03:01:59 #startmeeting openstack-cyborg 03:02:00 Meeting started Thu Oct 31 03:01:59 2019 UTC and is due to finish in 60 minutes. The chair is Sundar. Information about MeetBot at http://wiki.debian.org/MeetBot. 03:02:01 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 03:02:03 The meeting name has been set to 'openstack_cyborg' 03:02:08 #topic Who's here 03:02:13 #info Sundar 03:02:19 #info s_shogo 03:02:36 #info chenke 03:03:21 #info Yumeng 03:05:28 Hi all 03:05:46 Let's wait for 1 more minute for others. 03:06:00 Ok. 03:06:04 Agenda: https://wiki.openstack.org/wiki/Meetings/CyborgTeamMeeting#Agenda 03:07:49 We have waited long enough. Let's get started 03:08:02 #topic PTG planning 03:08:21 Started some tentative schedule at https://etherpad.openstack.org/p/shanghai-ptg-cyborg 03:09:01 This is only my suggestion: you all can suggest more times for specific topics. Also, I am not done yet :) 03:09:11 I'll try to complete after this meeting 03:10:08 Re. team dinner, there has been a suggestion to move it to Tuesday (Nov 5) eve. Is everybody ok with that? 03:10:50 I am ok for Nov.5 Nov.6 Nov.7 03:11:28 ok with that. ^ ^ 03:11:45 me too, no problem > team dinner. 03:11:59 metoo. 03:12:00 Excellent. Let's move it up then. 03:13:09 There is some confusion about project onboarding. When I asked about that to OpenInfra Foundation staff, I got an answer about project updates. 03:13:58 But I believe it is part of the PTG this time. So, we have to schedule it and get people to attend it 03:14:46 Since Wednesday is part of the summit, not sure if we should hold PTG discussions on that day. What do you all think? 03:16:46 I am not sure if there is time on the day. 03:17:00 time on Wednesday? 03:17:08 yes. 03:18:08 Well, we have a room reserved for Wed too: http://ptg.openstack.org/ptg.html If we are all busy with the Summit on that day, let's not schedule anything on that day then. Any objections? 03:18:54 OK, I got it. 03:19:05 Agree. 03:19:24 +1 03:19:30 Good. That means all PTG topics and project onboarding will fall on Thu/Fri. 03:20:02 Got it. 03:20:35 Shall we schedule project onboarding for Fri morning? Ideally, we'll get lots of candidates, an d hopefully some of you can present Cyborg and answer questions. I am not good with Mandarin. :) 03:22:13 I can help with that. 03:22:22 Cool. 03:22:30 Great. Thanks, @Yumeng 03:22:43 shall we present project update in a Mandarin version ? or something like that? 03:23:19 I'll put that in https://etherpad.openstack.org/p/shanghai-ptg-cyborg . I'll ask OpenInfra Staff how to advertise that. 03:23:39 Meanwhile, if you can pitch your friends and colleagues to attend that, that would be great. ;) 03:24:01 About 2 hours maybe? 03:24:49 2 hours enough. 03:25:34 We can also mention that in the project update 03:25:57 All right, anything else on PTG? 03:26:07 yes, good idea. maybe around 1 hour is enough. but we can reserve 2 hours, no harm. 03:26:59 Cool. Please update and respond the https://etherpad.openstack.org/p/shanghai-ptg-cyborg etherpad later 03:27:14 #topic Tempest parallel tests 03:27:27 ok. will do after meeting. 03:27:29 I updated https://review.opendev.org/#/c/685183/ tempest tests 03:28:11 It still works only if the tests are done serially. One reason: the previous code stored the instance object as self.instance in the test case. I fixed that. 03:28:27 There are 2 more reasons that I mentioned in that patch's commit message. 03:28:42 Fake device has only one accelerator resource. 03:29:17 We may need to modify the fake driver to publish a large number of resources, as they are all virtual/fake anyway 03:29:22 Any ideas? 03:31:19 do yo mean we add more resources like more vfs ? 