13:01:00 <Qiming> #startmeeting senlin
13:01:01 <openstack> Meeting started Tue Apr 18 13:01:00 2017 UTC and is due to finish in 60 minutes.  The chair is Qiming. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:02 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:04 <openstack> The meeting name has been set to 'senlin'
13:03:23 <XueFeng> hi,all
13:03:27 <elynn> o/
13:04:17 <Qiming> hi
13:04:41 <Qiming> yanyan just texted me that he won't be able to join us today
13:04:51 <XueFeng> ok
13:05:22 <Qiming> #topic pike work items
13:05:33 <Qiming> let's get started
13:05:38 <Qiming> #link https://etherpad.openstack.org/p/senlin-pike-workitems
13:05:49 <XueFeng> ok
13:05:58 <Qiming> API version conversion bug has been fixed
13:06:04 <Qiming> so .. I'm removing that item
13:06:29 <Qiming> VDU profile i.e. feature rich server
13:06:35 <Qiming> any update, elynn ?
13:06:42 <elynn> No update...
13:06:51 <elynn> haven't got time to do it...
13:06:57 <Qiming> okay
13:07:06 <elynn> I will try to upload patches before summit
13:07:27 <Qiming> cluster check redesign, no progress afaik
13:07:41 <XueFeng> yes, Qiming
13:07:51 <XueFeng> debug tacker+senlin
13:07:56 <XueFeng> these day
13:07:56 <Qiming> node adopt, two patches for review
13:08:02 <XueFeng> with xuhaiwei
13:08:21 <XueFeng> ok, will review
13:08:36 <Qiming> good, you may want to write another article when the integration is done, :)
13:08:44 <XueFeng> and CMCC yaofenghua send a mail to me
13:09:02 <XueFeng> about node adopt the also need
13:09:04 <Qiming> from suzhou or beijing?
13:09:16 <XueFeng> haha,Qiming, yes
13:09:21 <XueFeng> We will write
13:09:41 <Qiming> cool, will be good news for NFV users
13:09:41 <XueFeng> Maybe suzhou
13:10:30 <Qiming> alright, pls help review the two patches
13:10:39 <XueFeng> ok, I will:)
13:10:45 <Qiming> then I'll proceed on the api level changes
13:10:58 <Qiming> scaling improvements, mostly by ruijie
13:11:09 <lixinhui> hi
13:11:12 <Qiming> haven't heard a thing from him
13:11:24 <XueFeng> hi,xinhui
13:11:35 <Qiming> oh, our godess finally shows up
13:11:39 <lixinhui> Sorry for the late dail in
13:11:47 <elynn> haha
13:11:48 <lixinhui> :)
13:12:07 <XueFeng> ruijie also have serval patches need to be review
13:12:16 <Qiming> we are on senlin pike work items, https://etherpad.openstack.org/p/senlin-pike-workitems
13:13:17 <Qiming> I think I have left comments to most of his patches if not all
13:13:39 <XueFeng> yes
13:14:15 <XueFeng> I also reviewed some of them:)
13:14:42 <Qiming> about adding health check before deletion or scaling, I think we have reached a common understanding
13:15:22 <XueFeng> Yes, the cluster.health_check has been merged
13:15:23 <Qiming> he was tring to do the invocation from policy side, while I am more inclined to make them part of the node logics
13:15:32 <Qiming> yes
13:16:06 <Qiming> the policy decides whether health should be checked, but the actual check logic still should be part of cluster_action or node_action
13:16:31 <Qiming> that was something we have spent time digging into
13:16:41 <Qiming> I believe he is still on it these days
13:16:57 <Qiming> moving on
13:17:02 <Qiming> next thing is about RDO
13:17:07 <XueFeng> So this pathch still need updte https://review.openstack.org/#/c/456453/
13:17:19 <XueFeng> No update for RDO this week
13:17:20 * Qiming is looking
13:17:40 <Qiming> right, XueFeng, that one should be reworked
13:18:01 <XueFeng> Ok, I got the point you mentioned
13:18:21 <Qiming> I believe he was more affected by the load-balancing policy, where we embedded many operations into the policy itself
13:18:54 <XueFeng> yes
13:19:02 <Qiming> however, our design philosophy for polices are just some rules to be checked/enforced before/after action execution
13:19:51 <Qiming> it doesn't mean policies should intiate some operations on physical resources unless the policy is owning that resource, e.g. lb policy
13:20:21 <Qiming> next thing
13:20:28 <Qiming> health management
13:20:38 <Qiming> two patches about workflow support are in now
13:20:58 <Qiming> need to enable them in health policy and recover action
13:21:17 <Qiming> I have moved the logic from nova server up, to the general node level
13:21:49 <Qiming> because I think workflows are generic enough to invoke many openstack services
13:22:00 <XueFeng> ok
13:22:05 <Qiming> xinhui approved that change
13:22:14 <XueFeng> :)
13:22:45 <Qiming> I have moved the load-balancer check and event emitting work back to FEATURES.rst, if you haven't noticed
