13:00:26 <yanyanhu> #startmeeting senlin
13:00:27 <openstack> Meeting started Tue Feb 28 13:00:26 2017 UTC and is due to finish in 60 minutes.  The chair is yanyanhu. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:28 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:00:32 <openstack> The meeting name has been set to 'senlin'
13:00:33 <openstack> yanyanhu: Error: Can't start another meeting, one is in progress.  Use #endmeeting first.
13:00:39 <yanyanhu> hi
13:00:51 <elynn_> hi
13:00:56 <Qiming> hi
13:01:01 <yanyanhu> hi, elynn_ Qiming
13:01:19 <XueFengLiu> hi
13:01:23 <XueFengLiu> all
13:01:27 <yanyanhu> hi, XueFengLiu
13:01:39 <XueFengLiu> :)
13:02:03 <ruijie> sorry for late :)
13:02:04 <yanyanhu> hi, ruijie
13:02:09 <ruijie> hi yanyanhu
13:02:10 <yanyanhu> not at all :)
13:02:30 <yanyanhu> please review the agenda and add items you want to discuss
13:02:31 <yanyanhu> https://wiki.openstack.org/wiki/Meetings/SenlinAgenda#Agenda_.282017-02-28_1300_UTC.29
13:02:56 <ruijie> we are in pike now ?
13:03:03 <Qiming> same question, ;)
13:05:36 <yanyan> hi, sorry, just dropped...
13:05:40 <yanyan> network is very unstable
13:05:49 <yanyan> ruijie, yes, we are in pike now
13:06:49 <Qiming> created pike etherpad: https://etherpad.openstack.org/p/senlin-pike-workitems
13:07:08 <yanyan> hi
13:07:16 <yanyan> thanks, Qiming
13:08:52 <yanyanhu> hi, sorry, dropped again...
13:09:06 <Qiming> yanyanhu, seems your network connection is very bad
13:09:16 <yanyanhu> yes... don't know why
13:09:20 <yanyanhu> it is ok usually
13:09:34 <yanyanhu> so where are we now?
13:09:48 <Qiming> looking at API section I guess
13:09:54 <Qiming> the 436288 is merged
13:10:06 <yanyanhu> ok
13:10:08 <Qiming> need more patches on request object and api layer
13:10:49 <yanyanhu> and also test I guess
13:11:33 <yanyanhu> ok, next one
13:11:36 <yanyanhu> "Feature Rich" Nova Server
13:11:42 <yanyanhu> hi, elynn_
13:11:47 <elynn_> hi
13:11:50 <yanyanhu> so the first template works now?
13:11:54 <yanyanhu> saw your mail :)
13:12:11 <yanyanhu> could you please paste the link to etherpad as well so we can easily access it
13:12:17 <elynn_> yes, a simple template works now.
13:12:26 <Qiming> great
13:12:26 <elynn_> I will do it.
13:12:41 <yanyanhu> ok
13:12:43 <Qiming> nova server is never about a bare server resource itself
13:13:07 <elynn_> yes
13:13:19 <Qiming> so I'm wondering if we should merge it back to nova server profile
13:13:21 <elynn_> And still working on update operation
13:13:39 <Qiming> left some comments to both patches
13:13:47 <elynn_> too early to merge it back
13:13:53 <Qiming> yep
13:14:08 <elynn_> I saw them, will revise it.
13:14:19 <Qiming> how about we merge them in small patches?
13:14:59 <elynn_> Will it break the compatibility
13:15:01 <elynn_> ?
13:15:02 <Qiming> another option is one patch with a thousand lines of code, including unit tests and api/functional tests
13:15:20 <Qiming> it doesn't matter now because ocata is out
13:15:53 <elynn_> do you like the later option ??
13:16:06 <Qiming> what do you think?
13:16:25 <elynn_> after it's stable, I will merge it back with small patches...
13:16:38 <Qiming> that's fine, though
13:16:56 <Qiming> I was suggesting we work these two threads in parallel
13:17:25 <Qiming> if that is possible
13:17:29 <elynn_> I guess that will break many things...
13:17:33 <Qiming> say, add port support
13:17:47 <Qiming> then add other properties one by one
13:18:31 <elynn_> How about using a new version for this profile?
13:18:38 <elynn_> like os.nova.server-2.0?
