13:02:25 <Qiming> #startmeeting senlin
13:02:26 <openstack> Meeting started Tue Mar  1 13:02:25 2016 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:27 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:02:29 <openstack> The meeting name has been set to 'senlin'
13:02:44 <Qiming> sorry guys, a little bit late again
13:02:51 <yanyanhu> hi
13:02:54 <zzxwill> Hello. It's fine.
13:03:08 <haiwei> hi
13:03:11 <Qiming> welcome, zzxwill
13:03:21 <Qiming> #link https://wiki.openstack.org/wiki/Meetings/SenlinAgenda
13:03:35 <Qiming> please review agenda and see if you have things to add
13:03:51 <Qiming> #topic Austin summit planning
13:03:58 <yanyanhu> hi, Qiming, the etherpad link for summit is not added in agenda
13:04:13 <Qiming> okay, pls help
13:04:26 <yanyanhu> just can remeber the name
13:04:27 <yanyanhu> ...
13:04:39 <lixinhui> hi
13:04:48 <Qiming> newton-senlin-sessions
13:04:48 <yanyanhu> tried several different ones, all failed...
13:04:58 <Qiming> #link https://etherpad.openstack.org/p/newton-senlin-sessions
13:04:58 <yanyanhu> ah, missed the last 's'
13:05:15 <Qiming> following previous convention on naming the etherpad
13:05:53 <Qiming> as the summit is drawing near, we need to start brainstorm topics for discussion
13:06:23 <Qiming> the above link is an etherpad for proposals
13:07:12 <Qiming> we have a lot of things to discuss, especially those items from TODO.rst and FEATURES.rst
13:07:24 <yanyanhu> yes
13:07:59 <Qiming> for example, the node-create/delete actions handling could be a discussion topic
13:08:12 <Qiming> I'm not sure we will figure that out by mitaka release
13:08:13 <haiwei> so this time would be how long for senlin design summit?
13:08:49 <Qiming> Fishbow slots (Wed-Thu)
13:08:59 <Qiming> Workroom slots (Tue-Thu)
13:09:08 <Qiming> Contributors meetup (Fri)
13:09:28 <yanyanhu> just as usual
13:09:34 <Qiming> it is up to the team (you and me) to figure out how many sessions do we need
13:10:09 <Qiming> if possible, I'm getting back to ttx after we have a rough estimation
13:10:10 <haiwei> yes
13:10:40 <Qiming> in the past, for example, some projects have 2+12+2
13:10:57 <Qiming> some projects have 1+0+0
13:11:17 <Qiming> Some projects have 12+0+2
13:11:36 <Qiming> just focus on the topics you want to discuss
13:12:37 <yanyanhu> just added two topics
13:12:38 <Qiming> #action Everyone please input your proposals by end of Thursday (Feb. 3)
13:12:41 <yanyanhu> will try to add more
13:12:50 <Qiming> does that action item makes sense to everyone?
13:13:01 <yanyanhu> I believe there will be many items we want to talk f2f :)
13:13:07 <yanyanhu> sure
13:13:13 <yanyanhu> totally makes sense to me
13:13:19 <Qiming> On Friday, we need to do a consolidation
13:13:29 <haiwei> is it necessary to send a mail?
13:13:32 <Qiming> just a side note
13:13:34 <yanyanhu> ok, so we have 2 days for input
13:14:00 <Qiming> once we have a schedule, we can send an email to the dev list
13:14:44 <Qiming> there will be someone collecting inputs from all projects and work out a summary like this: https://wiki.openstack.org/wiki/Design_Summit/Mitaka/Etherpads
13:15:36 <Qiming> please mark your session's type: fishbowl or working
13:15:54 <yanyanhu> ok
13:16:13 <Qiming> let's move on
13:16:20 <Qiming> #topic Mitaka release schedule
13:16:25 <Qiming> #link http://releases.openstack.org/mitaka/schedule.html
13:16:43 <Qiming> this week is the m-3 milestone
13:16:47 <Qiming> ... time flies
13:17:22 <Qiming> it is a week for many things: feature freeze, requirements freeze, soft string freeze, final release of client libs
13:18:28 <zzxwill> As for soft StringFreeze, does it affect our project?
