01:00:56 #startmeeting zaqar 01:00:57 Meeting started Tue Oct 17 01:00:56 2017 UTC and is due to finish in 60 minutes. The chair is flwang. Information about MeetBot at http://wiki.debian.org/MeetBot. 01:00:58 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 01:01:01 The meeting name has been set to 'zaqar' 01:01:04 #topic roll call 01:01:08 anybody around? 01:01:52 gecong_: gengchc_: around? 01:01:56 yep 01:02:00 cool 01:02:21 yes 01:02:25 gecong_: firstly, congrats for joining the core team and thanks for all your contribution in the past days 01:02:33 gengchc_: thanks for joining 01:02:47 you are welcome 01:02:59 thanks 01:03:26 generally, we use this etherpad to track the agenda of zaqar meeting but we didn't use it much recently, see https://etherpad.openstack.org/p/zaqar-meeting-agenda 01:03:49 this week is the queens milestone 1 release 01:04:10 so today i'd like to revisit the plans/blueprints for queeens 01:04:20 to make sure if we're on track 01:04:20 ok 01:05:16 ok 01:05:37 https://etherpad.openstack.org/p/zaqar-queens-plan 01:06:00 this is the link we're using to track the tasks we'd like to do in Queens 01:07:06 do you have any topic to discuss today? 01:07:32 Support more backoff functions in Zaqar -- in work 01:07:54 Bascially pool group is useless in Zaqar, it is time to remove it -- in work bu gengchc 01:07:57 gecong_: yep, is it ready for review? 01:08:20 not yet, in a few days 01:08:49 ok, pls let us know when it's ready for review 01:08:56 Support Redis as mgmt storage backend -- in work bu gengch and gecong 01:09:18 yes,I work on pool group function, code is almostly completed, I work on unittest. 01:09:19 np , thanks 01:10:02 Support Token Session in Zaqar for improving performance -- I will contact with wanghao today 01:10:26 redis as the management of storage function, the code is almostly completed, I and gecong work on unittest. 01:10:27 I added comments on the redis mgmt spec https://review.openstack.org/#/c/504379/4/specs/queens/support-redis-as-mgmt-storage-backend.rst if you can address those comments, I'm happy to approve it 01:10:55 ok 01:10:58 ok 01:11:44 gecong_: I'm really interested in the spec Support Token Session in Zaqar for improving performance 01:12:11 ok , I will discuss with wanghao for this today 01:12:29 so you guys are working on 3 blueprints now, redis, pool group and retry policy, that's good enough for this cycle 01:12:47 and I know Lance, keystone ptl is working on the policy in code 01:12:54 yes 01:12:55 policy in code is inited by lance 01:13:00 yeah 01:13:01 and chandankumar is working on the tempest plugin 01:13:11 It's in good shape. 01:13:30 wxy_: yep, i saw you're reviewing it, thanks 01:13:57 generally, given the size of our team, I think we have enough things to do 01:13:59 Anyone who is interested in can push the follow patches. 01:14:07 wxy_: which one? 01:14:13 policy in code 01:15:20 Lance only complete queue and message refactor. 01:15:58 do we have to finish all of them in one cycle? 01:16:35 I think we should, it's not very hard. 01:17:32 And this one https://review.openstack.org/#/c/511749 . Need your review comment. 01:18:38 [https://review.openstack.org/#/c/511749] can't open. 01:19:00 seems gerrit is down? 01:19:18 anyway, I will take a look later, if it's necessary, i can pick it up 01:19:20 yes 01:19:37 i was busy recently, but should be available in next couple of weeks 01:20:12 prepare for summit? 01:20:28 I will review it today 01:20:38 wxy_: vote for TC and internal Horizon migration/performance 01:20:50 s/vote/run 01:20:56 I will review it too. 01:21:05 thanks, team. 01:21:08 I have voted for TC already 01:21:12 flwang: cool. 01:21:18 wxy_: any idea for this Support Token Session in Zaqar for improving performance ? 01:21:35 I have voted for TC too. 01:22:39 thanks for all the comments/updates. i think we're basically on track for the plans on Queens 01:23:08 flwang: No, It needs more detail I think. 01:23:19 wxy_: ok, no problem 01:25:03 gengchc_: gecong_any updates from the your production? 01:25:40 i mean zaqar on ZTE developer platform 01:26:39 * wxy_ leave for some internal work. 01:26:52 yesterday, I attend a meeting about openfog messageing 01:27:46 any info can be shared? 