02:39:16 <xinranwang> #startmeeting openstack-cyborg 02:39:17 <opendevmeet> Meeting started Thu Jun 3 02:39:16 2021 UTC and is due to finish in 60 minutes. The chair is xinranwang. Information about MeetBot at http://wiki.debian.org/MeetBot. 02:39:18 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 02:39:20 <opendevmeet> The meeting name has been set to 'openstack_cyborg' 02:39:39 <xinranwang> #topic meeting time proposal 02:40:26 <xinranwang> I have another weekly meeting conflicts with our meeting, can we start this meeting at 10:30 in the future? 02:40:43 <xinranwang> If you have other preference, please let me know. 02:40:44 <wenpingsong> let's start now， arthur is joining. 02:41:26 <xinranwang> wenpingsong: ok, thanks 02:42:26 <brinzhang> I cannot using the web page .. 02:43:10 <xinranwang> I use irccloud.com 02:43:33 <wenpingsong> i use this url https://kiwiirc.com/nextclient/ 02:43:56 <xinranwang> hi arthurdayne 02:43:59 <brinzhang> irccloud.com I have tried, but cannot join and didt receive the mail 02:44:46 <xinranwang> we are talking about new meeting time slot, I propose we start the meeting at 10:30, and waitng for other's feedback now 02:45:07 <arthurdayne> hi 02:45:27 <wenpingsong> 10:30 ok for me. 02:46:04 <xinranwang> what about you guys, brinzhang and arthurdayne 02:46:29 <arthurdayne> im ok, too 02:46:33 <brinzhang> 10:30 am I have a meeting but it's happen in each two week 02:47:26 <brinzhang> But I think it's ok, agree 02:47:33 <xinranwang> Hmmm, which time slot you guys are all avaible? 02:48:06 <xinranwang> then I can check my schedule 02:48:21 <brinzhang> no, follow your heart ^ 02:48:31 <xinranwang> lol 02:48:33 <wenpingsong> i am avaiable all time. 02:48:49 <xinranwang> ok, let's start at 10:30, and let's see... 02:48:56 <xinranwang> I will update the doc 02:49:07 <xinranwang> #topic spec review 02:49:47 <xinranwang> there are some pending spec, please review them. is there any block in the process? 02:51:09 <brinzhang> There is no pending 02:51:09 <xinranwang> about rootwrap, I suggest we switch to privsep directly, what do you think? 02:51:46 <xinranwang> I mean not merged spec... 02:52:25 <wenpingsong> we cannot get an accout to download vgpu software. seems nvidia rejects our corporate email address. 02:52:54 <wenpingsong> you means we can use privsep instead of rootwrap? 02:53:04 <xinranwang> wenpingsong: yeah 02:53:37 <xinranwang> https://docs.openstack.org/oslo.privsep/latest/user/index.html#converting-from-rootwrap-to-privsep 02:54:06 <xinranwang> wenpingsong: oh, you rebuild your gpu env, and have no software now? 02:54:35 <wenpingsong> ok i'll look into this doc. 02:54:45 <brinzhang> privsep is better than rootwrap, it has high performance 02:54:54 <wenpingsong> yes no software. 02:56:47 <wenpingsong> does nova has convert to privsep? 02:57:03 <xinranwang> wenpingsong: yes, nova use privsep now 02:57:24 <wenpingsong> cool 02:57:25 <brinzhang> yes, in some place nova already changed to use privsep 02:57:37 <xinranwang> wenpingsong: about the gpu, maybe you need check with internal colleagues... 02:57:45 <brinzhang> I remember I sent you some links from gerrit, you can check 02:58:35 <xinranwang> wenpingsong: please check the privsep stuff, and update the spec, after this, I think it is ready to go 02:58:44 <xinranwang> #topic doc improvement 02:58:56 <wenpingsong> ok, thanks. 