13:18:46 <Qiming> we'll create a b3 tag this week
13:18:46 <Qiming> that one is just a few commands
13:18:46 <Qiming> but there are other things where helps are needed
13:18:46 <Qiming> yes, zzxwill
13:18:48 <Qiming> just talked to Daisy today
13:19:09 <yanyanhu> one thing that is still unclear is when should we rework our driver for sdk release 0.8.0
13:19:21 <Qiming> if we have a focal for i18n, they can help try put senlin translation into their agenda
13:19:45 <Qiming> I'm seeing the translation to Japanese for senlin-dashboard is bumping back
13:20:07 <Qiming> that one is something I should have taken care of
13:20:14 <yanyanhu> it will really cost some time to finish this job
13:20:22 <Qiming> I have pushed sdk team to cut a version
13:20:24 <zzxwill> I am a translator for i18n, I wonder whether I can overtake the role. But I still have so many things to learn.
13:20:31 <Qiming> however, 0.8.0 is not ideal
13:20:34 <yanyanhu> Qiming, understand
13:20:41 <Qiming> it still misses some patches
13:20:50 <yanyanhu> oh, right, the path_arg patch is still hang there
13:20:55 <Qiming> not sure if we should push them to release a 0.8.1
13:21:20 <briancurtin> Qiming: if you have some that you want done i can easily do a 0.8.1 (path_arg is a part of a big refactoring, “path args” are going away)
13:21:22 <Qiming> another thing is the version bump at global requirements got blocked again and again, for no good reason
13:21:51 <Qiming> hello, briancurtin, that would be REALLY REALLY helpful
13:22:22 <Qiming> there is a workaround for that, briancurtin
13:22:27 <yanyanhu> hi, briancurtin, I noticed Terry abandoned his patches about path_args, does that mean this problem has been resolved?
13:22:51 <briancurtin> Qiming: i don’t want to interrupt your meeting (my irc client just highlights on SDK) so if you want to send me some of the reviews you would need for a 0.8.1
13:22:53 <Qiming> we only need a patch for neutron lb member I think
13:23:03 <Qiming> yanyanhu, pls help work we brian and see if it makes sense
13:23:07 <briancurtin> yanyanhu: no it’s very much still a problem, but i am refactoring the Resource class to eliminate it. the old ways were not enough
13:23:10 <Qiming> s/we/with
13:23:34 <yanyanhu> understand
13:23:42 <Qiming> the good thing is SDK is really great, it helps us eliminate all dependencies to xxxclients
13:23:50 <briancurtin> yanyanhu: it’s a pretty big change to the insides of the SDK, but the proxy interface so far is the same so the changes user experience should not be that much
13:24:35 <Qiming> final release for senlinclient
13:24:44 <Qiming> that is very possible now
13:24:50 <yanyanhu> I see. thank you so much, briancurtin.
13:25:00 <Qiming> thanks to di xiaoli, we have all OSC plugin patches landed
13:25:11 <Qiming> now you can do a 'openstack cluster list' for example
13:25:55 <Qiming> as for string freeze, need to do an examination if there are things need fixes
13:26:20 <Qiming> starting from m-3 release, we should be really careful about introducing new strings
13:26:43 <yanyanhu> cool
13:26:49 <Qiming> moving on
13:26:59 <Qiming> #topic Mitaka work items review
13:27:08 <Qiming> #link https://etherpad.openstack.org/p/senlin-mitaka-workitems
13:27:30 <Qiming> heat resource type support, it is done
13:27:38 <Qiming> thanks for the great effort fro ethan
13:27:52 <yanyanhu> so all resources are available now?
13:28:24 <Qiming> yes, all resources we planned to expose via Heat resource types
13:28:33 <yanyanhu> great
13:28:43 <Qiming> that blueprint has been marked 'completed'
13:28:47 <yanyanhu> may the next step is some example template
13:28:51 <yanyanhu> ok
13:28:56 <Qiming> testing side
13:28:57 <yanyanhu> s/may/maybe
13:29:19 <Qiming> yep, I think ethan has already started some work
13:29:41 <Qiming> that could be our focus in coming weeks
13:30:03 <Qiming> just curious if something new to share, yanyanhu ?
13:30:08 <yanyanhu> functional test for cluster update is in progress now. Have done some preparation by reworking test driver.
13:30:18 <yanyanhu> yes, made some progress
13:30:29 <yanyanhu> hope can propose the first patch tomorrow
13:30:50 <yanyanhu> for test of cluster update with flavor being changed
13:30:59 <Qiming> okay, we may need to start some stress testing next week
13:30:59 <Qiming> cool
13:31:23 <Qiming> next, health management
13:31:35 <yanyanhu> actually, I just added stree test topic into etherpad
13:31:48 <Qiming> xinhui has helped added db table and an initial version
13:31:53 <Qiming> still polishing it
13:31:59 <lixinhui> Yes, Qiming
13:32:00 <yanyanhu> I think we may need more discussion about it, from both developer and user point of view
13:32:12 <Qiming> hopefully, we can get it stable by RC1
13:32:13 <lixinhui> I will keep testing it and polish it
13:32:32 <Qiming> not sure we can finish the LB based health detection
13:32:39 <lixinhui> yanyanhu, agree to you
13:32:56 <yanyanhu> healthmonitor supported has added for lb policy
13:33:06 <lixinhui> Alarm has been created successsfully based on status change
13:33:16 <yanyanhu> patch is under review. It works in my env.
