03:04:34 #startmeeting openstack-cyborg 03:04:35 Meeting started Wed Mar 6 03:04:34 2019 UTC and is due to finish in 60 minutes. The chair is Li_Liu. Information about MeetBot at http://wiki.debian.org/MeetBot. 03:04:36 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 03:04:38 The meeting name has been set to 'openstack_cyborg' 03:05:08 #topic Roll Call 03:05:16 #info Sundar 03:05:18 #info Li_Liu 03:05:25 #info xinranwang 03:05:28 #info Yumeng 03:05:44 hmm 03:05:52 let's wait for coco for a bit 03:05:58 let me ping her on wechat 03:11:17 Hi all 03:11:40 Hi Coco_gao 03:11:40 Hi Coco 03:11:45 let's get started 03:11:51 OK 03:11:55 #topic Status Update 03:12:01 #info Coco_gao 03:12:22 Have any of you tried to run https://review.openstack.org/#/c/640616/ ? I am getting an error 03:12:56 ccess wasdenied; cyborg:deployable;create 03:13:07 I tried .. it worked 03:13:29 well, we will not merge it for now 03:13:29 Apart from sourcing openrc as admin, did you do anything else? 03:13:55 Li_Liu: the code looks good. I may be missing some step. 03:14:12 let 's find out what I did 03:14:13 I finished most of the functions, but not pass the test. So I still need some some to upload the diff code. 03:14:23 some time 03:14:29 Coco_gao, thanks for the update 03:14:30 maybe today 03:14:39 great 03:15:16 xinranwang, how is the conductor part? 03:15:27 I may not have enough time for the unittest 03:15:44 Coco_gao, that's should be fine 03:16:20 All: please review https://review.openstack.org/#/c/641184/ 03:16:22 After discussion, we found Coco and I have some duplicate work on conductor side, so I leave her some interfaces, and coco implement the diff code there 03:16:41 xinran's code is depended on me, so I will help update her patch. The code is ready, just need some time for co-test. 03:18:02 Hi Sundar, thanks for that patch 03:19:05 Coco_gao: We'll fix the pep8 issues. But please see if this is how we want to fix and deprecate v1 API 03:19:38 OK, after my code being merged. 03:19:45 What's the deadline? 03:19:50 We are just fixing the return value. We should also fix the documentation. Is anybody handling documentation now? 03:21:11 Coco_gao: official deadline is March 8 for features. Bug fixes can go in after that too. 03:21:18 Sundar, I can do the doc update. 03:21:32 Yumeng_: Thanks a lot 03:21:33 Li_Liu: Also, I have improve the fpga driver to match the newest DB design, I think we'd better have at least one driver works in this release. 03:21:41 here is the patch https://review.openstack.org/#/c/599960/ 03:21:56 Sundar, after March 8, I can join for the doc. 03:22:00 xinranwang: Rodrigo has a patch almost ready for OPAE-based driver 03:22:15 xinranwang, Thank you. 03:22:46 Sundar: yes, but better to have one workable driver in this release. 03:23:21 I am pulled into another meeting... might be delayed in response 03:24:51 xinranwang, thanks a lot for the effort 03:25:27 Is zhenghao around ? 03:25:31 Li_Liu Coco_gao np, pls review it when you got time 03:25:34 Could we make the GPU driver as well ? 03:25:59 I will contact zhenghao for that. 03:26:36 That'll be a dash for the weekend deadline 03:27:13 I think that part won't change much, he had a old patch. 03:29:52 wangzhh 03:30:04 we need you. 03:30:09 Hi Coco. 03:30:19 Update the GPU driver 03:30:28 😁 03:31:12 FPGA driver is ready 03:31:41 No problem. But I don't have gpu device now. So Maybe I can just unit test. 03:33:03 wangzhh yep that should be fine 03:33:12 Anyway, I'll improve the gpu driver these days. 03:33:27 Plz do that before the S release cut 03:33:41 When? 03:34:31 wangzhh, do you need help on that? 03:34:40 This weekend I think 03:34:53 Li should know the exact date 03:35:18 Mar 04 - Mar 08 R-5 Stein-3 milestone Feature freeze 03:35:39 Mar 18 - Mar 22 R-3 RC1 target week Hard StringFreeze 03:36:22 This work can't be split. And I think I can finish it in time. 03:36:37 Then we still have time to squeeze more stuff in lol 03:36:58 🤣 03:37:10 We still have time fix bug? 03:37:16 My understanding is, we can at least merge the features first this week 03:37:19 And Li, what's the priority of deprecate api and new gpu driver ? 