13:00:32 #startmeeting senlin 13:00:33 Meeting started Tue Jun 6 13:00:32 2017 UTC and is due to finish in 60 minutes. The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:37 The meeting name has been set to 'senlin' 13:05:41 seems no one is in? 13:06:15 hi QiMing 13:06:27 hi, XueFeng 13:06:36 hi 13:06:41 Meeting will start? 13:06:51 already started I think 13:06:55 :) 13:07:18 I'm terribly sorry for accidentally skipped two meetings during the past weeks 13:08:04 No matter:) 13:08:13 both were due to my daughter 13:08:31 she was sick last week, I was in hospital ... 13:08:53 and we have had quite a few holidays since boston summit 13:09:08 So we need your lead 13:09:13 let's get started, anyway 13:09:25 OK 13:09:40 it doesn't have to be me, when it is meeting time, anyone can start the meeting by typing '#startmeeting senlin' here 13:09:56 #topic pike work items 13:10:06 #link https://etherpad.openstack.org/p/senlin-pike-workitems 13:10:17 duno if elynn is online 13:10:29 I'm here 13:10:37 I saw he was pushing a new patchset for the vdu profile 13:10:42 evening, elynn 13:11:07 Oh, I almost thought you can't see my message... 13:11:11 evening 13:11:29 Yes, I push a init patch 13:11:33 https://review.openstack.org/#/c/467108/ 13:11:50 everyone pls help review when possible 13:11:57 Just porting some features from vdu to server profile. 13:12:08 And fixed some unittests. 13:12:31 Still need more unittests and integrated tests. 13:12:41 ah, right, it is a porting effort 13:12:46 okay 13:12:49 I'm still working on it, just a little slow... 13:13:19 pretty big a change, will jump onto it this week 13:13:50 thanks for pushing this forward 13:14:04 next item: check action 13:14:15 Hoping this patch don't breaking any existing functions, please review it thoroughly 13:14:28 no activity for a few weeks now: https://review.openstack.org/#/c/435822/ 13:14:56 elynn, really hope so, but review itself cannot reveal all problems, I promise you, :) 13:15:14 hi~ 13:15:15 XueFeng, still on that patch? 13:15:33 or you want a discussion on senlin channel later? 13:15:35 hi, shark 13:15:35 hi,Qiming 13:15:53 Yes ,need discuss 13:16:07 in senlin channel later 13:16:19 alright, moving on 13:16:24 node adopt 13:16:34 almost finished the first round 13:16:44 need to add api test and functional test 13:16:49 great 13:17:03 not sure if we can add integration test for this 13:17:03 Will test it:) 13:17:20 there are quite a few holes in the current poc 13:17:28 and shark is using it 13:17:47 yes 13:17:51 for most of them I have left some comments in the source code so that we won't forget them 13:18:11 so you mean it is already usable? 13:18:16 my goshhhhh 13:18:25 I need add some function 13:18:38 feel free to propose patches, shark 13:18:47 we (usually) don't bite 13:19:18 next item is about scaling improvement, Ruijie is onto it now 13:19:32 yes, effers are still needed 13:19:35 for deletion process 13:19:36 and Ruijie is also optimizing the DB interactions 13:19:46 ah, Hero is online 13:19:59 :) 13:19:59 okay 13:20:16 let's try get it fully landed in Pike 13:20:30 okay 13:21:06 I want to compete the bp existed-pool-support-in-lb-policy 13:21:07 Revise runtime_data implementation, I think Ruijie and I both did something in that space 13:21:22 no problem, shark 13:22:03 if you want that work thread tracked, please add it to the etherpad 13:22:33 backport to Ocata (the runtime_data optimization) seems pretty difficult 13:22:58 ok 13:23:01 next item: RDO packaging 13:23:15 is it working now, XueFeng ? 13:23:29 sorry no process these days 13:23:43 will continue to do it 13:23:51 okay 13:23:57 health management 13:24:23 I believe the master branch still needs some work regarding mistral workflow support 13:25:36 'run_workflow()' is now a method of Node, but it is not invoked anywhere 13:25:48 I'll fix it 13:26:07 ok 13:26:27 no progress on rally test or senlinclient functional test as far as I can tell 13:27:08 that's all for things on etherpad, anything to add, guys? 13:27:18 rally test problem seems solved in rally project 13:27:35 oh really? 13:27:58 Will check and update etherpad 13:28:03 then the work item should be revised to 'add more rally test cases'? 13:28:10 sounds great, thanks 13:28:22 yes 13:29:42 okay, moving on 13:29:56 #topic boston summit recap 13:30:11 anything new to share with team after the boston summit? 13:30:43 we had three sessions there, presentation was well prepared 13:31:13 however, the attendance was not exciting 13:31:31 QiMing, I want to know the session by you and xinhui 13:31:53 one of the reasons could be that we could have stress more on problems solved than the projects/services backing that solution 13:32:33 most of the materials were prepared by xinhui 13:32:55 I was there just to repeat "what senlin is about" ... 