13:33:34 <lixinhui> yanyanhu, I will test it also
13:33:37 <yanyanhu> but still need more tests
13:33:39 <Qiming> omg, you all unbelievable
13:33:44 <yanyanhu> lixinhui, thanks :)
13:33:53 <lixinhui> just the bug is under resovling
13:34:19 <lixinhui> yanyanhu :) you work so quicly!
13:34:20 <Qiming> somehow, somewhere, we need to document this
13:34:43 <Qiming> and we should warn users about the limitations of this naive solution
13:34:51 <Qiming> it is a starting point
13:35:04 <yanyanhu> lixinhui, my pleasure :)
13:35:11 <lixinhui> :)
13:35:21 <yanyanhu> try my best to clean the obstacle before you guys
13:35:25 <lixinhui> okay, Qiming
13:35:26 <Qiming> more advanced health monitoring is out of senlin's scope, we do provide some entry level support etc ...
13:35:49 <Qiming> policy documentation, have to resume that thread next week
13:36:04 <Qiming> this week I'll focus on the jobs mandated by release schedule
13:36:17 <Qiming> so far have documented deletion policy and scaling policy
13:36:34 <Qiming> revising the affinity is almost done, then I'll document it as well
13:36:48 <Qiming> when lb policy is finalized, we document it later
13:36:54 <lixinhui> super start, yanyanhu
13:37:03 <lixinhui> s/start/star
13:37:29 <Qiming> profile
13:37:43 <lixinhui> Qiming is machine again
13:37:44 <lixinhui> ...
13:37:46 <Qiming> maybe should move disk update to TODO?
13:38:10 <yanyanhu> maybe we should move the item of nova disk updating support back to TODO list
13:38:13 <Qiming> call me Machinagain
13:38:19 <yanyanhu> yes, have the same feeling
13:38:22 <lixinhui> :)
13:38:31 <Qiming> yup, don't drop the ball at least
13:38:59 <Qiming> I'll keep the NODE_CREATE/DELETE item there, until I'm sure I cannot finish it
13:39:13 <yanyanhu> ok
13:39:16 <Qiming> lock breaker, that one has been finished?
13:39:45 <yanyanhu> guess so
13:39:46 <Qiming> yanyanhu, are you seeing the error logs again?
13:40:00 <yanyanhu> the db warning happened during engine starts has gone
13:40:03 <yanyanhu> nope
13:40:28 <yanyanhu> ahh, sorry, that's another problem
13:40:57 <yanyanhu> but I really didn't see error logs again
13:41:03 <Qiming> okay, killing that item
13:41:23 <Qiming> ok
13:41:28 <Qiming> added a job for you
13:41:37 <Qiming> you are welcome, yanyanhu
13:41:56 <Qiming> haiwei, still there?
13:41:56 <yanyanhu> thanks :)
13:42:04 <haiwei> yes
13:42:15 <Qiming> do we have any progress on container support?
13:43:00 <Qiming> if we need more time to come up with a prototype, we can add a design summit session for this
13:43:07 <Qiming> it can be a fishbowl session
13:43:17 <haiwei> not really, I was learning Magnum, and try to figure out a good solution for container cluster
13:43:25 <Qiming> okay
13:43:49 <Qiming> just please keep the team posted when you have some ideas
13:44:04 <haiwei> currently there is not a clear image of container cluster for me
13:44:05 <Qiming> oh, missed some items
13:44:17 <Qiming> haiwei, that is something interesting
13:44:25 <yanyanhu> haiwei, any possible the bay can be built based on senlin cluster?
13:44:29 <haiwei> ok, I will try to update the spec file
13:44:29 <yanyanhu> :)
13:45:07 <Qiming> my only big concern is that we still need to involve a global scheduler for placing containers
13:45:28 <Qiming> other than that, bringing up container cluster over senlin is a piece of cake
13:45:31 <haiwei> I think the problem we need to think about first is shall Senlin use Magnum or not for our container solution
13:45:53 <Qiming> haiwei, I would think from the other direction
13:46:08 <haiwei> I think it depends on what kind of container service senlin will provide
13:46:09 <Qiming> I think Magnum should use senlin to create clusters
13:46:38 <Qiming> this topic has been brought up several times before
13:46:41 <yanyanhu> haiwei, I think magnum itself doesn't provide container management interfaces?
