02:05:31 <xinranwang> #startmeeting openstack-cyborg
02:05:32 <openstack> Meeting started Thu Jan 14 02:05:31 2021 UTC and is due to finish in 60 minutes.  The chair is xinranwang. Information about MeetBot at http://wiki.debian.org/MeetBot.
02:05:33 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
02:05:35 <openstack> The meeting name has been set to 'openstack_cyborg'
02:06:07 <xinranwang> #topic roll call
02:06:16 <xinranwang> #info xinranwang
02:06:21 <brinzhang_> #info brinzhang
02:06:26 <chenke> #info chenke
02:06:27 <songwenping_> #info songwenping_
02:06:52 <xinranwang> #topic vGPU progress
02:07:35 <xinranwang> songwenping_: do you have some updates about vgpu, I saw you and yumeng have update the patches.
02:07:37 <songwenping_> i am testing yumeng's patch and meet some problem
02:07:57 <chenke> what's problem
02:08:11 <chenke> songwenping_
02:08:31 <songwenping_> as the trait is OWNER_CYBORG, cyborg-conductor can't report data to placement
02:09:03 <songwenping_> so yumeng commit the patch of os-trait project: https://review.opendev.org/c/openstack/os-traits/+/770569
02:09:23 <chenke> it should add it in os_traits
02:09:27 <xinranwang> Oh, the customized trait must start with CUSTOM
02:09:47 <chenke> yumeng said she had talked with you and nova's guys.
02:09:53 <songwenping_> xinranwang: right
02:10:43 <brinzhang_> +1, add the CUSTOM_ prefix
02:10:49 <songwenping_> but donnot know what nova guys' opnion
02:11:11 <Yumeng> hi all
02:11:53 <xinranwang> Hi Yumeng, we are talking about the vGPU traits
02:12:02 <songwenping_> s opnion/opinion
02:12:04 <brinzhang_> Yumeng: we are talking vGPU patch, hopw you can give some info form nova guys
02:12:17 <brinzhang_> s/hopw/hope
02:12:42 <Yumeng> we already have an agreement in PTG, nova guys does not suggest using CUSTOM_ . pls check line 150 https://etherpad.opendev.org/p/nova-wallaby-ptg
02:13:51 <brinzhang_> I remember, they said should add CYBORG or NOVA subfix, the CUSTOM_ we shuold keep to use
02:14:07 <Yumeng> they prefer having OWNER_NOVA, OWNER_CYBORG in os-traits as a standard traits
02:15:03 <xinranwang> use OWNER as the prefix, right?
02:15:10 <xinranwang> a new kind of trait
02:15:33 <Yumeng> yes, it is a new trait
02:16:10 <Yumeng> setting them as standard traits means these traits can be used not only for GPU, but also for fpga, sriov etc.
02:16:40 <brinzhang_> That wrote in Line150
02:16:57 <chenke> I think this is a good idea when cyborg and nova jointly manage vGPU.
02:18:41 <chenke> songwenping_ if you want to test, you should modify os_traits first.
02:20:15 <songwenping_> ok i'll try
02:21:16 <xinranwang> sorry, just lost the connection
02:21:39 <xinranwang> what do you think about stephen's comments?
02:25:12 <Yumeng> xinranwang: aha, I got stephen's meaning.
02:25:18 <xinranwang> do we need to add the new traits definition in other place?
02:26:12 <Yumeng> seems we need place new traits in another path, maybe like the os-traits/owner.py
02:27:11 <xinranwang> Yumeng: yes, it seems the new prefix will be added according to the dir name.
02:27:50 <Yumeng> emm better find a env to have a try
02:28:11 <Yumeng> I will try that. and I think we also need nova guys to take a look this patch.
02:28:56 <xinranwang> Yes, sure. Thanks Yumeng
02:29:20 <xinranwang> Ok, do you guys have any other opens about vgpu?
02:30:15 <xinranwang> if not, let's move to next topic.
02:30:46 <Yumeng> nothing from my side
02:31:11 <xinranwang> #topic patch review
02:31:32 <xinranwang> there are some pending patches under reviewing.
