02:07:59 #startmeeting openstack-cyborg 02:07:59 Meeting started Thu May 27 02:07:59 2021 UTC and is due to finish in 60 minutes. The chair is xinranwang. Information about MeetBot at http://wiki.debian.org/MeetBot. 02:08:00 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 02:08:02 The meeting name has been set to 'openstack_cyborg' 02:08:54 Hi qiujunting! is it Fossen 02:09:08 I began to learn cyborg with cyborg doc, but when I read the cyborg arch pictures, I am a bit confused 02:10:23 xinranwang: yeah, qiujunting is Fossen named in Gerrit 02:11:22 So I decide to give a more detailed architecture picture to help anyone who want to learn cyborg 02:11:42 yeah 02:12:10 ArthurDayne:awesome!, your first patch has been approved, that you should add the nova-cyborg interaction patch, thanks 02:12:46 ArthurDayne: that's great! it helps a lot for the beginner, thanks for your patch 02:13:04 My pleasure 02:13:44 qiujunting: could you please also introduce yourself a little bit. I think you all know each other lol 02:14:25 qiujunting has learned to manage PMEM device by cyborg. 02:15:00 great! do you plan to support pmem in cyborg? 02:15:54 not sure. 02:16:51 when pmem is use as memory, it will be conflict with the vm maxmemory 02:17:04 xinranwang: whether need to support PMEM in Cyborg, need to be talk more 02:17:05 do you know this problem? xinranwang 02:17:12 cyborg support pmem is not friendly, PMEM device is not like other pci device. 02:17:17 and there some some issue as wenping said 02:18:14 can we support another mode, I remember pmem has 2 modes 02:18:17 I think xinranwang is not simarlier with PMEM, as I know PMEM was maintained by other teams 02:18:49 3 modes, memory mode, AD modes and storage over AD 02:18:58 now nova use memory mode, right? 02:19:10 brinzhang: got it 02:19:18 yes, nova treat it as memory mode 02:19:32 even we use it as storage disk 02:19:34 xinranwang: I plan to discuss some issue with pmem in hackthon with your nova team 02:20:14 at nova there are some problems about pmem models 02:20:40 let's discuss after meeting ok? 02:20:50 ok, let's talk about this during hackathon, there are some pmem experts 02:21:17 yeah 02:21:33 #topic vgpu support 02:21:53 #info wenpingsong 02:22:01 oh sorry 02:22:07 #topic roll call 02:22:15 #info xinranwang 02:22:22 #info wenpingsong 02:22:35 welcome qiujunting and ArthurDayne join our meeting, we are meet time on 10:00am, at Thurday in each week 02:22:39 #brinzhang 02:23:21 cyborg support vgpu complete？ 02:23:42 welcome qiujunting and ArthurDayne, hope we have more collaboration int the future 02:23:52 #topic vgpu support 02:23:53 conguratulation 02:24:56 let's merge the spec first. 02:25:06 many users are interested in vgpu support, I have received some mail which asking the status 02:25:22 my env is down, i cannot continue now. 02:25:27 I hope we can land it in x release if possible. 02:25:55 the spec in cyborg, I think it's ready to go 02:26:01 but yumeng's code is well on the whole. 02:26:27 brinzhang: yeah, I will review it during hackathon, to make it also an output lol 02:26:52 wenpingsong: did nova guys have some feedback on nova spec? 02:27:04 ack 02:27:37 xinranwang: no, i'll ping them today again. 02:27:40 I plan review the vgpu spec during hackathon and asked Alex to help 02:28:11 brinzhang: good method. 02:28:17 wenpingsong: ok, fyi, there is also a mdev framework proposed in nova 02:28:58 mdev framework? 02:29:01 wenpingsong: will you rebuild your env to continue? 02:29:20 yeah, i will continue soon. 02:29:58 wenpingsong: great 02:30:12 please check it, it may helps. https://review.opendev.org/c/openstack/nova-specs/+/792796 02:31:22 cool. i'll look into this patch. 