13:33:50 I have watched that video 13:34:07 a million thanks to xinhui and elynn 13:34:47 yes 13:34:50 it was almost 1 month ago, now it is time for a new milestone pike-2 13:35:00 we about cutting pike-2 this week 13:35:23 please help scan the review backlog and approve things you feel good about 13:35:38 ok , we wil 13:35:44 thanks. 13:36:03 next thing we need help is about blueprints review 13:36:13 #topic blueprints to review 13:36:23 we have got quite a few proposed recently 13:36:48 some can be merged, some needs more discussion, others are already good to go 13:37:08 for example, https://blueprints.launchpad.net/senlin/+spec/support-node-protect-operation 13:37:23 this one needs some clarification then it is good to go 13:37:47 this one: https://blueprints.launchpad.net/senlin/+spec/improve-vm-name 13:37:53 ok.seems good 13:37:58 is about naming VMs 13:38:04 Qiming, for this process, we are tring to use policy data first 13:38:08 we need a design for that support 13:38:15 yes, vm name need improve 13:38:32 Ruijie_, which one? 13:38:34 maybe we can change to use request.body first, so that the process is much easier to be controlled? 13:38:37 node proection? 13:38:41 about the bp 13:39:10 it says that we want to protect the node not to be destroyed by scaling actions 13:39:38 well, implementation details can be discussed during code review 13:40:00 what I'm curious about is the scope of such a 'locked' tag 13:40:28 is it only designed to protect node from being deleted, or we can use it for guarding node update as well? 13:40:37 I'd suggest we use request parameter to controll it, will leave comments ~ 13:41:22 maybe a 'locked' or 'protected' node should be tagged so in database, 13:41:27 Maybe adding a new property named 'protected'? 13:41:46 yes, that is about the implementation 13:42:12 we can use new property, we can add tag support, we can use metadata, whatever works for this 13:42:32 but the first thing I want to clarify is the definition of 'protected' 13:42:50 okay 13:43:00 then we can together figure out how big a scope is such a "lock" 13:43:14 and its impact on cluster/node actions and policies 13:43:26 yes 13:43:51 I have talked to Hongbin on this, but he said he is busy on his thesis work ... 13:44:15 my suggestion was that he draft a spec for team to review, and comment 13:44:30 the same goes with VM naming 13:44:37 current default naming is terrible 13:44:53 we need a better solution 13:45:25 there are other bps for review as well: https://blueprints.launchpad.net/senlin 13:45:28 I have some time to work 13:45:35 pls feel free to take a look on them 13:46:00 cool, chenyb4_, maybe you can grab a few BPs to work on, :) 13:46:21 post your questions to senlin channel when you need help 13:46:31 ok 13:46:58 alright, that's all on my mind for the meeting today 13:47:07 #topic open discussions 13:47:16 please speak up 13:47:23 we still have some time, :D 13:47:27 I want to discuss health policy how to handle the adopt node 13:48:08 how about just treat them the same as other nodes? 13:49:04 yes, we can think them the same as other node. 13:49:06 we rebuild user vm,is this good 13:49:07 nodes 13:49:33 I don't understand profile support snapshot and restore, the profile is about single vm configure, so the snapshot how can i create? 13:50:32 snapshot/restore is an operation supported by backend service, modeled into a profile operation in senlin 13:50:49 just compare it to 'update' 13:50:57 how about remove the node from senlin? 13:51:15 shark, you can do that 13:51:28 just as if you have nevered created such a node 13:52:16 wait ... did you mean delete a senlin node without deleting the physical resource (which could be a VM)? 13:52:27 yes 13:52:41 why are you doing that? 13:53:18 you can leave it as an orphan node (not member of any cluster) in senlin 13:53:52 I will not call it 'delete', it sounds more like an 'abandon' operation 13:54:07 yes 13:54:23 I'm interested in the use case -- why do you need it 13:54:23 becauser it is a adopt node 13:55:10 so ... we have to define "abandon" because we have defined "adopt" ? 13:55:45 then the design looks beautiful? :D 13:56:18 sounds feasible 13:56:46 it is doable, although ... I'm not sure it would be useful 13:56:54 I'm not against it, to be honest 13:57:13 ok 13:57:41 anything else? 13:57:46 3 mins left 13:57:46 Qiming, I have a last question today 13:57:48 Since we are going to have 'protected' node 13:58:09 and then force delete .. 13:58:14 maybe we should also need to consider health policy on them. 13:58:18 Intergation enterprise monitoring product, senlin and mistral for auto-healing.Can this session fall to the ground? I would like to refer to this architecture deployment 13:58:24 elynn, sure 13:58:41 that is why I think we need a spec for more discussions on the implications 13:58:57 XueFeng, it is 13:59:03 it is real 13:59:24 buy xinhui a rose, she will tell you the full story 13:59:44 time's up, guys, thanks for joining, gooooooood night 13:59:45 haha 13:59:48 #endmeeting