13:00:57 <Qiming> #startmeeting senlin
13:00:58 <openstack> Meeting started Tue Jul 11 13:00:57 2017 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:59 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:01 <openstack> The meeting name has been set to 'senlin'
13:01:15 <Qiming> evening, guys
13:01:39 <XueFeng> hi,all
13:03:04 <elynn> hi
13:03:04 <Qiming> anyone online?
13:03:09 <Qiming> hi, elynn
13:03:16 <XueFeng> hi ,Qiming
13:03:40 <elynn> Parallel attending other meetings at the same time...
13:03:44 <xinhuili_> hi, everyone
13:03:48 <Qiming> Thanks you drafting the agenda, XueFeng
13:03:59 <Qiming> hi xinhui and ruijie
13:04:05 <Qiming> let's get started
13:04:13 <ruijie> :)
13:04:13 <Qiming> #topic agenda
13:04:26 <Qiming> #link https://wiki.openstack.org/wiki/Meetings/SenlinAgenda
13:04:27 <Guest53985> h
13:04:28 <Guest53985> hi
13:04:32 <XueFeng> Hi, Qiming ,my pleasure
13:04:42 <Qiming> pls add items if you have one
13:04:55 <Qiming> #topic pike work items
13:05:13 <Qiming> let's see if there are any updates on the work items etherpad
13:05:19 <Qiming> #link https://etherpad.openstack.org/p/senlin-pike-workitems
13:05:22 <XueFeng> hi Guest53985,
13:05:44 <Qiming> just noticed the new item about API test
13:05:46 <XueFeng> use /nick name can change you nick
13:06:21 <Qiming> but no one has yet claimed that item
13:07:03 <Qiming> elynn has committed a new patch to "feature rich nova server"
13:07:31 <Qiming> https://review.openstack.org/467108
13:07:35 <elynn> I test it locally today, and I think it could work as expected.
13:07:37 <Qiming> it is still a WIP
13:08:08 <elynn> I didn't add any version information to those new properties,
13:08:17 <XueFeng> I think liyi is test api and do pathch for the doc. We can leave a comment in his patch
13:08:19 <elynn> So I just leave it as WIP
13:08:20 <Qiming> 486 lines added to the server profile, not counting the test cases, :)
13:08:38 <Qiming> got it, elynn
13:08:59 <Qiming> Guest53985 is liyi probably?
13:09:01 <elynn> What do you suggest about adding version info to those new properties?
13:10:16 <Qiming> ideally, I'd suggest we port all profile implementation to oslo.versionedobjects
13:11:02 <elynn> okay.. I will look into that.
13:11:08 <Qiming> however, if we can hide the details from end users, maybe it is still okay to use the current versioning support
13:11:10 <XueFeng> Guest53983 is yaofenghua I think
13:11:13 <XueFeng> https://review.openstack.org/#/c/482370/1
13:11:20 <Guest95123> yes
13:11:23 <XueFeng> I leave a comment in liyi's patch
13:12:13 <Qiming> XueFeng, thanks
13:12:41 <Qiming> I'm not aware of any other changes to senlin engine
13:13:28 <Qiming> do we still have a lot to do regarding runtime_data implementation?
13:13:52 <ruijie> Qiming, I want to add some data to action.outputs after execute is
13:14:16 <Qiming> em, that is good suggestion
13:14:45 <Qiming> although we don't have a strict schema for action inputs or outputs
13:15:09 <ruijie> but that may need to query databases, e.g  nodes created
13:15:21 <Qiming> I'd still prefer a systematic way to get it done
13:16:11 <Qiming> different actions produce different data as outputs
13:16:29 <Qiming> maybe we should document them in the design docs
13:16:42 <ruijie> yes, that should be formatted
13:17:16 <Qiming> so at least we are still not losing control of what was dumped in P version and what was added later in Q version ... so on and so forth
13:18:02 <Qiming> we don't need a strict protocol on that, right?
13:18:24 <Qiming> I mean we don't want to check if a field exists and if it is actually an integer for example
13:18:53 <Qiming> actions are primarily provided as a debugging tool
13:19:05 <ruijie> it could be something like the action payload?
13:19:40 <Qiming> just document what are expected as inputs and/or outputs would be okay
13:20:15 <Qiming> the action.outputs only surface when you do 'action show'
13:20:34 <ruijie> it is mainly used for later processing in policies and dumping events to MQ
13:20:40 <Qiming> although internally it may get checked by some policies
13:20:54 <Qiming> right, events ...