13:22:49 <Qiming> this is a sad move
13:23:01 <Qiming> we hoped that Octavia is open to such a change
13:23:20 <XueFeng> I know that
13:23:30 <Qiming> however, ... things didn't go smooth although I know xinhui and friends have tried hard on that
13:23:52 <Qiming> the whole thing has been there for one year, no comment to the BP, no comment to the patch
13:24:16 <XueFeng> Maybe we can have another choice to do this thing
13:24:18 <Qiming> if BP is not approved, patch won't be reviewed, (according to one of the reviewers)
13:24:43 <Qiming> we really don't want to do ping or HTTP get by ourselves
13:24:57 <XueFeng> en
13:24:57 <Qiming> it is not a big deal from implementation's perspective
13:25:15 <Qiming> just we don't want to reinvent the wheel, let the right service do the right thing
13:25:34 <Qiming> we want to be good citizen
13:25:42 <XueFeng> right
13:25:53 <Qiming> anyway, we can live with it
13:26:21 <Qiming> when we get more urgent requests, we can quickly stand up our own version of health checking
13:26:43 <lixinhui> yes, Qiming
13:27:13 <Qiming> the only barrier is that we need a place to issue the ping or HTTP GET call
13:27:22 <Qiming> that 'place' must be on the external subnet
13:28:09 <Qiming> ping the nodes from private subnet is not optimal, but it is better than nothing
13:28:33 <Qiming> moving on
13:28:39 <XueFeng> ok
13:28:41 <Qiming> tempest test
13:28:58 <Qiming> the local test for non-admin account is done, XueFeng ?
13:29:05 <XueFeng> Yes
13:29:07 <Qiming> cool
13:29:13 <XueFeng> I think so
13:29:15 <XueFeng> :)
13:29:19 <Qiming> I'm seeing that the whole patch is not breaking the gate
13:29:27 <XueFeng> Yes
13:29:37 <Qiming> bravo
13:29:40 <XueFeng> We also can pass in local
13:29:56 <Qiming> we have not yet started senlinclient functional tests
13:29:56 <XueFeng> for 21:25 < Qiming> just we don't want to reinvent the wheel, let the right service do the right thing
13:30:06 <XueFeng> yes
13:30:11 <XueFeng> no
13:30:26 <XueFeng> no time to to it now
13:30:26 <Qiming> ?
13:30:35 <XueFeng> to do
13:31:06 <XueFeng> It's priority is low
13:31:21 <XueFeng> About senlinclinet I think
13:31:25 <Qiming> questions on my previous sentence?
13:31:38 <XueFeng> So we need do senlin+tacker firslty
13:31:46 <Qiming> okay
13:32:16 <Qiming> just from service/client stability's point of view
13:32:43 <Qiming> having functional tests there will ensure that any breaking changes between the server/client will bubble up quickly
13:32:45 <XueFeng> yes
13:32:57 <Qiming> e.g. changes to sdk
13:33:19 <Qiming> sometimes break us in an unnoticeable way
13:33:48 <XueFeng> Yes, it is necesary
13:34:05 <Qiming> next thing is about cluster-recover improvement for lb policy
13:34:18 <Qiming> I know ruijie still have some ideas to improve it
13:34:28 <Qiming> so we'd better leave that item there
13:34:39 <Qiming> besides items on the etherpad
13:34:47 <Qiming> I'm looking into the performance overhead problem
13:34:51 <XueFeng> We can add to etherpad
13:35:17 <XueFeng> Some patches about these need review as well
13:35:49 <XueFeng> This one
13:35:50 <Qiming> elynn previously reported that scaling a cluster takes time
13:35:54 <XueFeng> https://review.openstack.org/#/c/455151/
13:36:04 <Qiming> it is mostly related to the cluster.rt and node.rt
13:36:15 <elynn> yes
13:36:21 <Qiming> we instantiate Cluster and Node instances
13:36:41 <Qiming> each instantiation of a Node in turn instantiates a Profile ...
13:37:08 <Qiming> when revisiting the code, my feeling is that we can avoid db calls in many places
13:37:17 <XueFeng> elynn, give a bug link
13:37:34 <Qiming> we can even avoid instantiate nodes and still break nothing
13:37:59 <Qiming> just populate cluster.rt['nodes'] with senlin.objects.node.Node instance is enough
13:38:05 <elynn> https://bugs.launchpad.net/senlin/+bug/1677268
13:38:07 <openstack> Launchpad bug 1677268 in senlin "scale-out API will take a long time to response when too many nodes in a cluster" [High,Triaged] - Assigned to Ethan Lynn (ethanlynn)
13:38:20 <Qiming> and that operation takes only one ovo call
13:38:48 <Qiming> if we still have performance problems, we can take a further step, recording only node ids in the runtime data
13:39:12 <Qiming> I'm not referencing that bug number in my patches
13:39:26 <Qiming> although ... they are actually remotely related
13:40:03 <Qiming> e.g. https://review.openstack.org/#/c/457493/
13:40:39 <elynn> :) I concern if these series of patches can back port to ccata...
