13:00:12 <yanyanhu> #startmeeting senlin 13:00:12 <openstack> Meeting started Tue Dec 20 13:00:12 2016 UTC and is due to finish in 60 minutes. The chair is yanyanhu. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:16 <openstack> The meeting name has been set to 'senlin' 13:00:31 <yanyanhu> hello 13:00:34 <Qiming> hi 13:01:13 <yanyanhu> hi, Qiming 13:01:13 <yanyanhu> hi, elynn 13:01:22 <elynn> hi yanyanhu 13:01:36 <zhouzhengxi> Hi yanyanhu. 13:01:42 <yanyanhu> hi, zhouzhengxi, you're using your full name :) 13:01:49 <zhouzhengxi> Hi elynn. 13:01:51 <yanyanhu> hi, Ruijie_ 13:01:58 <Ruijie_> hi, yanyanhu 13:02:14 <yanyanhu> lets wait for a while for other attenders 13:02:20 <yanyanhu> and here is the agenda: https://wiki.openstack.org/wiki/Meetings/SenlinAgenda#Agenda_.282016-12-20_1300_UTC.29 13:02:28 <yanyanhu> please feel free to add items 13:02:56 <yanyanhu> ok, lets start 13:03:01 <yanyanhu> https://etherpad.openstack.org/p/senlin-ocata-workitems 13:03:07 <yanyanhu> #topic ocata workitems 13:03:18 <yanyanhu> Testing 13:03:30 <yanyanhu> haven't work on it 13:03:47 <yanyanhu> and about rally plugin, Ruijie_ has added the support for cluster scaling 13:04:35 <yanyanhu> next one 13:04:40 <yanyanhu> health management 13:04:58 <yanyanhu> xinhui is not here I guess 13:05:27 <yanyanhu> noticed octavia team have left more comments on xinhui's patch 13:05:29 <Qiming> one of the workitems is about adding mistral support 13:05:39 <yanyanhu> ah, right 13:05:46 <yanyanhu> Qiming is now working on it 13:05:47 <Qiming> yup, those are all reasonable comments 13:05:55 <yanyanhu> about supporting mistral workflow in senlin 13:06:37 <yanyanhu> the purpose is leveraging mistral's ability to drive the workflow of failure recovery 13:06:43 <zzxwill> Changed to my gerrit username. Thanks, yanyanhu. 13:06:44 <yanyanhu> Qiming, is that right? 13:06:47 <yanyanhu> :) 13:06:56 <yanyanhu> zzxwill, :) 13:07:01 <Qiming> yes, sort of an amendament 13:07:20 <Qiming> there are always work we cannot model, for recovery 13:07:33 <Qiming> we can leave them as user provided hooks 13:07:43 <yanyanhu> yep 13:07:56 <Qiming> users can hook their own workflow, workbook to the recovery process 13:08:10 <yanyanhu> for users are the only people can decide how to make the recovery in some cases 13:08:22 <Qiming> this is the patch: https://review.openstack.org/#/c/412389/ 13:08:52 <yanyanhu> yea, the API version problem 13:09:05 <Qiming> once mistral is returning a "discoverable" version, we can identify it in openstacksdk 13:09:13 <Qiming> then we add resources into sdk 13:09:23 <yanyanhu> great 13:09:29 <Qiming> and finally, we know how to retrieve and trigger a workflow 13:10:05 <yanyanhu> how about the current mistral support in sdk? 13:10:15 <Qiming> nothing yet 13:10:19 <yanyanhu> need big effort to complete it? 13:10:22 <yanyanhu> I see... 13:10:35 <Qiming> only a base committed: https://review.openstack.org/#/c/412079/ 13:11:19 <yanyanhu> maybe we can start from those basic and necessary ones 13:11:41 <Qiming> it won't be big I think 13:11:52 <yanyanhu> ok 13:12:03 <Qiming> for example, I have completed the whole bare-metal support in about one day: https://review.openstack.org/#/q/project:openstack/python-openstacksdk+message:bare-metal 13:12:28 <yanyanhu> WoW 13:12:31 <Qiming> checked mistral resource types, it is not big 13:12:51 <yanyanhu> Qiming, you're a real superman :P 13:13:11 <yanyanhu> thanks a lot for the effort 13:13:15 <Qiming> one thing I'm not quite sure is whether the API doc for mistral is latest: http://docs.openstack.org/developer/mistral/developer/webapi/v2.html 13:13:29 <Qiming> need to double confirm 13:14:08 <yanyanhu> yes... those API docs are almost the only reference for us... 13:14:18 <yanyanhu> besides the source code 13:14:36 <Qiming> em, maybe I was too optimistic, those resources do have a lot of attributes ... 