13:18:42 <Qiming> the implication for code in contrib is that we may throw it away
13:18:54 <Qiming> yes, we will bump versions along the way
13:18:55 <elynn_> could we do that?
13:19:19 <elynn_> and also leave the code of version 1.0 I guess?
13:19:25 <Qiming> the version support is very limited
13:19:27 <yanyanhu> if so, should we maintain os.nova.server-1.0 and os.nova.server-2.0 at the same time?
13:20:00 <yanyanhu> or we add deprecated warning for os.nova.server-1.0 profile once 2.0 is supported?
13:20:03 <elynn_> I'm more afraid of compatibility issues.
13:20:11 <yanyanhu> me too...
13:20:15 <Qiming> a valid concern
13:20:50 <Qiming> that's the reason I suggested we do it in small patches
13:21:18 <Qiming> we cannot add 500 lines of new code and ensure back compatibility
13:21:33 <yanyanhu> I think Qiming's suggestion is feasible
13:21:47 <Qiming> and ... once the POC is there, people usually hate to rewrite it again
13:21:48 <yanyanhu> we enhance the current nova server step by step
13:22:02 <yanyanhu> until we meet uncompatible issue
13:22:07 <yanyanhu> then we stop to figure it out
13:22:18 <yanyanhu> and them keep going go
13:22:21 <Qiming> right, we don't know how difficult it will be yet
13:22:38 <yanyanhu> s/them/then
13:23:03 <elynn_> that
13:23:19 <elynn_> that's right.
13:23:26 <yanyanhu> just one more question, does this template meet the requirement raised by xinhui at the beginning?
13:23:37 <Qiming> one lesson we learned in the past is about deploying container clusters
13:23:49 <Qiming> that has been done separately on github
13:23:56 <elynn_> This template is what xinhui found.
13:24:05 <Qiming> but code is kind of abandoned
13:24:11 <elynn_> When xinhui gets back, I will check with her.
13:24:25 <Qiming> alright, it's up to you, sir
13:24:35 <yanyanhu> elynn_, I see
13:24:50 <yanyanhu> great
13:25:09 <Qiming> if bran has been involved into this, I'm optimistic on that, yanyanhu
13:25:24 <yanyanhu> ok :)
13:26:03 <yanyanhu> so lets move on if there is no more question on this item?
13:26:27 <yanyanhu> Engine Improvement
13:26:41 <XueFengLiu> hi,yanyan
13:26:44 <yanyanhu> hi, XueFengLiu
13:26:52 <yanyanhu> any new progress on cluster check
13:26:59 <XueFengLiu> Two patches has been merged
13:27:18 <XueFengLiu> the api for target delete
13:27:46 <XueFengLiu> And delete actions when delete cluster.
13:28:02 <yanyanhu> nice
13:28:10 <XueFengLiu> This is the a new one for cluster check actions
13:28:22 <yanyanhu> so action number will be reduced now
13:28:56 <yanyanhu> so, please keep working on it, thanks a lot :)
13:29:02 <XueFengLiu> https://review.openstack.org/#/c/435822/
13:29:07 <XueFengLiu> Need review
13:29:16 <XueFengLiu> and modify
13:29:30 <yanyanhu> great, that is what we want
13:29:44 <yanyanhu> will help to review it
13:29:52 <XueFengLiu> Thanks:)
13:29:58 <yanyanhu> :)
13:30:02 <yanyanhu> ok, next one
13:30:07 <yanyanhu> - Add senlin to RDO
13:30:12 <yanyanhu> XueFengLiu, still your turn :)
13:30:22 <yanyanhu> Qiming said you have filed bug for it?
13:30:31 <XueFengLiu> Yes
13:30:53 <XueFengLiu> I report a bug to rdo project.
13:31:03 <XueFengLiu> And next
13:31:15 <XueFengLiu> We need write a spec for senlin
13:31:34 <yanyanhu> like the spec in openstack?
13:31:41 <yanyanhu> or it is a template for building package?