02:32:17 #topic doc improvement 02:34:22 I got some feedback from users, they asks the nova/cyborg version requirement and said the doc is not very friendly for a beginer. I hope we can improve the doc in this release as well 02:35:01 yeah, cyborg has some interactions with nova and cyborg, but in the doc, we cannot find some detail about that 02:35:40 Yep, that we need to improve the docs readability 02:36:13 agree 02:36:21 so is ther some suggestions from user? 02:36:26 ArthurDayne: that you can try to wirte, and we are all review and can hep to complete it 02:36:29 s/ther/there 02:37:07 ArthurDayne: if you find something confused or need to improved during your ramp up, please let us know. you can also improve it by yourself. 02:37:08 brinzhang: ack 02:37:37 thanks 02:37:40 xinrangwang: imo, we could add a section to introduce the interaction about nova and cyborg 02:37:53 wenpingsong: I think we don't have suggestion from deployer about the version requirement now 02:38:22 your can start a new page, and we can import it to the cyborg index page 02:38:25 we havenot the interaction doc between nova and cyborg? 02:38:35 https://specs.openstack.org/openstack/nova-specs/specs/ussuri/implemented/nova-cyborg-interaction.html 02:38:58 there is a spec to explain that. but it is in nova 02:39:32 oh, then we need to add this in cyborg side. 02:40:19 yeah, we can make this a ref and explain the overview simply 02:41:14 we should reword these description to the cyborg docs, not just a hyper link, it's not friendly 02:41:34 brinzhang: of course, this is a reference 02:41:51 yes, it's similar with the nova spec. 02:41:55 ArthurDayne, wenpingsong: we can reference this nova-cyborg interaction spec to competed cyborg's instruction 02:42:23 s/complete/competed 02:42:41 there are 2 point of view. one is for deployer, one is for developper 02:42:47 yes 02:43:57 so... anyone is availble to do this? 02:44:08 we are developer, not good at doc write :( 02:44:42 let me to do it 02:45:19 from a beginner's view to write it? 02:45:25 ArthurDayne: cool, thanks a lot. please remember to create a bp on launchpad about this 02:45:43 ok 02:45:52 cool 02:46:36 I think you can start by deployer's view, like how to boot up a VM with accelerator, what configuration admin need to do for this. Then we can dive into the code flow. 02:46:53 what do you think 02:47:16 xinranwang: good idea 02:47:43 ok, thanks ArthurDayne 02:47:49 agree with xinranwang 02:48:20 we have only 10 mins left. 02:48:35 #topic smartnic support 02:49:08 the implementation is rebased, and we plan to review it during hackathon. 02:49:26 #topic disable/enable api 02:50:21 #topic spec review 02:50:39 i need the rootwrap spec to review. 02:51:00 I noticed there are some pending spec from wenping, and also the disable/enable api. all of this will be review during hackathon 02:51:14 let's review them face to face, it's more efficient 02:51:30 wenpingsong: during hackthon we plan to compete your rootwrap spec 02:51:38 that's my opinion. what do you think 02:51:57 ok cool 02:52:56 #topic AoB 02:53:11 anything else you want bring up? 02:53:39 nothing from me 02:53:50 https://review.opendev.org/c/openstack/cyborg/+/791084 i think we need update rp inventory 02:53:56 when rc is changed 02:55:03 yeah 02:55:16 and do we need to cleanup old trait and rc if they are changed? 02:56:05 this week we will switch freenoe with OFTC, and the details I have not time to review, if you have time, please review it 02:56:17 this is the ML http://lists.openstack.org/pipermail/openstack-discuss/2021-May/022718.html 02:57:24 wenpingsong: I think it's not neccessay. but we can do this by not raising a exception in rc is still used 02:58:19 brinzhang: thanks for the link 02:58:20 ok let's discuss offline. 02:58:37 if nothing else, let's wrap up the meetin 02:58:37 yeah, we can catch and logged it, but not be raised 02:58:53 byebye 02:58:58 #endmeeting