13:00:46 #startmeeting tricircle 13:00:47 Meeting started Wed Nov 9 13:00:46 2016 UTC and is due to finish in 60 minutes. The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:48 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:50 The meeting name has been set to 'tricircle' 13:01:18 #topic rollcall 13:01:24 #info joehuang 13:01:34 #info Yipei 13:01:43 #info xiulin 13:01:49 #info jiawei 13:02:15 #info liukun 13:03:24 #topic tricircle big tent application info share 13:03:48 \ 13:04:19 the application was discussed in TC weekly meeting 13:04:31 and TCs in the meeting voted 13:04:32 #info Junsik 13:04:47 the vote result could be found here: https://review.openstack.org/#/c/338796/ 13:04:49 #info zhiyuan 13:05:15 some TCs are absent, so not all TCs finished vote yet 13:05:32 need to wait one more week for the conclusion 13:06:12 to Junsik, what's your mail address in OpenStack review system? 13:06:49 you can add yourself to the patch for review: https://review.openstack.org/#/c/394372/ 13:07:22 o, Junsik's session just quit 13:07:38 welcome back Junsik 13:07:58 could you add yourself to the patch https://review.openstack.org/#/c/394372/ as reviewer? 13:08:36 Sure, I'll do. 13:09:03 so that other guys know your handle and can add you as reviewer for his/her patch 13:09:25 come back to the big tent application 13:09:39 most of TCs voted +1 13:10:01 and Armando's comment is very valueable for our design improvement 13:11:03 The local neutron plugin inherits from the ML2 plugin whereas the central neutron plugin inherits from a base class usually known as monolithic base. IMO, both design decisions may lead to a fragile neutron solution (in other words, neutron code changes may break tricircle all the time), in that neither are 'blessed' pluggable points. ML2 drivers and service plugins are, but nothing else. 13:11:09 I did look at the code a bit and I must admit I am not sure what to think. Some decisions look like are taken without any feedback from the neutron team. I personally find the concept of 'real' core plugin a bit troubling [1]. This means two core plugins are running at any given time in each local neutron deployment with no care of transactionality (what happens in partial failure scenarios?). 13:11:15 Network services composability does not seem to be taken into account. What happens to L3+ services? All of this is unclear 13:12:12 what's your comment about the local plugin and central plugin inherit point? 13:13:54 Can the central plugin also be inherited from ML2 plugin but not the base class? 13:15:43 if inherit from ML2 then we add mechanism drivers to central Neutron server which is not needed 13:16:39 yes, unless we override the MD driver loading 13:17:57 so it's weird to inherit from ML2 while remove the mechanism drivers 13:18:30 how about if we inherit from ML2 plugin, but override all interfaces but just keep the interface same as ML2 plugin 13:19:49 this is a big question, don't have to give an answer today, the change is too big 13:21:04 and other thoughts on Armando's comment? 13:21:10 then what's the gain? just to let both local and central inherit from the same ML2 plugin? 13:22:11 from currently point of view, no big gain I think 13:22:17 composability needs further validation if the real_core_plugin is not ML2 13:23:03 seems most of SDN controllers now provide ML2 MD driver 13:24:04 i see. actually the reason local plugin inherits from ML2 is that we would like local plugin to handle RPC 13:24:46 so if other core plugin has its own RPC, local plugin may fail to handle RPC 13:25:02 the most important one is the point being able to insert UUID from the central Neutron 13:25:42 that's right, if Neutron API supports creating resource with ID, we don't need a local plugin 13:26:48 Am I correct? 