13:14:37 <yanyanhu> anyway, hope there won't be much difficult 13:15:06 <yanyanhu> hope their API document is accurate :) 13:15:20 <Qiming> me too 13:15:37 <yanyanhu> oh, I have added this sub workitem to HA section 13:15:51 <yanyanhu> for further tracking 13:16:09 <yanyanhu> and will contact xinhui for that patch to octavia 13:16:22 <lxinhui> hi, yanyanhu 13:16:34 <yanyanhu> hi, lxinhui :) 13:16:34 <lxinhui> have micheal accepted the BP? 13:16:37 <yanyanhu> nice to see you 13:16:42 <yanyanhu> lxinhui, not yet 13:16:47 <lxinhui> sorry for the late 13:16:50 <yanyanhu> and their left more comments on that patch 13:16:55 <yanyanhu> lxinhui, no problem 13:17:07 <lxinhui> last week, I traveled to Shenzhen for Open-O meetup 13:17:13 <yanyanhu> I guess they still have some questions about that change 13:17:22 <lxinhui> and saw many old friends there 13:17:26 <yanyanhu> for NFV :) 13:17:36 <yanyanhu> nice 13:17:36 <lxinhui> that does not make sense 13:17:54 <lxinhui> since BP represents the requriements only 13:18:04 <lxinhui> which should be speparated with implement 13:18:20 <lxinhui> if he has questions about the change 13:18:39 <yanyanhu> link to the patch: https://review.openstack.org/#/c/402296/ 13:18:39 <lxinhui> why not accpet the BP then leave comments on the change 13:19:18 <yanyanhu> lxinhui, I guess they may want to understand more detail about the change 13:19:32 <yanyanhu> anyway, BP is the basement 13:19:49 <lxinhui> is German one of the core members? 13:20:03 <yanyanhu> lxinhui, not sure 13:20:19 <lxinhui> that is what I am discussing 13:20:28 <Qiming> their suggestions to make notifications configurable is reasonable 13:20:47 <lxinhui> no matter reasonable or not 13:20:50 <yanyanhu> lxinhui, maybe you can leave some reply on the patch and ask for them for further review on that bp 13:20:54 <lxinhui> micheal need approve the BP firstly 13:21:07 <lxinhui> as his own comment 13:21:16 <lxinhui> leave it till the BP approved 13:21:16 <yanyanhu> have got response from michael and he said they will spend more time to understand it 13:21:34 <lxinhui> no need to understand the chnage 13:21:58 <lxinhui> just echo back to us if he accept the requrirment or not 13:21:59 <lxinhui> ... 13:22:31 <lxinhui> or lese, I update new patch means nothting 13:24:05 <yanyan> sorry, just dropped 13:24:43 <Qiming> yes, need some clarification on the review 13:24:43 <lxinhui> All of us should go to Alanta 13:24:49 <lxinhui> for the middle cycle 13:25:02 <lxinhui> and f2f coding there :) 13:25:10 <yanyan> lxinhui, that is diffult for us :( although we really hope so 13:25:11 <Qiming> or else we will get blocked 13:25:29 <yanyan> difficult 13:25:43 * Qiming not a strong believer of f2f coding 13:26:03 <yanyan> anyway, having more discussion on that BP is urgent 13:26:13 <yanyan> we have only 1 month left for Ocata 13:26:18 <yanyan> less than 1 month 13:26:18 <lxinhui> Yes 13:26:38 <Qiming> just leave a comment on that patch 13:26:38 <lxinhui> another thing I am trying to get time to do is integrate mistral with Senlin 13:26:59 <Qiming> whether we should wait for the BP to be approved before submitting another patchset 13:26:59 <yanyan> lxinhui, Qiming is working on it too :) 13:27:06 <lxinhui> ha 13:27:12 <lxinhui> I will work with him toegtehr 13:27:18 <lxinhui> or qiming can handle it 13:27:26 <lxinhui> I will do something else 13:27:46 <lxinhui> since he is much more familar about SDK 13:28:00 <yanyan> I think you two can have some discussion to see whether there are some gaps between what we need and what they have 13:28:26 <lxinhui> yes 13:28:26 <yanyan> Qiming is now working on sdk support and also need some fixes on mistral itself 13:28:35 <lxinhui> yes 13:28:38 <yanyan> lxinhui, maybe you can consider more from senlin's pespective 13:28:45 <lxinhui> we have has a short discussion in the weekend 13:29:09 <lxinhui> ha 13:29:13 <yanyan> yes? 13:29:22 <Qiming> https://review.openstack.org/#/c/412389/ 13:29:26 <lxinhui> about what you just mentined 13:29:29 <Qiming> patch for fixing the version problem 13:30:10 <lxinhui> coo 13:30:12 <lxinhui> l 13:30:30 <lxinhui> Then I can learn that in the vacation 13:30:43 <lxinhui> and see what I can help there 13:30:47 <yanyan> lxinhui, btw, German is not the core reviewer of octavia, but he does contribute a lot to it 13:31:14 <yanyan> so he should be very familiar with it 13:31:46 <Qiming> I'd suggest we focus on getting things landed, instead of the people behind it 13:32:14 <yanyan> Qiming, agree totally 13:32:22 <yanyan> so maybe just reply to the questions 13:32:38 <yanyan> and have more discussion on the proposal 13:32:49 <yanyan> try to involve related people 13:33:02 <yanyan> and see their feedback 13:33:21 <yanyan> if we do all those things but can get it, that won't be our fault 13:33:30 <yanyan> s/can/can't 13:33:41 <Qiming> leave a comment there 13:33:47 <Qiming> asking for suggestions 13:34:00 <Qiming> should we wait for bp approval before submitting another patchset? 