13:47:00 <haiwei> I think it does, yanyanhu
13:47:13 <yanyanhu> ohh
13:47:21 <Qiming> the Magnum team was always saying no, because 'senlin is a stackforge project', 'senlin is not an official project' ... 'we don't want dependency to non-official services ...'
13:47:35 <haiwei> from magnum client, you can all create a container by 'magnum container-create'
13:47:40 <yanyanhu> so we can treat it as a container management service just as k8s or mesos?
13:47:45 <haiwei> but we need a container cluster
13:47:46 <Qiming> haiwei, why I'm doing that?
13:48:03 <Qiming> why don't we call docker api directly?
13:48:36 <Qiming> any benefits users will get?
13:48:40 <haiwei> that is a solution, but you need to handle the problems like scheduling, network and so on
13:48:49 <Qiming> suppose we are going down this path
13:49:06 <Qiming> does magnum do scheduling?
13:49:13 <haiwei> I am afraid senlin will have to do some jobs which has already done by other projects
13:49:31 <Qiming> yes, senlin's scope is always limited
13:49:33 <haiwei> no, Magnum uses docker swarm things
13:49:56 <Qiming> then why we need magnum in the middle?
13:50:27 <haiwei> as I said, we should think about the use case first, what kind of container cluster senlin will provide
13:50:37 <Qiming> I'm open to any use case
13:50:52 <Qiming> to any benefits we or the users can get by having senlin to invoke magnum
13:51:27 <Qiming> if I'm a user, I don't care there is a cluster at all
13:51:53 <Qiming> the thing I care is, I have a docker image, please help me deploy it to your cloud
13:51:55 <haiwei> yes, we should make user feel by using Senlin they can manage container cluster better, or else, they will use Magnum or other things
13:52:30 <Qiming> I want the service capable of handling autoscaling, load-balancing, high-avaialbility etc. all those non-functional requirements
13:52:51 <Qiming> haiwei, I don't think so
13:53:03 <Qiming> users really don't care the name of your project
13:53:32 <lixinhui> agree to Qiming on this point
13:54:07 <Qiming> it is not the glue layer that matters to users
13:54:26 <haiwei> yes, I am not focusing on using which project, but on what kind of service
13:54:32 <Qiming> it is the eventual user experience, service level you provide to users that matters
13:55:17 <Qiming> how about a service that can deploy and manage container clusters, while capable of handling autoscaling, load-balancing, high-avaialbility
13:55:37 <haiwei> that is great
13:55:58 <Qiming> how far are we from that goal?
13:56:12 <Qiming> what are the specific feature gaps we need to fill?
13:56:47 <haiwei> I can't give a clear answer now
13:56:47 <Qiming> for each feature missed, are there any existing OS-style service we can leverage so we can focus on the end goal?
13:57:35 <Qiming> that is something we need to think about
13:57:39 <yanyanhu> Qiming, I think the problem is which layer we want to be located
13:57:49 <Qiming> that is a good question
13:57:49 <lixinhui> yanyanhu
13:57:55 <yanyanhu> whether it's just a IaaS cluster management service
13:57:58 <yanyanhu> or more
13:58:09 <lixinhui> I think that does not matter
13:58:09 <Qiming> yanyanhu, to be honest, I don't care
13:58:22 <yanyanhu> yes, but that will decide the interface you expose
13:58:23 <Qiming> I don't think any users care at all
13:58:28 <lixinhui> if senlin can provide XXXXX service for container
13:58:34 <yanyanhu> e.g. app/service-oriented
13:58:45 <lixinhui> that will be good
13:58:54 <yanyanhu> or just bare container management
13:59:10 <yanyanhu> yes, but they are two different things
13:59:11 <Qiming> bare container management is the corner stone
13:59:24 <yanyanhu> Qiming, yes, I also think so
13:59:26 <Qiming> the upper layer interface is what users see
13:59:39 <Qiming> as a service, you do all the labor
13:59:44 <yanyanhu> right
13:59:56 <Qiming> that is why people call you a service
14:00:06 <yanyanhu> so we should be a service on PaaS layer
14:00:16 <Qiming> drawing a hard line between iaas and paas won't help solve the problem
14:00:23 <yanyanhu> if we want to provide that kind of support
14:00:25 <Qiming> sorry, time's up
14:00:34 <Qiming> back to senlin channel pls
14:00:39 <yanyanhu> ok
14:00:39 <Qiming> #endmeeting