13:21:30 <Qiming> go ahead and do it, :)
13:21:41 <ruijie> okay :)
13:22:03 <Qiming> moving on
13:22:18 <Qiming> we got some progress on RDO packaging, right?
13:22:28 <XueFeng> Yes
13:22:29 <XueFeng> Qiming
13:22:35 <XueFeng> https://bugzilla.redhat.com/show_bug.cgi?id=1426551
13:22:37 <openstack> bugzilla.redhat.com bug 1426551 in Package Review "Review Request: Senlin - is a clustering service for OpenStack" [Unspecified,Assigned] - Assigned to jpena
13:22:47 <Qiming> someone has taken over that job
13:22:47 <XueFeng> Got  process these days
13:22:54 <Qiming> cool
13:23:12 <XueFeng> https://review.rdoproject.org/r/#/c/7470/
13:23:27 <XueFeng> and in rdo project code is in review
13:23:28 <Qiming> if there are things needed from senlin side, just let yell in #senlin channel
13:23:42 <XueFeng> ok
13:25:22 <Qiming> any other updates regarding pike work items?
13:26:30 <XueFeng> https://blueprints.launchpad.net/senlin/+spec/improve-vm-name
13:27:39 <Qiming> any updates to that bp, XueFeng ?
13:27:59 <xinhuili_> what kind of order? could you explain more,XueFeng
13:28:31 <XueFeng> I want to ask which state about this bp
13:28:50 <Qiming> it is there
13:30:00 <Qiming> no follow up after the review
13:30:06 <Qiming> no new patch
13:30:14 <Qiming> no response to the comments
13:30:42 <XueFeng> So the better way is  to add a column to the cluster table, followed by an upgrade of the cluster apis?
13:30:55 <xinhuili_> why index is important
13:31:05 <xinhuili_> I do not get the point of this BP
13:31:14 <Qiming> index will be used as part of node name
13:31:41 <Qiming> if name can be predictable, the 'hostname' of a VM is predictable
13:31:44 <xinhuili_> is any dependency on the node name
13:31:54 <xinhuili_> by exited adoption?
13:32:03 <xinhuili_> exsited
13:32:04 <Qiming> and it will help applications running in nodes
13:32:31 <Qiming> currently, there are two possibilities, all nodes named the same
13:32:42 <Qiming> or names have random suffixes
13:32:58 <Qiming> neither one meets the requirement sketched above
13:33:52 <Qiming> if anyone is interested in working on this, I can share my thoughts on it
13:34:22 <XueFeng> ok
13:35:01 <Qiming> actually, I wrote some ideas in the first review
13:35:06 <XueFeng> current the name is the same, it's better to improve I think
13:35:32 <elynn> If we don't specify name in profile, I think it's different?
13:35:37 <Qiming> it is definitely a good feature
13:36:00 <Qiming> different is not enough
13:36:07 <Qiming> users want the name predictable
13:36:38 <Qiming> say, 'k8sworker-1', 'k8sworker-2'
13:36:49 <elynn> Like cluster_name-index?
13:36:50 <Qiming> or 'redis-1', 'redis-2', ...
13:36:55 <Qiming> yes
13:37:20 <elynn> Yes, that's reasonable.
13:37:22 <Qiming> then user can predict the node names
13:37:41 <Qiming> not 'node-blah', 'node-xyz123', ...
13:37:51 <elynn> I remember our node name have some rules like that.
13:38:13 <elynn> Why not just make it the same as node name?
13:39:25 <xinhuili_> or in order words, what kind of name protocal will be preferred
13:39:25 <elynn> It's easy to do so, some monitoring system is based on vm name I guess.
13:39:35 <xinhuili_> for this BP
13:39:40 <xinhuili_> easy to map?
13:39:45 <xinhuili_> keep unified?
13:39:54 <xinhuili_> fir for different use cases?
13:39:59 <xinhuili_> fit
13:40:36 <Qiming> xinhuili_, read the first comment I left there
13:41:09 <Qiming> #link https://review.openstack.org/#/c/461673/
13:41:32 <Qiming> I was suggesting we may this configurable
13:41:46 <xinhuili_> reading ...
