need to agree on a reasonable goal for Ocata 09:16:26 My opinion is that it would be good for Ocata to have a Blazar that is compatible with the release and that we can tag with a version for people to use 09:16:28 lets list the items and start owning 09:16:33 that will speed up the work 09:16:53 do folks know the Blazar API ? 09:17:10 priteau: agreed. 09:17:21 because we have 2 ways of reserving resources, and one should probably be deprecated IMHO 09:17:44 bauzas: I am familiar with v1, not so much with v2 09:17:47 bauzas: it's meaning V1 API and V2 API? 09:17:48 bauzas: VM reservation should be deprecated, right? 09:17:56 hiro-kobayashi: yup 09:18:14 bauzas: Do you know if v2 is still relevant? 09:18:21 I think we honestly made a wrong assumption on what Blazar should cover 09:18:37 priteau: REST API v2 is the current API, and I don't see any problems using it 09:19:16 priteau: v1 could even probably be deprecated and eventually removed 09:19:38 We use v1 in Chameleon. I will have to double-check, but I thought v2 was never fully merged 09:20:10 priteau: there could be some features missing, maybe 09:20:15 that's old memories to me 09:20:37 yeah. I thought it's still under implementing when I see the codes 09:20:51 #action Review state of API versions: v1 vs v2 09:20:57 here, I just want to express that I'd love to see people tech refreshing Blazar and probably cutting things they consider no longer valuable 09:21:23 if not, you'd have tech debt to honor 09:22:10 I would be OK with deprecating virtual instance reservation 09:22:26 anyway, my opinion is thinking what we need in either v1 or v2 in the future. 09:22:27 and focusing on physical host reservation for now 09:22:43 masahito: that's a reasonable point 09:23:02 here, I think you need a couple of usecases to express and see what's missing for those 09:23:08 masahito: We should have only one API, otherwise it's going to be a pain to maintain 09:23:10 priteau: Perhaps OPNFV Promise may require VM reservation 09:23:12 From Promise side, we need instance reservation feature in the future. 09:24:01 priteau: I agree that we focus the host reservation first 09:24:10 FWIW, we wrote a few usecases here https://wiki.openstack.org/wiki/Blazar/Use_Cases 09:26:16 OPNFV Promise also have usecases here #link http://artifacts.opnfv.org/promise/colorado/3.0/docs/requirements/index.html#document-usecase 09:27:44 looks nice. but we could add real-usecase to make it more concrete, discuss it and then decide it. How do you think? 09:28:04 oh. for previous blazar usecase 09:29:31 So for our first new release, we will choose use cases that we can support and ensure that they are working 09:30:05 ++ 09:30:09 +1 09:30:12 +1 09:30:19 Obviously I would like to see compute host reservation 09:30:31 I think the best approach is to identify what's functional now as short-term and release 09:30:35 I think it can be well supported (it already is working quite well in the current implementation) 09:30:53 and then once we have this solid release, we can work on more long term features 09:31:20 I don't push the instance reservation is complete by O release as this cycle is and short sprint. 09:32:03 reservations of non-compute objects (like volumes or Heat stacks) will likely involve working with other OpenStack groups and might take a long time to be supported 09:32:08 priteau, masahito: I agree because VM reservation may need long time to be fixed 09:32:42 #agreed Focus on supported use-case for the first release: compute host reservation 09:32:45 we should make current implementation stable for short term goal first. 09:33:17 masahito: Blazar is not part of the released-managed projects so you don't need to feel tied to the OpenStack cycle for the moment 09:33:42 what you folks need to identify is a target date and a target state, iterate over that, and cut the release 09:33:51 bauzas: right, but if we had, we can easy to focus. 09:34:05 then, you could see the benefits of trying to align with the general cadence 09:34:35 I don't think we would follow the release cycle as in freeze, RCs, etc. 09:34:52 I would see bringing Blazar to the big tent as another priority for near to mid term 09:34:55 but it would be nice to say "there's the commit that we know works with ocata", and tag it so we can use it 09:35:09 GeraldK: that's not mutually exclusive 09:35:23 GeraldK: you can ask for being in the Big Tent without following the general release cadence 09:35:35 I mean following major release, not FF and so on 09:35:38 bauzas: okay, got it. 09:35:50 GeraldK: agreed. Demonstrating that we have a working system that follows OpenStack conventions is a big part of it 09:36:16 so probably goal for after our next release 09:36:53 OK, we're already more than half way through the meeting, I would like to go over the Etherpad list 09:37:26 first items are related to using Blazar with DevStack 09:37:48 this is almost fixed now, just one change in review 09:37:54 #link https://review.openstack.org/#/c/405630/ 09:38:44 I am hoping it can be merged this week 09:39:06 Next we have actually making it work :-) 09:39:38 :-) 09:40:03 #link https://review.openstack.org/#/c/398716/ Update the Blazar Nova