13:00:33 #startmeeting senlin 13:00:34 hi 13:00:34 Meeting started Tue Aug 4 13:00:33 2015 UTC and is due to finish in 60 minutes. The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:35 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:37 The meeting name has been set to 'senlin' 13:01:21 hello guys 13:01:24 hello 13:01:27 hello 13:01:28 o/ 13:01:29 hi 13:01:39 let's get started 13:02:01 please add items to the agenda if you have things to share 13:02:06 #link https://wiki.openstack.org/wiki/Meetings/SenlinAgenda 13:03:10 #topic liberty-3 milestone targets 13:03:33 refer to etherpad: https://etherpad.openstack.org/p/senlin-liberty-workitems 13:04:09 we still have some items left from l-2 13:04:34 yes, and we have 5 goals want to achieve 13:04:55 right, those five are about l-3 13:05:00 i am closing out on mine by tomorrow 13:05:29 if everything goes smooth, I think we should try propose to switch the namespace, ;) 13:05:50 that sounds fantastic 13:05:59 is trigger also one of our goals? I saw it in backlog but not in goal list 13:06:11 I hope it is :) 13:06:17 it is in backlog, should have been completed by l-2 13:06:25 ok 13:06:38 need more reviews, you know 13:06:46 right 13:07:07 switch the namespace means to be a openstack project? 13:07:17 the big things ahead: container clusters, which is the topic for the summit talk 13:07:33 yes, haiwei__ we should definitely try it 13:07:42 ok 13:08:05 another big thing is senlinclient unit tests 13:08:10 i am working on the container clusters. behind where i need to be, i have spent some time trying to figure out magnum 13:08:15 we have no a single line for that 13:08:45 thanks, jruano, that is an important show case 13:09:33 we are not supposed to only propose senlin server, leaving senlinclient on stackforge, that makes no sense 13:10:06 yep, so maybe we need to have a plan about the unit test of senlinclient 13:10:10 yes, so the client test should be started soon 13:10:16 maybe also file some bugs about it 13:10:25 as for placement policy, it would be part of our efforts to make senlin capable of supporting VM/app HA scenarios 13:10:50 +1 on filing some bugs now 13:11:11 this is one of the most important feature we want to support in the first release I think 13:11:27 the placement policy to support cross az/region node creation 13:11:52 it is important for two reasons 13:12:15 we have some commitments to demonstrate cross-region autoscaling 13:12:52 placement policy would be also a show case for the upcoming OpenNFV summit in San Francisco 13:13:25 that was the reason I'm putting it there in the list as an important goal 13:13:45 you will git a demo about it? 13:13:57 hopefully, Cindia can help us get that sorted out soon 13:14:07 yes, haiwei__ 13:14:10 we talked with a customer last week around tosca. they mentioned placement policy at the top of their list as well 13:14:32 it was a proposal submitted with China Mobile, VMware 13:15:03 just submitted today 13:15:15 Qiming, I think we may need to start work on this item asap to ensure we can give a prilimary implementation by the end of l-3 deadline 13:15:17 when is the summit 13:15:40 it's Nov 08-14? 13:15:59 i thought it was in october 13:16:10 tokyo is end of october this year 13:16:20 27-30 i think? 13:16:27 yes, hope no conflict 13:16:31 right 13:16:52 openstack summit Oct 27-30, OpenNFV summit Nov 08-14 13:18:06 if we have supports from carrier vendors, we are really on a good track to get our work utilized 13:18:16 HA policy is also needed as well as the placement policy? 13:18:36 placement policy is only part of the story, HA ... is a big topic 13:18:58 ok, good news for now :) 13:19:08 hopefully we can gain some traction, with placement policy as a starting point 13:19:25 we don't have to jump into HA part in next month 13:19:32 ok 13:19:35 again, I'm hoping that Cindia will lead that effort and get started soon 13:19:47 that will be the best :) 13:19:51 it is a separate thread 13:20:19 who is Cindia :) 13:20:22 do we need to prioritize these items? 13:20:27 lixinhui :) 13:20:33 haiwei__, Cindia-Blue, :) 13:20:50 ok, absent today? 13:21:05 no, Qiming, I think we just need to ensure every item has some guys to work on it 13:21:35 alright, sign your name after each item 13:21:46 then we can know whether we really have bandwidth to finish these items 13:22:06 it is okay to have more than one name signed on each item, right? 13:22:13 sure :) 13:22:47 I'm signing on the senlinclient test cases 13:22:57 yap, but don't conflict 13:23:39 sure, if two or more people are working on the same item, we need to sync more frequently 13:23:51 hope I can finish the functional test setup 13:24:03 btw, jruano, do you need help on the unit tests? 