13:42:07 <xinhuili_> good idea to make the rule configurable
13:42:42 <Qiming> okay, back to the agenda
13:43:30 <Qiming> don't know if people have interests in standing up a k8s cluster using senlin
13:44:06 <Qiming> I don't see that a big challenge, just need hands to work it out
13:44:10 <XueFeng> Qiming , share the idea to us
13:44:30 <Qiming> XueFeng, not today, the agenda is already too long
13:44:31 <XueFeng> More detail
13:44:43 <Qiming> we are not supposed to overrun
13:45:19 * Qiming trying to find a pointer
13:45:50 <Qiming> here: https://github.com/kubernetes/kubernetes/tree/master/cluster/openstack-heat/kubernetes-heat
13:46:07 <Qiming> some hints for those who are interested
13:46:44 <Qiming> it is part of official k8s code
13:46:48 <elynn> I guess that heat template is not easy to use :P
13:47:03 <Qiming> yes, that is where we can help
13:47:59 <Qiming> recent feedbacks: users still have lock problems
13:48:12 <Qiming> maybe because they were using old version
13:48:19 <XueFeng> Yes ,in Ocata version
13:48:23 <Qiming> the complaint I got is about 3.0.0
13:48:44 <ruijie> I am :)
13:49:06 <Qiming> ruijie, you got lock problems recently?
13:49:27 <XueFeng> Can't clean lock when restart engine
13:49:27 <ruijie> didnt't see that problem again
13:49:36 <Qiming> okay
13:49:48 <ruijie> I said I am interested in the interations of senlin & k8s
13:50:00 <Qiming> please ring the bell when you see things abnormal
13:50:05 <XueFeng> http://eavesdrop.openstack.org/irclogs/%23senlin/%23senlin.2017-07-07.log.html
13:50:25 <Qiming> #topic sydney proposals
13:50:31 <Qiming> it is drawing near
13:50:49 <Qiming> pls think about what you want to propose if any
13:51:17 <XueFeng> ok
13:51:33 <Qiming> if you want, you can pass your draft across the team for comments
13:51:50 <elynn> Do you have any proposals?
13:51:50 <XueFeng> About adoptation we can do a proposal
13:52:20 <Qiming> adoption be one, but the story is not that attractive for users
13:52:37 <Qiming> I'd like to see k8s standing up on senlin
13:52:55 <Qiming> with very few parametes to tune
13:53:04 <XueFeng> Good
13:53:21 <elynn> That's another proposal I guess
13:53:30 <XueFeng> And if we use k8s+ senlin
13:53:33 <Qiming> but ... I myself is overrrrr committed at the time
13:53:48 <XueFeng> Should we intergate kuryr?
13:54:12 <elynn> I could try to do some poc codes if you like
13:54:16 <Qiming> XueFeng, it depends
13:54:28 <XueFeng> I am interested
13:54:32 <Qiming> so .... good elynn
13:55:17 <Qiming> from k8s side, they are assumming the cloud provider provides a clustering service
13:55:30 <Qiming> on openstack that interface has not been claimed
13:55:43 <elynn> But you have to share more info to us, ur the expert of k8s :)
13:55:52 <Qiming> we are almost a perfect fit for that role
13:55:56 <elynn> Like how lb and network work.
13:56:01 <XueFeng> https://blueprints.launchpad.net/senlin/+spec/integration-kuryr-for-container-network
13:56:08 <Qiming> lb should be hidden, ideally
13:56:15 <XueFeng> This is the bp fo integration kuryr
13:56:29 <Qiming> just expose one IP for a cluster as the service IP
13:56:56 <elynn> I remember that k8s has api service and can be integrate with lb I assume?
13:57:02 <Qiming> thanks for proposing that, XueFeng
13:57:22 <Qiming> it would be better if someone has started the coding/testing work, :D
13:57:30 <XueFeng> My pleasure
13:57:36 <Qiming> yes, k8s has that
13:57:52 <Qiming> but ... I'm very disappointed by lbaas on openstack
13:57:59 <Qiming> tbh
13:58:51 <Qiming> I'd prefer calling out to haproxy or nginx directly ...
13:59:05 <elynn> You mean we can get rid of laas in k8s cluster?
13:59:26 <elynn> Hmm, that's another approach
13:59:27 <Qiming> avoid ocativa when possible, I mean
13:59:33 <Qiming> it is terrible
13:59:40 <XueFeng> We can do in this way
13:59:41 <Qiming> sorry, running out of time
13:59:53 <Qiming> thanks for joining boys & girls
13:59:57 <Qiming> good night
14:00:02 <Qiming> #endmeeting