13:01:11 <Qiming> #startmeeting senlin
13:01:12 <openstack> Meeting started Tue Sep  6 13:01:11 2016 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:15 <openstack> The meeting name has been set to 'senlin'
13:01:48 <yanyanhu> hello
13:01:52 <zzxwill> Hi.
13:01:55 <Qiming> hi
13:02:11 <Qiming> just started meeting in senlin channel, the meeting bot is working there too, ;)
13:02:25 <yanyanhu> :)
13:02:57 <zzxwill> I felt strange just now as I saw a meeting room in senile channel.
13:03:00 <elynn> hi
13:03:08 <Qiming> yup, a mistake
13:03:14 <Qiming> let's get started
13:03:21 <zzxwill> okay
13:03:22 <Qiming> #topic newton work items
13:03:30 <Qiming> #link https://etherpad.openstack.org/p/senlin-newton-workitems
13:03:58 <yanyanhu> rally support
13:04:02 <Qiming> testing side, the profile context one is finally in
13:04:06 <yanyanhu> profile context patch has been merged
13:04:07 <yanyanhu> yep
13:04:18 <yanyanhu> will work on cluster context in next step
13:04:37 <yanyanhu> and have proposed patch to revise existing test job in senlin repo
13:04:49 <yanyanhu> to use profile context support
13:04:55 <Qiming> okay, just keep working on it
13:05:02 <yanyanhu> https://review.openstack.org/365878
13:05:05 <yanyanhu> yep
13:05:40 <Qiming> can remove the working item related with cmcc?
13:05:48 <yanyanhu> hmm, feel so
13:05:49 <Qiming> don't think they are investing on that
13:06:00 <yanyanhu> I think they are still using old version rally to perform the test
13:06:12 <Qiming> sigh
13:06:13 <yanyanhu> since their environment is also not latest one
13:06:26 <yanyanhu> so we'd better help them mirgate to latest one
13:06:32 <yanyanhu> will talk to eldon about it
13:06:40 <Qiming> okay, thx
13:06:46 <yanyanhu> no problem
13:06:54 <Qiming> don't think xinhui is online, I don't have update on health management
13:07:06 <Qiming> no update on documentation either
13:07:25 <Qiming> the tutorial doc has pretty much what we have got today
13:07:33 <Qiming> that item on line 21 can be removed
13:07:40 <Qiming> just need to polish the wiki
13:07:42 <yanyanhu> ok
13:07:56 <yanyanhu> versioning support on policy/profile, it will be a work session in Barcelona
13:08:02 <Qiming> yes
13:08:13 <Qiming> before that, we can do some experiments now
13:08:20 <yanyanhu> yes
13:08:27 <Qiming> base everything that should be versioned on oslo.versionedobjects
13:08:43 <yanyanhu> ok
13:09:06 <yanyanhu> hi, ruijie, welcome
13:09:16 <Qiming> we can remove the whole item and add a version control item into the FEATURES.rst?
13:09:17 <ruijie> hi´╝îyanyan,evening
13:09:28 <Qiming> hi, ruijie!