13:40:42 <Qiming> this one: https://review.openstack.org/#/c/457508/
13:41:56 <Qiming> patch 457508 directly exposes profile.created_at into node dict, so that we don't need to visit node.rt['profile'] when searching nodes with the oldest profile
13:42:22 <Qiming> we can use ovo Node object in cluster.rt['nodes']
13:42:39 <Qiming> will propose related patches to get this whole thing done
13:43:11 <Qiming> em, I self appoved #457493
13:43:16 <Qiming> :P
13:43:24 <Qiming> it is in now
13:43:26 <XueFeng> hh
13:43:32 <elynn> great
13:43:51 <lixinhui> cool
13:43:54 <Qiming> that's all for pike work items so far
13:43:58 <Qiming> moving on
13:44:03 <Qiming> #topic boston summit prep
13:44:07 <Qiming> where are we?
13:44:19 <Qiming> our godess?
13:45:08 <Qiming> XueFeng, yours?
13:45:35 <XueFeng> May 11
13:45:39 <XueFeng> afternoon
13:46:24 <Qiming> I mean how are the preparation now?
13:46:29 <Qiming> do you need any help?
13:46:59 <XueFeng> https://review.openstack.org/#/c/449471/
13:47:18 <XueFeng> In tacker ,this patch is in review
13:47:31 <Qiming> yes?
13:47:51 <XueFeng> And about demo
13:47:56 <XueFeng> debuging
13:48:10 <Qiming> okay
13:48:40 <XueFeng> We can deploy VNF from tacker+heat+senlin
13:48:58 <Qiming> do you guys think it necessary to spend some time together, maybe later this week or early next week for a face to face discussion?
13:49:10 <XueFeng> And now we change to the real image
13:49:55 <Qiming> ya, an imporant step
13:50:06 <XueFeng> I can come
13:50:27 <Qiming> by the way, yanyan is being relocated to Singapore
13:50:33 <lixinhui> I am on-site service these days
13:50:45 <lixinhui> I heard about that
13:50:48 <lixinhui> happy for him
13:50:54 <Qiming> we are "cheaping hime away"
13:51:03 <Qiming> jian4 xing2
13:51:37 <lixinhui> we should dinner together before his leaving
13:51:56 <XueFeng> haha
13:51:57 <Qiming> em, he told us not to mention this to anyone, and he at the same time has already told everyone, :D
13:52:05 <elynn> glad he have a better future
13:52:17 <elynn> haha
13:52:55 <XueFeng> what is jian4 xing2?
13:53:11 <Qiming> 贱行, in Chinese
13:53:53 <XueFeng> oh:)
13:54:01 <Qiming> #topic open discussion
13:54:02 <lixinhui> 豁出去了
13:54:28 <lixinhui> 不减肥了
13:54:35 <Qiming> if you need any helps on preparing the talks, you can reach out to the team
13:54:53 <elynn>13:54:54 <Qiming> you were never on that direction as far as I know it
13:55:03 <lixinhui> ...
13:55:08 <Qiming> *never*
13:55:15 <elynn> Nice job
13:55:41 <lixinhui> Will send out a draft agenda for your comments
13:56:02 <Qiming> top priorities in life: 1 eat, 2 eat, 3 goto 1
13:56:10 <Qiming> okay
13:56:30 <Qiming> I'm mostly stuck by Go and k8s these days
13:56:38 <Qiming> it is a bigger ... mess
13:56:40 <XueFeng> hhh
13:56:52 <lixinhui> you need more coffee
13:57:21 <XueFeng> Qiming
13:57:22 <Qiming> no .... I need more sleep, seriously
13:57:30 <XueFeng> Another thing
13:57:32 <Qiming> my girl is too naughty
13:57:45 * Qiming listening
13:58:05 <XueFeng> Can senlin run in openstak K version
13:58:33 <Qiming> not quite sure, but we did successfully get it up in Juno
13:58:47 <Qiming> you need to mind the sdk version used
13:58:51 <lixinhui> XueFeng
13:58:57 <lixinhui> Senlin can
13:59:01 <XueFeng> ok
13:59:09 <XueFeng> good news
13:59:10 <lixinhui> I once tried and ran up senlin with VIO K version
13:59:25 <Qiming> we are not bound to any specific version of any services, except for sdk
13:59:37 <Qiming> thanks for the confirmation, xinhui
13:59:50 <lixinhui> my pleasure
13:59:50 <Qiming> we are running out of time, thanks for joining, boys and girls
13:59:57 <Qiming> good night
13:59:58 <XueFeng> Ok, if need in K version, will ask you guys
14:00:05 <Qiming> sure
14:00:08 <Qiming> #endmeeting