13:24:18 just noticed that you have claimed quite some of them 13:24:22 and provide tests for some most important cases 13:24:57 yanyanhu, that would be cool, giving both users and us more confidence on using the service 13:25:03 i think im good. i finished node_action, and started cluster_action 13:25:14 yes, that's our goal 13:25:27 the new ones i claimed yesterday were for drivers 13:25:46 okay, jruano, if the plate is too full for you, we can share, :) 13:26:10 let's move on 13:26:15 sure np. 13:26:21 #topic plan of functional testing 13:26:26 yanyanhu, 13:26:34 ok 13:26:49 now we plan to add functional test for Senlin service 13:27:24 currently, we only have unit test and all of them are located in senlin/tests 13:27:54 so we tried to relocate these existing tests and reorganize our test cases 13:27:57 https://review.openstack.org/#/c/208944/ 13:28:18 this patch move all unit tests into senlin/tests/unit directory 13:28:39 and new added functional test will be put into senlin/test/functional directory 13:29:13 so some of you guys may need to rebase some existing works on unit test :) 13:29:48 already get some report of merge failure, :) 13:29:57 yes ;) 13:30:27 first, we will try to add basic support of functional test locally 13:30:50 functional test is mostly about ensuring the service behaves as expected when processing REST requests 13:31:19 yep, so we don't touch other services in this part 13:31:57 only senlin services will be tested through REST API 13:32:33 so, basically, you will do 'source senlinrc; tox -e functional' 13:33:30 I hope by the end of l-3, we will have major workflows covered 13:34:17 this is also a big task 13:34:29 absolutely 13:34:45 sorry, just dropped 13:34:51 once the first workflow is running 13:35:04 adding other workflows would be simple 13:35:12 yes 13:35:22 just send a REST request and assert the response 13:35:25 hope can finish the first step in a week 13:35:48 cool! 13:35:59 scenario and integration test will be future work :) 13:36:31 it seems other projects are doing function tests in Tempest 13:36:42 integration test means we will be testing senlin's interaction with other services, using real Profile 13:36:55 we will also move it to tempest maybe one day? 13:36:58 hmm, most of them have moved functional test out of tempest I think 13:37:19 saw related patches and bps in different projects 13:37:35 Qiming, yes 13:37:41 yanyan, oh, my information is too old 13:37:45 yes, doing functional tests there would make the gating job run very long 13:37:55 I also got this info from Qiming :) 13:38:27 in heat, we used to call them heat-slow tests 13:38:33 and heat-very-slow tests, :) 13:38:44 I used to write some function tests in tempest for nova 13:38:49 :p 13:39:04 anyway, functional tests in each project, that makes sense, we will do it 13:39:15 yanyan, shall we move on? 13:39:16 right 13:39:17 cool, maybe you can share some experience about it 13:39:19 ok 13:39:21 ok 13:39:25 lets move on 13:39:30 #topic clusters of containers 13:39:52 jruano, anything new to share? 13:40:38 i am a little behind here... i spend some time digging through magnum and it has been more difficult than i expected 13:40:53 magnum requires a coe to manage containers 13:41:01 okay, that's true 13:41:29 so for the poc i am not sure we need all of that 13:41:45 there are things we cannot control out there 13:42:58 so for the short term i am focused on trying to get a cluster of containers deployed under coreos 13:43:08 maybe senlin is not capable of assuming the role of a coe in near future 13:43:39 i'd hope it can 13:44:02 magnum is wanting to be generic, but it is semantically very close to k8s 13:44:03 it seems to me that Magnum is still brainstorming their horizontal scaling support 13:44:10 #link http://lists.openstack.org/pipermail/openstack-dev/2015-August/071216.html 13:44:54 to be honest, we have been asking the question again and again, what are the values magnum bring to users 13:45:10 lol 13:45:16 i am struggling with it 13:45:37 fundamentally to me it appears to be a k8s abstraction 13:45:49 maybe the magnum team has some ideas I'm not aware of 13:46:00 do we need a k8s abstraction? 13:46:09 when i talk to folks in tosca, they want a generic container service 13:46:34 the 'container engine' could be docker or k8s, right? 13:46:35 as far as I know, magnum is just a tool to install k8s 13:46:56 yes, that is my understanding as well, haiwei__ 13:47:01 yes haiwei_ that is what it looks like from the code 13:47:07 and it is using Heat to do the VM provisioning 13:47:11 yep 13:47:56 a unified abstraction above docker and k8s? that is challenging 13:48:01 so what is our goal to work with magnum? 13:48:11 we want magnum to use senlin? 13:48:14 not just from technical perspective 13:48:42 we believe Senlin can do a better job managing the VMs they use to run containers 13:49:34 anyway, let's keep investigating 13:49:52 so that should be magnum->k8s->docker->senlin->vm->containers? 13:50:15 that is where i am at. i will have more over the next couple of days 13:50:34 the vision is: magnum talks to senlin to provision VMs, it then talks to k8s or docker to privision containers 13:51:27 maybe senlin should help manage the containers directly, :P 13:51:43 thats where i was headed 13:51:55 oh, really? 13:51:58 but magnum for now seems to want to just push that to the underlying coe 13:52:20 magnum doesn't want to reinvent things 13:52:26 true 13:52:38 k8s has a big community, docker has another 13:52:51 you are not compete head to head with them 13:53:27 so, for Senlin, we just see where we can help 13:53:43 okay, we don't have much time now 13:53:48 #topic open discussions 13:54:32 anything we have missed? 13:54:47 nope from me 13:54:59 guess no 13:55:02 nothing from me 13:55:10 see you guys, thanks for joining 13:55:12 it's ok for me 13:55:14 #endmeeting