13:09:30 <yanyanhu> Qiming, sure
13:09:41 <yanyanhu> will add it
13:09:50 <ruijie> hi,qiming
13:10:05 <Qiming> container support is pretty much there already
13:10:26 <Qiming> a lot of features to be added though, but as a POC, it is working now
13:10:37 <Qiming> finally haiwei has got cycles to work on it
13:10:46 <yanyanhu> great
13:11:05 <yanyanhu> we have a work session on this as well
13:11:10 <Qiming> yep
13:11:23 <Qiming> maybe (some of) you have noticed
13:11:36 <Qiming> we have added a 'dependents' column to the node table
13:11:54 <Qiming> that property will be deserialized and exposed to users
13:12:28 <Qiming> we will use that column to record that a node is being depended by another node (maybe of a different profile)
13:12:50 <Qiming> the current usage scenario is that a VM node may be hosting one or more containers (nodes)
13:13:07 <Qiming> without getting those containers deleted, the VM node cannot be deleted
13:13:26 <Qiming> this will be further extended into other usage scenarios in future
13:13:37 <yanyanhu> ok
13:13:49 <yanyanhu> I believe that field will be useful in more scenarios
13:13:56 <Qiming> for example, a DB node may be treated as depended by a web node, thus it cannot be removed
13:14:02 <yanyanhu> where dependency relationship exists
13:14:11 <yanyanhu> yep
13:14:40 <Qiming> just want to leave a note here that the 'dependency' we are talking about is pretty generic
13:15:08 <Qiming> I was trying the ceilometer driver using opentacksdk recently
13:15:25 <Qiming> hoping that we can leverage that metering service for container scheduling
13:15:57 <Qiming> we don't have to monitor loads on all VM nodes because ceilometer (maybe also monasca?) has already done so
13:16:00 <zzxwill> Use ceilometer to meter the burden of each container?
13:16:31 <Qiming> we just need to check the statistics from ceilometer, and select the node with the lightest node to run a new container
13:16:49 <Qiming> not sure if container monitoring is supported by ceilometer
13:16:58 <zzxwill> Got it. Thanks.
13:16:59 <yanyanhu> zzxwill, the burden of vm which will be used as the host for container cluster :)
13:17:37 <Qiming> still need some time to understand the pros/cons of ceilometer doing load monitoring
13:18:16 <Qiming> our best hope is that it can do a good job in this, though I know some drawbacks not yet fixed so far
13:18:29 <yanyanhu> one potential issue is the metric data could be inaccurate
13:18:51 <Qiming> even if users are not using ceilometer, they are using something else for monitoring, that 'something' should expose an API for query
13:19:07 <Qiming> so we don't bother to do the monitoring outselves
13:19:41 <Qiming> yup, cpu overcommit is not yet handled well as far as I know, memory consumption may not be accurate
13:20:34 <Qiming> network bandwidth consumption is another complex area, it is really up to how the network is set up (i.e. how many nics, which one is for business data ...)
13:21:07 <Qiming> anyway, we can assume that the metering project can solve those problems
13:21:42 <yanyanhu> sure, that's why we want to rely on them :)
13:21:45 <Qiming> next thing is receiver support
13:22:03 <yanyanhu> working on trust building between enduser and zaqar service user
13:22:05 <Qiming> yes, a soft dependency
13:22:21 <yanyanhu> basic workflow is ok I think. Just some authentication issues need to figure out
13:22:38 <yanyanhu> e.g. the roles to specify when building trust
13:22:40 <Qiming> yep, pls think of a more generic multi-tenancy support scenario
13:22:48 <yanyanhu> yes
13:22:54 <yanyanhu> and need more tests here I think
13:22:56 <Qiming> right, also the authentication thing
13:23:10 <yanyanhu> hopefully, basic support can be done before newtone release
13:23:29 <Qiming> receivers are not supposed to be reused across projects, right?
13:23:41 <yanyanhu> Qiming, sure
13:24:14 <yanyanhu> ideally, each user should create their own message type of receiver(s)
13:24:18 <yanyanhu> for different purpose
13:24:19 <Qiming> okay, that is one of the rules we are not supposed to break
13:24:24 <yanyanhu> yep
13:24:53 <Qiming> also security perspective
13:25:10 <yanyanhu> right, that part need more thorough thinking
13:25:17 <Qiming> how much do I have to know if I want to trigger an operation on a cluster/node I don't own?