13:34:45 <yanyan> thanks Qiming 13:35:02 <lxinhui> Qiming, That is what I mean by help 13:35:10 <lxinhui> you are the god can do everything 13:35:13 <yanyan> we can add octavia's core reviewers to reviewer list 13:35:20 <lxinhui> I will see anything can do to help landing 13:35:54 <lxinhui> What comments 13:35:55 <yanyan> great, thanks a lot 13:35:57 <lxinhui> about 13:35:59 <Qiming> cool 13:36:03 <lxinhui> octavia? 13:36:08 <Qiming> yes 13:36:08 <lxinhui> where is your comments 13:36:10 <lxinhui> or 13:36:12 <lxinhui> mistral 13:36:21 <Qiming> "whether we should wait for the BP to be approved before submitting another patchset" 13:37:02 <lxinhui> can not see it 13:37:07 <lxinhui> maybe metwork problem 13:37:18 <yanyan> could be gerrit's problem 13:37:19 <Qiming> never mind 13:37:33 <yanyan> anyway, hope there can be more progress 13:37:43 <yanyan> ok, so lets move on? 13:37:51 <lxinhui> yes, please 13:37:59 <yanyan> document, no progress? 13:38:10 <yanyan> versioned request 13:38:28 <yanyan> support for credential_get, get_revision are still unfinished 13:38:32 <yanyan> will work on them soon 13:38:46 <yanyan> all others have been done I think 13:39:03 <yanyan> container profile 13:39:14 <yanyan> Qiming's patches have been merged 13:39:29 <yanyan> two patches to support dependency between node and profile 13:39:51 <yanyan> event/notification 13:39:54 <Qiming> there are some more patches about validation 13:40:02 <yanyan> yes 13:40:03 <Qiming> blocked now by gate problem 13:40:13 <yanyan> yes, have a topic about this 13:40:21 <yanyan> really a headache 13:40:40 <yanyan> ok, lets finish the discussion on workitems first 13:40:49 <yanyan> configuration options for event/notification have been added 13:41:01 <yanyan> so all basic support has been done I guess? Qiming 13:41:07 <yanyan> for event/notification 13:41:31 <Qiming> yes 13:41:37 <Qiming> need to add some documents 13:42:17 <yanyan> Qiming, great 13:42:59 <yanyan> ok, last one 13:43:07 <yanyan> NFV support/Baremetal cluster 13:44:03 <yanyan> Qiming, do we need further sync with tosca team? 13:44:29 <Qiming> yes, we'd better do that 13:45:06 <yanyan> Qiming, ok, since the bp for tacker is almost ready(I think), maybe need talk with tosca team for some detail 13:45:25 <lxinhui> I think there already have been a disucssion in their weekly meeting 13:45:26 <yanyan> to support translation between senlin resource and tosca resource 13:45:28 <lxinhui> right? 13:45:36 <lxinhui> how about that time 13:46:10 <Qiming> yes, ideally, tosca templates can be translated to senlin profile/policy specs directly 13:46:16 <yanyan> yes 13:46:48 <Qiming> but this will become useless if tacker insists that they will orchestrate things using heat 13:47:01 <lxinhui> ... 13:47:20 <yanyan> Qiming, if they insist, it'ok I think 13:47:25 <Qiming> yes 13:47:36 <Qiming> but some day, they will realize ... 13:47:37 <lxinhui> why they still insist on the integration 13:47:40 <lxinhui> with heat 13:47:45 <lxinhui> instead of senlin 13:47:47 <yanyan> just they wont get benefit from some useful features senlin provide 13:47:54 <Qiming> it gives them a single dependency 13:48:04 <Qiming> it makes things a lot simpler 13:48:20 <yanyan> yes 13:48:22 <lxinhui> all the functions they need can be done by Senlin? 13:48:34 <yanyan> that's why they reply on heat to deploy everything 13:48:50 <Qiming> cluster management can and should be done via senlin 13:49:15 <yanyan> Qiming, yes, it is 13:49:26 <yanyan> just senlin cluster could be created through heat as well 13:49:29 <yanyan> as a heat resource 13:49:31 <lxinhui> what other functions make them dependent on heat> 13:49:37 <lxinhui> ? 