03:37:25 and fix them for another week? 03:38:15 That's good, cause I found some bugs and don't have time to fix right now. 03:39:03 I would caution deprecate anything before we have microversion in place 03:39:17 Bug fixes are needed ! 03:40:04 Hi all, I have to drop off, pls help review my patches this two days for the emergency. 03:40:11 these 03:40:11 zhipeng: The accelerators table got removed, so the API for accelerators is effectively deprecated 03:40:37 I will send my link in the wechat group. 03:40:41 bye 03:42:34 Sundar is it possible to at least return error code for that api call now ? 03:43:01 I'm a little bit cagy about removing APIs 03:43:11 zhipeng: with https://review.openstack.org/#/c/641184/, we treturn None or []. 03:43:12 In fact, we return None or [] now. 03:43:34 We should also update docs, as wangzhh said 03:43:43 Yep. 03:44:41 zhipeng: Totally understand your position. But, even if we put back the accelerator table, it will be useless as it doesn't get populated. 03:45:36 And I 'd like to add extra log info in these api. 03:45:45 *apis 03:46:07 wangzhh: Could you elaborate? Do you want something in the returned value, or cyborg-api logs? 03:46:37 Not return value. It is cyborg-api logs. 03:47:36 Sure 03:51:09 Anything else? 03:52:19 We keep getting questions on docs 03:52:27 I am still in another meeting >< 03:52:32 We need to update docs for API etc. 03:52:49 Sundar, we can do docs after the code freeze 03:52:50 I haven't looked at it. But I think it can be done after March 8? 03:52:58 yup 03:52:59 Li_Liu: ok 03:53:06 sorry, for late, just want to sure should we deprecate cyborg v1 api in Stein release? 03:54:04 yikun: No, we are not deprecating all v1 API. Only accelerator APIs 03:54:25 +1 03:54:37 Sundar: OK, and accelerator will return None/[], right? 03:55:00 yikun: Yes. Please review https://review.openstack.org/#/c/641184/ 03:55:16 Sundar, the deployable api does not work with new db schema. 03:55:22 Sundar: Cool, will do 03:55:31 wangzhh: Li_Liu's patch fixes that 03:55:51 ^ https://review.openstack.org/#/c/640616/ 03:56:34 Li is in internal meeting now, so he will response slowwww 03:57:02 OK any other matter we should take care for S ? 03:57:03 I saw it. But the defination of deployable is different, how to let others know? 03:57:40 wangzhh: Who is using these APIs today? 03:58:22 wangzhh: I see your point. Not sure how we do that without bumping versions 03:59:26 zhipeng, all: is anybody using Cyborg APIs today? 04:00:08 I just know another lenovo guys use it. And even though nobody use it, but somebody know the original concept of deployable. 04:00:54 We have product development in progress now 04:00:54 We should let them know we had changed this part. 04:01:55 wangzhh: Are you recommending moving to v2 in Stein? And deprecate v1? Then we will need v3 in Train 04:03:04 FWIW giving the time limit I would suggest postpone the version changes to T 04:04:57 Sure. I mean it's a long term work. We should maintain /v2/accelator, /v2/deployable, not mix v1/xxxx and /v2/xxx 04:05:16 in next version 04:05:58 I think we still keep deployable v1 for now, but fix the part according to our new db change 04:06:15 to make it compatible 04:09:18 +1 04:10:07 All right. 04:11:13 AT the PTG, let's talk about forming an API WG and a Documentation WG. 04:11:30 I'll work on getting microversions in 04:11:41 Cool. 04:12:24 Do we have etherpad to record what to disscuss now? 04:12:31 So let's try to get the driver code in before s-3 04:12:49 And bug fixes and doc fixes before RC1 04:13:12 wangzhh: Yes https://etherpad.openstack.org/p/cyborg-ptg-train 04:13:25 Sundar, Great. 04:14:04 zhipeng 04:14:09 Got it. 04:14:18 zhipeng: +1 04:16:38 Time for lunch, I'll drop off. 04:16:44 Bye :) 04:17:26 AoB? 04:17:34 #topic AoB 04:17:58 out engineering sync is tomorrow right? 04:18:03 Yes 04:18:31 great, will further sync with you guys tomorrow then 04:18:43 Bye! 04:18:44 thanks guys, have a good day/night 04:18:52 #endmeeting