01:27:51 * wxy_ is coming 01:29:17 it is about edge computer 01:30:13 gecong_: ok 01:30:34 they need messgae solution 01:30:55 as you know, the adoption of zaqar is still low, so please feel free recommend Zaqar wherever you can :) 01:31:11 I will discuss with them in following days 01:31:45 yes, I will introduce zaqar to them 01:31:48 if it's not a private mailing thread, you can involve me 01:32:09 sure 01:32:27 I am glad to hear that 01:32:51 cool, thanks 01:33:25 and one thing you could mention to them is, zaqar can work without the other openstack services 01:33:49 the only work they may need is a plugin for auth 01:34:03 they need most is communction between DCs 01:34:24 ok, we can discuss the details offline 01:34:34 ok 01:34:37 #topic bugs 01:35:01 Luckily(unfortunately), now we don't have much bugs on backlog 01:35:11 but I expect bugs in our code 01:36:04 I think if we commit our bp, there will be a lot of bugs 01:36:14 as you may know, zhenyuyang from YiYun is working on deploying zaqar 01:36:19 :) 01:36:34 so I may work with him to figure out any potential issue and get them fixed 01:36:59 that's the plan 01:37:05 #topic open discussion 01:37:06 ok 01:37:23 is there anything else you want to discuss? 01:37:44 not from me 01:37:57 Please register the bugs if possible. 01:38:05 wxy_: absolutely 01:38:08 I have a question: Does anyone know what's the current situation about the CI/CD? I found both Jenkins and Zuul. What's the relationship? 01:38:48 I think they will immigrate to zuul V3 01:38:49 actually, Zaqar is also on Catalyst Cloud roadmap for a while, but we have to deploy LBaaS and Magnum firstly due to the low requirements in New Zealand for messaging :( 01:39:35 CI/CD is not stable recently 01:39:53 wxy_: i didn't dig into the details of the CI, but I assume the Zuul v3 will replace Jenkins job? 01:40:05 gecong_: Which one is the primary? Now When I do recheck, it seems that only jenkins goes, but zuul does nothing. 01:40:39 and py27 is always fail 01:40:47 flwang: So do we need to do some work in project-config? 01:41:19 wxy_: i assume no, but will see 01:41:26 flwang: OK. 01:41:48 I don't know , I think they are debugging the zuul V3 01:42:19 Ok, That's all for me. THanks. 01:42:36 BTW, another thing in my mind randomly is, is it possible to rewrite zaqar with Go? 01:42:52 or at least some parts of Zaqar to improve the performance 01:43:02 but the idea is very unclear now 01:44:03 flwang: what's the progress that Go in Swift? I think we can learn something from it. 01:44:29 If We can improve performance, can you consider using go or c language? 01:44:39 wxy_: the problem is zaqar is only one process 01:44:48 so it's a little bit hard to rewrite part of it 01:45:09 Emm, yeah. 01:45:10 gengchc_: that's my thought, but I'm not sure if language is really the bottleneck, TBH 01:45:25 maybe just the websocket part> 01:45:26 ? 01:45:51 i don't know, just random thinking 01:47:29 I use two nodes to test zaqar based on redis+mongo+haproxy, almost 2000 messages /seconds,but there is big different from tecent(120000/s) 01:48:01 because the status of Trove makes me thinking, we may need to do some changes to get a big improvement 01:49:09 不破不立 01:49:20 lol 01:49:25 support 01:49:34 I agree. 01:50:01 wxy_: my long term goal, just like Gnocchi, actually Zaqar could be an open source messaging project out of OpenStack world 01:50:22 Zaqar standalone. 01:50:35 All projects now want to standalone. :) 01:50:43 so my goal is creating an open source messaging service with HA, low latency and good scalability 01:50:55 wxy_: no, I say we can 01:51:00 but we don't have to 01:51:09 in a visible future 01:51:22 but I mean we need to have a big goal 01:51:39 yes, sir. 01:51:45 yes sir 01:52:09 gengchc_: I'm interested in the performance you mentioned " from tecent(120000/s)" 01:53:07 I see it in tecent website. 01:53:07 I will create a separated etherpad for performance tunning 01:53:43 and we can collect the performance from different public clouds and work out a reasonable goal 01:53:47 then focus on that 01:53:57 ok 01:54:02 ok 01:55:02 ok, i think that's us 01:55:17 cent website: https://cloud.tencent.com/product/cmq 01:55:18 another thing: FYI, Keystone has a bp called unified limits. It's related to Quota. Now Lance and I are trying to take over it. Zaqar can wait for it for the quota requirement IMO. 01:55:29 anything else? we can continue discussion in zaqar channel 01:55:47 tecent website: https://cloud.tencent.com/product/cmq 01:55:47 wxy_: yep, good to know. 01:56:00 gengchc_: thanks, i will take a look 01:56:07 ok, i'm going to close the meeting 01:56:10 thanks for joining 01:56:12 ok 01:56:16 #endmeeting