13:49:42 <lxinhui> besides this 13:49:43 <Qiming> combination of things 13:49:43 <yanyan> lxinhui, network, storage, etc. 13:50:06 <lxinhui> ... 13:50:14 <lxinhui> once we discussed this 13:50:33 <lxinhui> right when Qiming mention the resource pool management 13:50:37 <Qiming> they don't get my metaphor: heat provides you a struct, while senlin provides you an array 13:50:49 <lxinhui> they do not need maybe 13:50:56 <lxinhui> or identify the importance 13:50:59 <Qiming> they will first try model everything as a struct, that is fine 13:51:29 <Qiming> just a little worried about their future changes, .... maybe I'm worrying too much 13:51:41 <lxinhui> could you give me an example of "model everything as a struc" 13:51:49 <yanyan> Qiming, anyway. senlin will be the backend service to manage cluster resource 13:51:50 <Qiming> .... 13:51:58 <yanyan> that will be helpful for them 13:52:06 <Qiming> every resource combined into a template, deployed as a stack 13:52:10 <lxinhui> ... 13:52:15 <Qiming> that is a struct, lxinhui, right? 13:52:15 <lxinhui> ..... 13:52:28 <lxinhui> anyway 13:52:55 <lxinhui> I agree to extend Senlin for more network/storage 13:52:57 <lxinhui> functions 13:52:58 <Qiming> operating a cluster thru heat is doable, but you are losing all the operations we have worked hard to provide 13:53:06 <lxinhui> as a real resource manager 13:53:19 <Qiming> actually we can 13:53:27 <lxinhui> but not today 13:53:42 <Qiming> it is not about things we CAN do, it is really about things we SHOULD do or NOT 13:54:05 <yanyan> Qiming, yes, that's the point 13:54:06 <lxinhui> then you can not claim senlin as a resource manager 13:54:15 <Qiming> e.g. we can invoke libvirt directly, bypassing nova 13:54:17 <lxinhui> as a resource manager 13:54:28 <lxinhui> you need consider different kinds of reosurces 13:54:33 <Qiming> so that we can manage physical hosts, and do VM placement 13:54:35 <Qiming> :D 13:54:44 <lxinhui> you need to think how to control them 13:54:57 <lxinhui> reserve 13:55:04 <lxinhui> and so on 13:55:26 <lxinhui> or it just a toy 13:55:28 <Qiming> so far I still believe we should focus on getting cluster management done 13:55:34 <lxinhui> to let audience to know 13:55:46 <lxinhui> that is good to use it for clusteting management 13:55:55 <lxinhui> since 13:56:08 <yanyan> kinda agree with Qiming, lets just well play our role as clusering service in current stage :) 13:56:13 <lxinhui> clustering should not be only a group of somethin,, 13:56:30 <Qiming> thinking from users perspective 13:56:32 <lxinhui> but also managemet functions related to it 13:56:41 <lxinhui> I am a user 13:56:42 <Qiming> we can somehow extend the nova server profile 13:56:42 <lxinhui> :) 13:57:03 <Qiming> to add more useful features regarding storage and network 13:57:20 <lxinhui> yes 13:57:35 <lxinhui> reason why every company want hire nove dev 13:57:46 <yanyan> ok, 2 minutes left, guys :) 13:57:46 <Qiming> we don't have to get ourselves limited/constrained by the abstractions provided by a single service (e.g. nova) 13:58:01 <lxinhui> because it is indepensible 13:58:19 <Qiming> if users need a nova server with a floating IP 13:58:27 <lxinhui> I am trying to think the way to make senlin indepensible 13:58:33 <Qiming> we should add add a floating IP property to nova server profile 13:58:46 <lxinhui> instead of some toolkit nice to have 13:59:09 <Qiming> ... openstack is not indispensible ... 13:59:13 <lxinhui> more than that 13:59:26 <lxinhui> precondition is openstack is alreay chosed 13:59:31 <lxinhui> then senlin should be chosed 13:59:38 <Qiming> I agree 13:59:39 <elynn> Maybe at least make it indispensable at NFV world 13:59:39 <Qiming> :D 13:59:39 <lxinhui> that is what I mean 13:59:49 <lxinhui> agree elynn 13:59:56 <yanyan> hi, since time is almost over, lets move back to senlin channel for further discussion? 14:00:24 <yanyan> thanks all you guys for joining, lets move back the senlin channel 14:00:31 <yanyan> #endmeeting