03:10:31 #startmeeting openstack-cyborg 03:10:33 Meeting started Thu Dec 5 03:10:31 2019 UTC and is due to finish in 60 minutes. The chair is Sundar. Information about MeetBot at http://wiki.debian.org/MeetBot. 03:10:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 03:10:36 The meeting name has been set to 'openstack_cyborg' 03:10:44 Sorry for joining late 03:11:03 Anybody here? 03:11:09 #info chenke 03:11:16 hi sundar. 03:11:24 #info Li_Liu 03:11:32 #info Yumeng 03:11:34 Hi Sundar 03:11:51 Hi Li_Liu and Sundar 03:11:52 Hi all 03:12:03 #info s_shogo 03:12:31 Sorry, I haven't posted the agenda for today 03:12:47 Could we start with functional testing, Li_Liu? 03:13:14 ok, as I discussed last week 03:13:17 https://etherpad.openstack.org/p/cyborg-ussuri-functional-test-plan 03:13:30 the test plan is posted here. and prob needs more inputs 03:13:52 it's welcome to provide comments and inputs in the etherpad 03:14:16 Great. Are there any considerations to mock the db interface? 03:14:55 I think we will need that 03:15:12 Let me do some research for it 03:15:17 Yea, cool 03:15:30 I think we are doing similar mocking in unit tests 03:16:08 Yea. The db mocking could be a fixture. 03:16:54 Like https://review.opendev.org/gitweb?p=openstack/nova.git;f=nova/tests/fixtures.py;hb=refs/changes/35/673735/20#l2463 03:17:38 Is Shaohe here? 03:18:18 Multinode is another important area. 03:18:42 Hi s_shogo, I know the client patch is waiting for review. 03:19:18 Did you get any comments from outside Cyborg? 03:19:31 On openstacksdk, etc.? 03:19:42 Hi Sundar, Thanks for your adding reviewers. 03:20:11 No comments,yet. 03:21:19 Ok. Is it better to ask for outside reviews now, or after we add /v2/devices etc. in the client? 03:23:59 s_shogo: ^ 03:24:00 IMO, adding of Device doesn't effect openstack sdk ,significantly. Thus, I would like to ask review now. 03:24:17 Ok, I'll try to find somebody. 03:24:31 Hi shaohe_feng 03:24:42 Thanks for the patch on fake driver 03:24:43 Thank you Sundar ! I also try to send e-mail and notify someone, after that. 03:25:38 Welcome, s_shogo 03:26:18 shaohe_feng: What else do you need for the multinode? Are you getting reviews for the config table? 03:28:40 Not sure if he is here -- I saw his join message. 03:28:52 s_shogo I have put some comments in your client patch. 03:29:31 Thank you cheke ! I'll check it. 03:29:43 chenke: Do you mean https://review.opendev.org/681391 ? 03:29:59 Ye. 03:30:26 s/cheke/chenke/ 03:30:39 Cool. chenke, what is the next step for https://review.opendev.org/#/c/691549/ ? 03:31:26 his job should be done in Phase-2: milestone-1 -> milestone-2 03:31:55 Ah I see. Sure. 03:32:08 ok. 03:32:35 Next week, we should go through Storyboard, cyborg client patches and tempest plugin patches. We don't look much at those. 03:32:53 Anything else for today? 03:34:01 seems lot of cyborg patch need to be review. 03:34:40 Yea. I've been focusing on the Nova patches, because we have the reviewer attention now and will lose it in week, as people head for holidays. 03:34:46 *in 1 week 03:35:30 oh. that should be high priority. 03:35:41 Please review the patches we have, and I'll catch up in the last 2 weeks of Dec 03:36:17 will do it. 03:36:38 Thanks. Anything else for today? 03:36:46 About RBAC policy. 03:36:48 I've read most of the materials provided on this page: https://wiki.openstack.org/wiki/Consistent_and_Secure_Default_Policies_Popup_Team 03:36:54 Sundar config table is create by me 03:36:55 We need to reassess each operation and associate which scope, or scopes, are appropriate. Mybe just like this: https://wiki.openstack.org/wiki/Barbican/Policy 03:37:20 I need all of you review it 03:37:22 then it will need all your comments. 03:38:01 Yumeng: Sounds good. Would you be creating that wiki? Do you need any help? 03:38:24 yep. wiki is good. I will do that. 03:38:51 I will try first. 03:38:54 Great. I'll add this to the agenda for next week. 03:39:34 now any others have comment on the config table? 03:41:08 Any suggestion on how to improve the program enable on agent? 03:42:06 shaohe_feng: Please post the etherpad link again here, to ask for review 03:44:11 #linkinfo https://etherpad.openstack.org/p/cyborg-ptg-ussuri 03:45:20 shaohe_feng and all, let's wait 1 more week for review. We'll take that as as the final table. Ok? 03:45:49 OK 03:46:08 so let's focus on agent program 03:46:15 any comments? 03:46:27 You mean the patch? 03:47:01 If so, we should get the comments there 03:47:09 Sundar s_shogo yikun, Yumeng, wangzhh, chenke 03:47:45 yes, we have discussed them yesterday. 03:47:52 on wechat 03:48:08 no agent just support fpga program 03:48:27 how should agent support fake driver program? 03:48:57 BTW, in terms of holiday planning, we should probably cancel the meetings on Dec 25/26 and Jan 1/2. Probably keep the meeting on Dec 18/19? 03:50:59 OK. 03:51:31 s_shogo Yumeng chenke Li_Liu: ^ 03:51:33 NP > schedule 03:51:43 let's load all drivers in agant manager init? 03:51:58 if it is configure in setup.cfg 03:52:08 shaohe_feng: Are we not doing that today? 03:52:23 and choose the right one 03:52:40 yes, I have paste a picture in wechat 03:52:59 for program we just support fpga 03:53:13 let me show you the code. 03:53:28 shaohe_feng: agent manager init calls into resource tracker and loads drivers: https://opendev.org/openstack/cyborg/src/branch/master/cyborg/agent/resource_tracker.py#L47 03:53:47 no this code 03:54:11 We call program API only on fpga driver today 03:54:26 We should expand that to fake driver also 03:54:55 you show the resource tracker code 03:55:21 https://opendev.org/openstack/cyborg/src/branch/master/cyborg/agent/manager.py#L41 03:55:36 shaohe_feng: We get the driver name in fpga_program_v2 API: https://opendev.org/openstack/cyborg/src/branch/master/cyborg/agent/manager.py#L64 03:56:10 here we init the self.fpga_driver 03:56:12 We can use that to call the right driver. No need to hardcode self.fpga_driver. 03:56:24 https://opendev.org/openstack/cyborg/src/branch/master/cyborg/agent/manager.py#L70 03:56:38 so here we just let fpga do program 03:57:05 shaohe_feng: ^ 03:57:26 OK, let's cancel the self.fpga_driver init 03:57:50 add self.drivers in AgentManager 03:58:23 all drivers if implement program method 03:58:30 let's do it. 03:58:41 no matter it is fpga 03:59:01 We already have the list of drivers in self._rt.acc_drivers 03:59:44 The program_v2 shold check if the driver implements the program() API and then call it. 04:00:19 Do others have any thoughts, comments, objections, .... ? 04:01:08 OK, let's review it in patch changes. 04:01:28 Anything else, folks? 04:02:12 #topic AoB 04:02:25 If nothing else, we can call it a day. 04:03:19 Tanks, everybody. Have a good day! 04:03:23 *Thanks 04:03:29 #endmeeting