03:31:27 Yes 03:31:29 Enrichment of the resource in fake driver is good. ( when testing, I sometimes change that manually) 03:31:45 Are you saying that support more complicated fake-driver scenarios? 03:32:19 It can support launching multiple VMs in the tempest env, to run different tests in parallel. Otherwise, they all have to run serially 03:33:18 Great. 03:34:15 Cool. May be somebody can propose a patch to the fake driver for that 03:34:46 I can modify tempest to address the other problem, that all tests use the same device profile name today 03:35:16 #topic Core reviewers 03:35:26 Welcome, chenke, as a core reviewer! 03:35:52 Thanks! 03:36:12 I am very happy to meet you at cyborg. 03:36:17 cool! 03:36:42 congrats chenke! 03:36:48 Sundar: I just talked to zhurong, if we wants to create more than one VMs, he would suggest run serially instead of parallelly. maybe we can add in etherpad and discuss that. 03:36:58 One person listed as a core reviewer has not been active since Feb 2018 (almost 2 years): Justin Kilpatrick. I wrote an email to him but no response for 1 week. Shall we remove him as core reviewer? 03:37:38 Yumeng: I see. Sure, let me add that for PTG. Hope zhurong can attend that part. 03:38:38 I will invite him to make some time for us. ^^ 03:38:53 Yumeng, s_shogo, chenke: On core reviewer removal? 03:39:28 +1. Sundar 03:39:38 +1. 03:39:56 Cool, I'll follow up in openstack ML and follow due process. 03:40:40 #topic Code reviews 03:40:54 https://review.opendev.org/#/c/685542/ -- the famous Glance API servers patch :) 03:41:02 haha. 03:41:26 zhurong and Me test this code in myenv. 03:41:28 I don't understand the changes honestly. But, if you all understand it enough to maintain it, I'll go along with that 03:41:52 Hi all 03:42:00 Your test results show that this is also ok, right? 03:42:04 chenke: My concern is, without FPGAs, the programming code paths that invoke Glance don't get tested 03:42:04 Hi shaohe. 03:42:10 hi shaohe_feng 03:42:18 Sorry for late. seems something wrong with network. 03:42:31 I change a new proxy, it can works now. 03:42:31 Hi shaohe_feng : Would you be able to test https://review.opendev.org/#/c/685542/ in your local env with real FPGAs? 03:42:50 let me have a look. 03:43:04 DO others understand the change? 03:43:58 If shaohe can help verify it, that would be great. 03:44:15 the evn was destroyed by Xinran. 03:44:24 :') 03:44:29 chenke: yes, my tests with real FPGAs worked. But do we all understand the changes enough to maintain it if issues arise in the future? One part touches pecan WSGI and another part touches keystone 03:44:32 she used for other test. 03:45:01 Sundar do you have a FPGA env? 03:45:13 I can test it on it. 03:45:21 shaohe_feng: You still have FPGAs, right? Is Xinran using it for the demo? Yes, I do have FPGAs and I tested chenke's patch 03:46:00 yes. 03:46:26 I need to check with xinran 03:46:41 maybe we really need to set a FPGA gate 03:46:50 I understand your fears. Zhurong and I are good friends. He is confident in this code. 03:47:15 Sundar who can support a server with a FPGA? 03:47:22 Ok. Yumeng, s_shogo, shaohe_feng: please review this patch. If you all give +1/+2, i is automatic +2 from me. :) 03:47:31 OK. 03:47:46 chenke can you introduce what's change in the patch? 03:48:03 Ok. 03:48:18 maybe you have introduce in the meeting. But I missed it. 03:48:24 OK, I also test this with FPGA. 03:48:25 chenke thanks 03:49:02 in /etc/cyborg/cyborg.conf api_serveris useless actually. 03:49:31 The patch does this thing. 03:50:09 But When we remove this line. The past code occured some error. 03:51:23 So zhurong help to modify some error that the past code occured. 