13:31:47 <XueFengLiu> No , for rpm
13:31:52 <yanyanhu> ok
13:32:14 <XueFengLiu> I generate a spec by python command automic
13:32:19 <yanyanhu> so the maintainer should be responsible for defining how to build the rpm
13:32:30 <XueFengLiu> but it seems need do some change
13:32:39 <XueFengLiu> Yes
13:32:58 <Qiming> better raise questions in the #rdo channel if helps needed
13:33:08 <XueFengLiu> OK
13:33:12 <yanyanhu> XueFengLiu, may thanks for starting this job
13:33:16 <yanyanhu> many
13:33:37 <XueFengLiu> My pleasure
13:33:39 <XueFengLiu> :)
13:33:50 <yanyanhu> ok, lets move on?
13:33:54 <XueFengLiu> OK
13:33:56 <yanyanhu> functional test for client
13:33:59 <yanyanhu> https://review.openstack.org/#/c/436338/
13:34:05 <yanyanhu> it is ready for review now
13:34:22 <Qiming> wrt to rpm spec, I'm not sure if the fully fleged version is created manually
13:34:25 <yanyanhu> btw, we met some problems which failed our gate in last several days
13:35:27 <yanyanhu> and that problem has been fixed
13:35:44 <Qiming> partially
13:36:08 <Qiming> still not sure if we should move post-config stuff to project-config
13:36:08 <yanyanhu> so, I guess we still need some change on our test hooks?
13:36:16 <XueFengLiu> QiMing, I tested to genereate by python setup.py bdist_rpm
13:36:17 <Qiming> need to keep an eye on it
13:36:25 <yanyanhu> Qiming, yes, looks sean suggested that in the mail
13:36:35 <yanyanhu> but there is still no concensus I guess
13:36:46 <Qiming> yep, both approaches have drawbacks
13:37:10 <XueFengLiu> And the spec and rpm are generated. But not whole
13:37:12 <Qiming> moving things to gate will increase infra workload
13:37:25 <Qiming> yes, XueFengLiu, I know that
13:37:51 <yanyanhu> Qiming, yes, that means any change could need revision in project-config
13:38:03 <Qiming> I mean, for the generation of a FULL version, there may be other tools we don't know
13:38:05 <yanyanhu> that will significantly increase the overhead of infra team
13:38:47 <Qiming> keeping those settings in individual projects will be a pain for devstack tools
13:39:01 <Qiming> it may occasionally break this project or that
13:39:19 <yanyanhu> yes, there is no perfect solution currently I guess...
13:39:53 <yanyanhu> anyway, about the functional test for client, once the gate job is ready, we can start working on the base test in senlinclient side
13:40:10 <Qiming> yep, that is a great starting point
13:40:10 <yanyanhu> XueFengLiu, will work together with you to setup the basic one
13:40:23 <XueFengLiu> OK
13:40:29 <XueFengLiu> Thanks
13:40:36 <Qiming> just want to remind everyone, we are supposed to write functional tests for OSC plugin only
13:40:50 <yanyanhu> yes
13:40:56 <yanyanhu> that's what we need
13:41:22 <yanyanhu> so lets move on?
13:41:36 <Qiming> sure
13:41:39 <yanyanhu> Support CLUSTER_RECOVER action in LB policy
13:41:45 <yanyanhu> hi, ruijie
13:41:49 <yanyanhu> your turn :)
13:41:53 <ruijie> yes yanyanhu
13:42:03 <ruijie> I proposed 1st patch today
13:42:15 <yanyanhu> https://review.openstack.org/438889
13:42:16 <yanyanhu> this one
13:42:21 <ruijie> https://review.openstack.org/#/c/438889/
13:42:38 <ruijie> yes, basicly will add logic to post_op()
13:42:51 <ruijie> to handle the data passed from do_recover(...)
13:42:55 <yanyanhu> this is important if we want lb policy works with cluster recovery action
13:43:29 <Qiming> maybe we will need some logic for CLUSTER_CHECK as well?
13:43:32 <yanyanhu> ruijie, also we may need to consider the situation that the IP doesn't change
13:43:37 <yanyanhu> after node is recovered
13:43:44 <yanyanhu> Qiming, yes
13:43:47 <yanyanhu> good idea
13:44:16 <ruijie> yanyanhu, you mean the server ip will changed during running?