13:26:56 how to notify the central Neutron if no local plugin 13:26:57 or I miss the port creation issue 13:27:30 no, even supports the UUID insertion through API, local plugin is still needed 13:27:37 yes 13:27:48 you missed the port creation issue 13:28:38 currently our implementation may be still the best solution 13:29:40 ok, let's think over Amando's comment offline, and next topic 13:29:50 #topic Ocata feature development discussion 13:30:21 could you share your recent progress in ocata feature development 13:30:25 well, since port creation is done before mechanism driver process, a MD is also not enought to handle port creation 13:31:02 to zhiyuan, agree 13:31:24 I am working on the bridge network issue, to support DVR, it's better to attach brige network just as external network 13:32:17 then the cross-pod traffic will first go to the central SNAT namespace, then go out of the pod 13:32:43 which reduce the VTEP and tunnels Tricircle needs to handle 13:33:07 I investigated multi-region gate/check test environment, no multi-region environment supported for tempest test in OpenStack yet 13:33:25 sorry for late 13:33:30 could you submit a spec for this, zhiyuan 13:33:34 welcome ronghui 13:33:47 yes, i am writting the spec 13:34:01 we are in the topic "Ocata feature development discussion" 13:34:32 okok 13:34:48 to Junsik, is the installation according to the guide successful? 13:35:53 Yes I succeed to install tricircle by following the guide. 13:36:08 mutli-node? 13:36:25 Yes multi nod2 13:36:35 great news, that means the guide can work now 13:37:18 would you like to contribute on configuration guide or user guide? 13:38:28 or someone is interested in the contribution in configuration guide or user guide? 13:38:49 Is there something we need to fix the configuration guide? 13:39:15 Yes I'll try to work on the guides 13:39:47 need to create a totally new configuration guide and user guide, currently we have no these two guide 13:40:23 only installation guide is there, and only for devstack based installation 13:40:26 i just install the tricircle right now 13:40:44 to Ronghui, does it work? 13:41:27 the devstack still running 13:42:14 and i think we need a guide to install all the tricircle (tricircle and trio2o) 13:42:40 these two are seperated, no need a guide for two, 13:42:44 :) 13:43:07 tricircle and trio2o are two seperated projects 13:43:12 ok 13:43:15 i got it 13:43:40 and what you mean for the totally new configuration guide and user guide 13:43:42 have you successfully install two nodes tricircle? 13:43:50 no 13:43:55 just singlenode 13:44:23 we don't have these two guides now, and need to writhe the guide 13:44:26 what you mean for the totally new configuration guide and user guide wthout devstack? 13:45:34 the guide should work even in non-devstack based openstack 13:45:56 great 13:46:55 i just carry out an openstack for Newton and i think i can help for this guide 13:47:14 To Junsik, Ronghui, which guide you would like to contribute 13:47:38 user guide 13:47:42 and one more contribution point is to remove the quota related files in tricircle 13:48:06 me too. user guide. 13:48:50 no need quota management in tricircle 13:49:17 so I hope someone can submit a patch to remove quota and it's unit test files 13:49:32 i will try it 13:49:44 To Junsik/ronghui, thank you 13:50:31 and for another configuration guide, i still a little confusion 13:50:58 could you give me some examples about the totally new configuration guide 13:51:18 yes 13:52:15 i install the devstack and openstack many times, so i think i can help to complement this configuration guide 13:52:33 this is the neutron configuration guide:http://docs.openstack.org/newton/config-reference/networking/networking_options_reference.html 13:53:53 I am looking for user guide 13:54:11 hi joehuang 13:55:07 hello, junfeng 13:55:15 seems first time to meet you here 13:55:25 Are you Tricircle project's core member? 13:55:30 yes 13:55:33 ok 13:55:54 could you introduce yourself briefly 13:56:04 yes 13:56:23 I'm from beijing china 13:56:52 I am a openstack developer 13:57:10 great! which project you mostly worked in 13:57:47 are you interested in to contribute in Tricircle 13:58:09 Recently I was studying your project 13:58:57 welcome to join 13:59:31 thanks 13:59:47 time is up, could you send mail in the maillist with [tricircle] in the subject 14:00:22 have to end meeting, thank you for your attending, will discuss in m-l 14:00:29 #endmeeting