13:25:25 <yanyanhu> to avoid potential security risk
13:25:35 <Qiming> yup
13:26:00 <yanyanhu> Qiming, that is problem. Actually haven't got a clear idea about it yet
13:26:07 <Qiming> sometimes a cluster is owned by Alice, but she trusts Bob to trigger some operations on that cluster
13:26:15 <yanyanhu> need more thinking here after basic workflow is ok
13:26:19 <Qiming> okay
13:26:46 <yanyanhu> good news is zaqar trust support works well till now :)
13:26:50 <yanyanhu> as expected
13:26:56 <Qiming> the initial step could be very restrictive, we can gradually relax the constraints when we feel comfortable
13:27:08 <yanyanhu> so we can rely on it to build our solution
13:27:19 <yanyanhu> Qiming, yes, that makes sense
13:27:29 <yanyanhu> starts from most strict one
13:27:30 <Qiming> if we work in the other direction, we will receive a lot of complaints in near future
13:27:45 <yanyanhu> yep
13:28:22 <Qiming> events/notifications, we don't have them yet
13:28:41 <Qiming> maybe cannot get them implemented this cycle due to short of hands
13:28:58 <yanyanhu> understand
13:29:09 <Qiming> #topic planning for RC releases
13:29:33 <Qiming> #link https://releases.openstack.org/newton/schedule.html
13:29:58 <Qiming> Next week would be the RC-1 release
13:30:21 <Qiming> RC2 is 2016-09-26, 20 days away
13:30:43 <Qiming> I'm planning to fix two things before RC2
13:31:05 <Qiming> 1. exception handling, regarding the interaction with nova/heat drivers
13:31:26 <Qiming> 2. expected/desired capacity, when there are failures
13:31:45 <Qiming> now I'm working on the former one
13:31:46 <yanyanhu> ok
13:31:55 <yanyanhu> these two things are not related to features
13:32:05 <Qiming> yes, if all these can be completed
13:32:19 <Qiming> I'll resume the work on versioning support
13:32:27 <yanyanhu> cool
13:33:01 <Qiming> using ovo to control request/response/notification etc
13:33:31 <zzxwill> What's ovo?
13:33:42 <yanyanhu> we may also need a plan for features we want to include to newtone release
13:33:45 <Qiming> also need to sync up with eldon on summit preparation
13:33:54 <Qiming> oslo.versionedobjects, zzxwill
13:33:59 <yanyanhu> since I feel we may not be able to finish all items we planned before for newton cycle
13:34:04 <zzxwill> Thanks Qiming.
13:34:04 <yanyanhu> in last summit
13:34:34 <Qiming> what's in your mind yanyan, regarding features?
13:35:04 <Qiming> we are almost there
13:35:08 <yanyanhu> hmm, basic HA support, message receiver, basic versioning support, tests(almost done), doc(almost done)
13:35:33 <yanyanhu> container cluster(partially done)
13:35:45 <Qiming> versioning is a bigger problem than we thought, --- because we got too ambitious
13:35:47 <yanyanhu> Qiming, yes, almost there
13:35:57 <Qiming> container cluster is a POC now
13:36:06 <yanyanhu> Qiming, seems so :)
13:36:13 <elynn> basic HA support means service HA? yanyanhu
13:36:15 <yanyanhu> version everything
13:36:30 <Qiming> not perfect, need more work on network/storage
13:36:34 <yanyanhu> hi, elynn, VM HA
13:37:08 <Qiming> maybe also a very restrictive message receiver?
13:37:16 <yanyanhu> service HA is another topic actually, but we have provides some basic supports for it I think :)
13:37:26 <yanyanhu> Qiming, yes
13:37:32 <yanyanhu> hopefully, it will be there
13:37:41 <Qiming> the whole health monitoring loop is completed and working
13:37:57 <Qiming> though I believe there are things to improve, as always
13:38:14 <Qiming> that is the reason we put it as a design summit topic
13:38:15 <yanyanhu> sure. Further improvement can be done step by step
13:38:22 <yanyanhu> yea
13:38:24 <Qiming> btw, we proposed 4 slots
13:38:49 <yanyanhu> no much, but I think it's sufficient
13:38:54 <Qiming> all working sessions
13:39:14 <yanyanhu> hope we all can be there to have some f2f discussion :)
13:39:15 <Qiming> I'd like to join more sessions from other projects, interact with them
13:39:24 <yanyanhu> sure
13:39:29 <yanyanhu> like, zaqar, taker
13:39:39 <yanyanhu> also HA I think
13:39:43 <yanyanhu> etc.