03:52:34 Well, could you clarify what's happening at https://review.opendev.org/#/c/685542/6/cyborg/api/hooks.py@79 ? 03:54:09 Sorry. I am not familiy with this part code now. But zhurong had explain this. 03:54:25 He said: The rootcause is less copied request context code when init the project. the main change is add necessary info into the context, 03:56:45 Or we can wait a little longer. I will do a detailed investigation of this code. 03:57:03 Are there reference code, from other component? ( if there, I would like to check this patch with that.) 03:57:12 Yes. 03:57:49 Quoted from other components, not his original. 03:57:52 Will this be needed if we move to openstacksdk? 03:58:09 why we need a X_SERVICE_CATALOG, but old code does not need this? 03:58:42 old code doesn't go this. 03:59:05 Sundar I am not sure. This patch was suggest by eric. 03:59:56 why we need to get service_catalog now? 04:00:09 @sh 04:00:20 shaohe_feng The if condition of the old code prevented this code from being executed. 04:01:58 no if condition in old code. 04:02:16 Wait. 04:02:51 https://review.opendev.org/#/c/685542/6/cyborg/image/glance.py@116 04:04:49 My suggestion: configuring the Glance API servers is not good, but introducing mysterious code to get rid of that is not necessarily better, May be we should focus more on moving to openstacksdk? Sorry, chenke, I know you put in a lot of work into this. 04:05:05 Ok. 04:05:09 I understand. 04:05:23 we can workflow -1 for this patch. 04:05:44 When will have free time. We can check it again. 04:06:12 Thanks for your understanding 04:06:26 Next patch: https://review.opendev.org/#/c/691766/ 04:07:28 seems like so many questions. we can dig more to make the whole problem clear. then decide whether workflow -1 or not 04:07:30 This is failing but the error seems to be trivial: https:// -587e1fde8c10362d45d985729e2fba7d.ssl.cf2.rackcdn.com/691766/4/check/cyborg-tempest-py3/56025ea/job-output.txt 2019-10-29 21:26:30.155717 04:08:31 looks like need to figure out the mysterious part first. I can help. 04:08:51 Yumeng: sure 04:10:02 For the py3 tempest patch, if this issue gets fixed, are we good with moving to Python 3 alone for tempest? Any other factor to consider? 04:12:29 suggest that we wait for other components. 04:13:12 ok 04:13:21 #topic AoB 04:13:24 Does everyone notice that openstack-tox-py37 test often timeout? 04:13:32 Yes. 04:13:38 shaohe_feng: Yes 04:13:40 Xinran's patch 04:13:45 what's the reason? 04:14:22 Me too, but no idea for the reason. 04:14:43 in several patches, I have seen timeout 04:14:50 From the zuul log. I didn't find the reason. 04:17:35 We can bring this up in openstack ML. If one of you want to write it up, good. Or, you can send me some concrete examples that I can cite, and I'll bring it up 04:18:28 Anything else for today, before we head to the summit? 04:18:57 BTW, to all, if you have time pls help to review ARQ and device_profile api doc :https://review.opendev.org/#/c/691306/, https://review.opendev.org/#/c/690539/ 04:19:10 Yes, thanks Yumeng. Will do. 04:19:35 Thanks! 04:19:47 hope to meet you all in shanghai. 04:20:04 Yumeng good job. 04:20:23 chenke I will not in shanghai. :') 04:20:38 will missing team dinner again. 04:20:39 See (most of) you next week in Shanghai. Shaohe_feng, we'll meet at the next Summit ;) 04:20:49 shaohe_feng: it's a pity :( 04:20:58 It's a pity. 04:21:14 Never have chance to attend the team dinner. 04:21:36 You were there at Berlin, right? 04:22:05 But not team meet in Berlin 04:22:16 No PTG in Berlin 04:22:41 team dinner is for developers, for PTG 04:22:42 I see. Anyways, safe travels to all. Bye. 04:22:46 #endmeeting