13:44:20 <yanyanhu> but one question is we are not sure whether lbaas support to manually set the health status of a member
13:44:49 <yanyanhu> ruijie, I mean if the IP of node doesn't change during recovery progress, maybe we don't need to perform lb member update operation
13:45:22 <ruijie> yes, I noticed the cluster_recover is under revising to sync the status of server
13:45:44 <ruijie> will add logic to cluster_recover() too
13:45:51 <yanyanhu> ruijie, thanks :)
13:45:55 <Qiming> if a server is recreated, the IP may get changed
13:46:24 <ruijie> yes, if the server is just soft-reboot, will not remove it from lb
13:46:25 <yanyanhu> Qiming, yes, that is likely to happen
13:46:26 <Qiming> so no matter recovery is a success or not, we need to check it
13:46:40 <yanyanhu> sure, the check is necessary
13:47:23 <ruijie> I am thinking we still use "creation"/"deletion" to handle the data
13:47:39 <yanyanhu> ruijie, you mean?
13:48:09 <ruijie> creation data contains the node is that need to be added to lb, and deletion means need to be removed
13:48:17 <ruijie> in the action layer we check that
13:48:27 <Qiming> if they are not sufficient
13:48:51 <Qiming> feel free to dump the related node info into the 'outputs' field of the action
13:48:54 <ruijie> and then pass it to post_op
13:49:17 <ruijie> okay Qiming, will propose another tomorrow
13:49:22 <ruijie> patch
13:49:40 <yanyanhu> ruijie, great, and please ping me if you need any help on lb policy
13:49:42 <yanyanhu> thanks
13:49:49 <ruijie> sure yanyanhu:)
13:49:59 <Qiming> I believe you have a good understanding about how policy works
13:50:15 <ruijie> :)
13:50:24 <yanyanhu> :)
13:50:32 <yanyanhu> so lets all in the list I think
13:50:47 <yanyanhu> and we have some items in Wishlist or with Low Priority
13:51:02 <yanyanhu> but I guess no progress on them in this week
13:51:16 <yanyanhu> that's all, sorry misclicked...
13:51:39 <yanyanhu> any more items in progress?
13:51:59 <yanyanhu> ok, open discussion now
13:52:03 <yanyanhu> #topic Open discussions
13:52:12 <yanyanhu> any more topics from you guys?
13:52:30 <yanyanhu> Qiming, xinhui is still in travel?
13:52:37 <Qiming> not sure
13:52:43 <Qiming> one question ...
13:52:47 <yanyanhu> looking forward to the news about ptg from her :)
13:53:01 <Qiming> anyone is interested in trying deploy a kubernetes cluster using senlin?
13:53:27 <ruijie> I am working on deploy a cloud foundry cluster ..
13:53:40 <Qiming> wow ...
13:53:56 <ruijie> no very familiar with k8s
13:54:00 <ruijie> not
13:54:40 <Qiming> it would be great to learn any experiences / lessons from you
13:55:07 <ruijie> very happy to share it after I finish it :)
13:55:22 <yanyanhu> ruijie, thanks :)
13:55:35 <XueFengLiu> good boy:)
13:55:44 <yanyanhu> :P
13:55:48 <Qiming> don't hesitate if there are things we can help from senlin side
13:55:51 <ruijie> lol ..
13:56:18 <ruijie> the role of node will be a big problem ..
13:56:28 <Qiming> ha
13:56:34 <ruijie> since the recover and scaling are both need to based on that
13:57:02 <Qiming> should we extend 'role' to be a list?
13:57:28 <Qiming> many other apis should be made role aware, I guess
13:57:52 <yanyanhu> enhancing role support is critical for many cases I believe
13:57:58 <yanyanhu> so may need more attention on it
13:58:18 <ruijie> agreed
13:58:50 <yanyanhu> so lets think more about it :)
13:58:55 <yanyanhu> ok, that's all for today I guess
13:59:21 <yanyanhu> thanks all you guys for joining, have a good night :)
13:59:36 <Qiming> bye
13:59:38 <XueFengLiu> Hi,YanYan. About receiver will ask you tomorrow
13:59:39 <yanyanhu> bye
13:59:41 <XueFengLiu> Bye
13:59:44 <yanyanhu> XueFengLiu, sure
13:59:45 <ruijie> have a good night
13:59:46 <yanyanhu> please ping me
13:59:52 <XueFengLiu> OK
13:59:55 <XueFengLiu> good night
14:00:03 <yanyanhu> goog night
14:00:08 <yanyanhu> oops...
14:00:10 <yanyanhu> good night :)
14:00:11 <yanyanhu> #endmeeting