13:40:12 <Qiming> magnum, zun, etc
13:40:35 <yanyanhu> sure
13:40:55 <Qiming> #topic open discussions
13:40:59 <yanyanhu> looking forward to listen to hongbin and eli's topic on zun :)
13:41:29 <yanyanhu> oh, BTW, I have removed integration test work item from etherpad
13:41:40 <yanyanhu> but actually it is broken again after latest patch is merged...
13:41:57 <yanyanhu> the reason is zaqarclient was not setup correctly...
13:41:58 <Qiming> but the integration fix is merged this afternoon, right?
13:42:16 <yanyanhu> but I doubted that problem should be fixed in zaqar not devstack-gate
13:42:26 <yanyanhu> Qiming, yes, mysql service is there now
13:42:34 <yanyanhu> but zaqarclient installing failed
13:42:37 <Qiming> https://review.openstack.org/#/c/365875/
13:42:41 <Qiming> this one ^
13:43:04 <Qiming> we have to install zaqarclient?
13:43:18 <Qiming> aren't we interacting with zaqar thru sdk?
13:43:21 <yanyanhu> yep. very appreciated for Andreas and yolanda's quick approval
13:43:38 <yanyanhu> Qiming, yes, but client setting up is part of zaqar devstack setting up
13:43:42 <Qiming> broken gate is always a high priority
13:43:49 <Qiming> ...
13:44:05 <yanyanhu> since we enabled zaqar as plugin, zaqar client will be setup automatically during zaqar installation
13:44:17 <yanyanhu> it failed during this progress
13:44:20 <Qiming> that is not something to be fixed at gate side
13:44:22 <Qiming> right?
13:44:26 <yanyanhu> yes, feel so
13:44:40 <yanyanhu> there could be some tricky things here
13:44:43 <yanyanhu> will dig it
13:44:46 <Qiming> I saw zaqar has functional tests
13:44:58 <Qiming> they may have some tricks on this
13:45:05 <yanyanhu> Qiming, yes. and I also noticed zaqar rally test job failed for a while
13:45:25 <yanyanhu> so not sure whether that is caused by the same problem
13:45:28 <Qiming> okay, pitty
13:45:39 <yanyanhu> anyway, will investigate it
13:45:54 <Qiming> okay
13:46:00 <yanyanhu> after message receiver support is done, will add integration test for it
13:46:35 <Qiming> okay
13:46:43 <Qiming> anything else?
13:46:55 <yanyanhu> nope from me
13:47:03 <Qiming> elynn, zzxwill, ruijie_ ?
13:47:05 <elynn> no from me
13:47:07 <zzxwill> me neither.
13:47:10 <ruijie_> Actually, I got one
13:47:24 <ruijie_> I jsut registered a BP for supporting cluster-replace-nodes
13:47:29 <ruijie_> from the TODO.rst
13:47:52 <yanyanhu> great, thanks for working on it :)
13:48:06 <Qiming> cool
13:48:30 <ruijie_> So, I'm going work on it and try to figure it out asap
13:48:40 <yanyanhu> any question about it, just feel free to ping me or mail me :)
13:48:52 <Qiming> don't hesitate if you got questions ...
13:48:55 <ruijie_> Sure, I will :)
13:49:18 <Qiming> or, better yet, stay on irc channel
13:49:22 <Qiming> ask your questions
13:50:10 <yanyanhu> yep, using irc is better :)
13:50:18 <Qiming> mail is fine, but sometimes it is personal or private communication, not the community way, ;)
13:50:35 <ruijie_> yea, I will attend IRC whenever I am free :)
13:50:44 <Qiming> okay
13:50:49 <Qiming> thank you all for joining
13:50:51 <yanyanhu> ruijie_, or you can leave msg if we are not online
13:51:07 <Qiming> take your 10 minutes back into your sweet dreams
13:51:12 <Qiming> #endmeeting