filter to the new API since it causes errors in nova-scheduler 09:40:21 this was proposed by hiro-kobayashi and is under review 09:40:57 I made a note on the etherpad about ^ 09:40:58 as far as I know this is the only big problem with making the compute host reservation work 09:41:10 oops 09:41:12 wrong 09:41:32 nevermind my previous comment, I thought it was on different change 09:41:42 bauzas: that's something else, that's the compute host scheduler filter 09:41:49 and yeah, the Nova scheduler filter interface changed a lot 09:42:04 I can review that change 09:42:16 as I also made the change for the nova side 09:42:18 to merge the patch, the patch #link https://review.openstack.org/#/c/406009/ should be merged 09:42:33 bauzas: Thank you very much! 09:42:47 hiro-kobayashi pushed a new patch set a few hours ago, I haven't reviewed yet but the previous versions were looking good 09:43:41 hiro-kobayashi: how do 406009 and 398716 depend on each other? I thought they were completely independent? 09:43:58 priteau: 398716 got a Jenkins -1 09:43:58 or is it for fixing the unit tests? 09:44:05 right 09:44:30 398716 test fails because of other part 09:44:43 and 406009 fixes the part 09:44:55 hiro-kobayashi: could you maybe rebase 398716 on top of 406009 ? 09:45:05 so we could have a pile of reviews 09:45:07 but 406009 is just a workaround and VM reservation still doesn't work 09:45:14 #action bauzas to review https://review.openstack.org/#/c/398716/ 09:45:45 bauzas: OK I'll do that 09:45:49 #action hiro-kobayashi to rebase 398716 on top of 406009 09:46:19 what would be the procedure for deprecating VM reservation? 09:46:31 and yeah, I assume VM reservation can't work with Nova API v2.1 09:47:09 To enable the VM reservation, API extensions should be merged to Nova tree 09:47:18 priteau: as we don't have microversions in place, I'd suggest you to stabilize Blazar, raise a warning when people are using the VM reservation plugin, tag the release and then drop the code 09:47:48 hiro-kobayashi: as I said in the etherpad, API extensions are not functional with Nova v2.1 09:47:58 bauzas: that sounds like a good approach 09:48:15 hiro-kobayashi: and nova v2.1 is the default API version since 2 releases 09:48:50 but people could still be using an old Nova release with Blazar, that really depends on how you plan to integrate with the other projects 09:49:16 if you say that you want to support only N releases of nova for Blazar, then you can just drop the code now 09:50:18 Do we have evidence of anyone using VM reservation? 09:50:44 If it needs to come back in a different form later, it may be easier to drop it now 09:51:21 What do you think hiro-kobayashi? 09:52:28 We, me and hiro-kobayashi, are doing gap analysis between current VM reservation and Promise requirements. 09:52:42 I am OK with keeping the code as long as it doesn't slow us down too much 09:52:53 if possible, we want to keep the code in the repo. 09:53:09 priteau: thanks. 09:53:22 we need the some reference. 09:53:31 Are you OK with following bauzas approach for now? "stabilize Blazar, raise a warning when people are using the VM reservation plugin, tag the release" 09:53:42 and post-release we decide what happens with the code 09:53:50 +1 09:53:53 +1 09:54:06 +1 09:54:16 at least some very clear message that says that Blazar can't work with Nova /v2.1 endpoint if you wanna use VM reservation 09:54:18 #agreed VM reservation plugin: raise a warning when people are using it but keep the code for now. Revisit after our next release 09:54:55 #info at least some very clear message that says that Blazar can't work with Nova /v2.1 endpoint if you wanna use VM reservation 09:55:15 OK, only 5 minutes left so let's continue 09:55:32 two big items next on the list: convert to Oslo and fix Tempest 09:55:57 I'm working on fixing tempest test. 09:56:07 We are lucky to have other members of OpenStack helping out with oslo because it is a goal for O to get rid of all incubator code 09:56:16 I guess I could push the patch by this weekend. 09:56:20 So please review their changes 09:56:30 I am working on a patch to use oslo.db 09:56:40 I hope to finish it this weeek 09:56:52 masahito: great news 09:57:28 I think we know what we're doing on these points 09:57:50 we have discussed the Nova plugin already 09:57:57 migrate namespace climate to blazar 09:58:21 it would be good to finish before the release, but maybe after we get all the important fixes in? 09:58:29 to avoid all kind of merge conflicts 09:58:35 right. 09:59:00 +1 09:59:03 don't need to be hurry 09:59:21 OK 09:59:31 Finally status of v1 API vs v2, I can take as action item to review this 09:59:45 And I will share my findings at the next meeting 10:00:02 priteau: thanks! 10:00:19 great 10:00:54 We haven't had time to go over future improvements and new features 10:01:10 for next meeting's agenda? 10:01:19 +1 10:01:22 +1 10:01:22 +1 10:01:36 Let's focus on getting patches merged for this week 10:01:44 OK 10:02:00 let's also assign people to Ocata prio tasks in next week meeting 10:02:34 GeraldK: I will create an agenda for next meeting, feel free to edit it 10:02:49 Thanks everyone for joining us today! 10:03:21 #endmeeting