02:59:52 <xinranwang> arthurdayne: is there anything we can help with that? 03:00:10 <arthurdayne> i have submit a bug in launchpad 03:00:22 <arthurdayne> i am working in progress 03:00:48 <arthurdayne> https://bugs.launchpad.net/openstack-cyborg/+bug/1930404 03:00:50 <opendevmeet> Launchpad bug 1930404 in Cyborg (OpenStack) "Docs: Missing The Section about Cyborg - Nova Interaction" [Undecided,Confirmed] - Assigned to Arthur Dayne (palagend) 03:02:03 <brinzhang> arthurdayne: cool 03:02:20 <arthurdayne> i am planning to tranfer the some content of "nova - cyborg intercation' to cyborg, ok? 03:02:21 <xinranwang> arthurdayne: ok cool 03:02:48 <xinranwang> could you copy the link here? 03:02:49 <brinzhang> while you add the docs instruction for nova-cyborg interaction, that you can ping us 03:03:38 <xinranwang> arthurdayne: ah ok, I got your point. Sure, it is valuable for cyborg 03:03:54 <brinzhang> xinranwang: the link just a reference, we also need a docs page to show nova cyborg how to work together 03:04:03 <arthurdayne> i am working it locally, have not yet submit to gerrit 03:04:35 <xinranwang> brinzhang: yes 03:04:43 <arthurdayne> yes 03:05:02 <xinranwang> arthurdayne: ok, you can feel free to push it one it's ready. Please ping us if any question 03:05:10 <xinranwang> #topic AoB 03:05:20 <arthurdayne> ok, thanks 03:05:31 <xinranwang> that's all from my side, anything else you wanna bring up here/ 03:07:21 <arthurdayne> noting else 03:07:33 <wenpingsong> dose any guys know this patch https://review.opendev.org/c/openstack/cyborg/+/793103? 03:07:35 <brinzhang> arthurdayne: as I talked with you, bug 1927056 and bug 1927057 you can also fixed 03:07:37 <opendevmeet> bug 1927056 in Cyborg (OpenStack) "Docs: Missing parameter list in Accelerator Requests APIs" [High,Confirmed] https://launchpad.net/bugs/1927056 - Assigned to Brin Zhang (zhangbailin) 03:07:38 <opendevmeet> bug 1927057 in Cyborg (OpenStack) "Docs:Missing parameter list in devices APIs" [High,Confirmed] https://launchpad.net/bugs/1927057 - Assigned to Brin Zhang (zhangbailin) 03:07:51 <wenpingsong> why we define ARQ_DELETING as 'failed' state? 03:09:02 <brinzhang> <arthurdayne> there is a commit fixed the devices APIs that you can see https://review.opendev.org/c/openstack/cyborg/+/792928 03:09:13 <xinranwang> wenpingsong: that's also my question...lol 03:09:50 <wenpingsong> i think it's deleting state. 03:10:15 <wenpingsong> let me find who submit this code. 03:11:53 <xinranwang> ok, maybe the "deleting" status return "failed" directly, but i think it is not reasonable 03:12:26 <arthurdayne> brinzhang:ok 03:12:28 <wenpingsong> it's Sundar 03:12:37 <brinzhang> The delete ARQ task we are using sync mothod, deleting may very fast, otherwise the device is lost, and we can get failed 03:12:58 <xinranwang> need double check it 03:14:00 <xinranwang> why arq turn to "deleting", i think it will return quickly with success or failure 03:14:25 <brinzhang> I agee with the fix now, for arq_deleting we dont need the failed status 03:15:40 <xinranwang> agree 03:15:50 <xinranwang> let's keep the deleting status 03:15:52 <wenpingsong> Each ARQ has its own state. However, the bind status sent to Nova should be 'completed' or 'failed'. The logic to do that conversion should not be in nova_client.py, to keep it free of ARQ state details. So it has been added in get_arq_bind_status() in ext_arq_job.py. 03:16:32 <brinzhang> xinranwang: one scenario, if the device is using by a instance, may we cant delete it quickly, such as deleting a instance's attached volume, if there is IO read/write it will return delete failed 03:17:21 <brinzhang> but's the delete action will get a failed status, rather trhan deleting_failed 03:18:23 <wenpingsong> from Sundar's commit message, it works for arq binding. 03:19:12 <wenpingsong> let's discuss offline. 03:20:22 <xinranwang> sorry 03:20:33 <xinranwang> I just lost my connection 03:21:11 <xinranwang> wenpingsong: sure, let’s discuss offline 03:21:15 <xinranwang> Anything else 03:21:23 <brinzhang> nothing, let's end of this meeting 03:21:28 <xinranwang> I use my cellphone now... 03:21:34 <xinranwang> Ok 03:21:42 <xinranwang> Thanks guys 03:21:48 <xinranwang> #endmeeting