02:32:22 <xinranwang> the first one is https://review.opendev.org/c/openstack/cyborg/+/766629 from Shiyan
02:33:19 <openstackgerrit> wdd proposed openstack/cyborg master: Revert remove py37  https://review.opendev.org/c/openstack/cyborg/+/770693
02:34:39 <xinranwang> it is useful for a production env
02:35:20 <xinranwang> Please help to review it
02:36:26 <xinranwang> I have a 710 driver patch under review, please help to review it as well
02:36:50 <openstackgerrit> wdd proposed openstack/cyborg-tempest-plugin master: Revert remove py37  https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/770694
02:38:00 <songwenping_> sure will look into
02:38:03 <chenke> ok
02:38:12 <chenke> i have a advice.
02:38:14 <brinzhang_> ack
02:39:17 <chenke> after irc meeting. If we have some urgent issues, we can communicate in Chinese
02:40:12 <brinzhang_> if possiable, it can be done in wechat
02:40:19 <chenke> yes
02:40:25 <xinranwang> lol, sure.
02:41:44 <chenke> haha
02:42:03 <xinranwang> It will be more efficient
02:42:12 <xinranwang> :)
02:42:15 <brinzhang_> in IRC, EN is better, because we have many people in this channel, may they will see the irc log after they are wake up
02:42:17 <chenke> about your 710 driver. have your create instance ?
02:42:35 <chenke> brinzhang_ agree.
02:42:42 <xinranwang> chenke:  yes, I have a  env to test it
02:43:21 <chenke> xinranwang ok, will learn about it.
02:43:24 <xinranwang> I am still updating it according to Yongli's comments.
02:43:43 <xinranwang> Your comments are welcome, so that I can update at once :)
02:44:55 <xinranwang> that's all from my side.
02:45:28 <chenke> I have a question. If we use this, do we need to set some properity in nova ports?
02:46:26 <chenke> It is recommended that you include this usage example.
02:47:02 <xinranwang> yes, actually there are a set of changes in Nova/Neutron and Cyborg if you want to boot up a VM with 710
02:47:20 <xinranwang> As for discovery function, the Cyborg patch is suffisant
02:48:03 <xinranwang> https://review.opendev.org/q/topic:%2522bp/sriov-smartnic-support%2522+
02:48:28 <chenke> thanks, will learn it.
02:49:04 <xinranwang> the guideline is in nova patch's commit message. Please refer to https://review.opendev.org/c/openstack/nova/+/758944
02:49:20 <chenke> If this smart network card is incorporated, it will be a big improvement for our cyborg.
02:49:21 <xinranwang> chenke:  thanks ;)
02:50:35 <xinranwang> chenke:  yes, that's true. Haha
02:50:54 <xinranwang> #topic AoB
02:51:40 <xinranwang> Do you have any other things to bring up?
02:52:00 <chenke> no for me.
02:52:55 <songwenping_> plz review my SSD driver spec: https://review.opendev.org/c/openstack/cyborg/+/766629 thanks
02:53:33 <songwenping_> sorry not this patch
02:53:41 <songwenping_> https://review.opendev.org/c/openstack/cyborg-specs/+/759542
02:53:45 <xinranwang> https://review.opendev.org/c/openstack/cyborg/+/723310 this one
02:54:31 <brinzhang_> and there is the cyborg shelve/unshelve support, nova team give some coments, wenping will update later
02:54:32 <songwenping_> spec first
02:54:48 <chenke> cool
02:54:54 <xinranwang> Oh, there are both spec and implementation.
02:55:25 <songwenping_> thanks
02:55:31 <xinranwang> brinzhang_:  cool, thansk songwenping_
02:55:48 <xinranwang> I will review SSD spec by today.
02:56:44 <chenke> songwenping_ recently, I am so busy. will learn your patch as soon as possible.
02:56:48 <songwenping_> nothing from myside.
02:57:37 <songwenping_> chenke: no hurry :?
02:57:41 <songwenping_> :)
02:57:59 <brinzhang_> in the end of year, everyone is busy, so we can promote the privority of the spec review
02:58:23 <brinzhang_> before the deadline, we can deal with our spec
02:59:31 <xinranwang> the end of next week is cyborg spec freeze.
03:01:17 <songwenping_> the vgpu spec in nova side is pending
03:02:18 <songwenping_> i am afraid it will be frozen.
03:03:17 <xinranwang> Do we have any change to push it in? Or it still need time to improve it?
03:04:06 <songwenping_> a small question from buzas
03:04:41 <songwenping_> Yumeng and i plan to use a new function to attach mdev
03:05:37 <songwenping_> he want to reuse the old function
03:05:38 <songwenping_> https://review.opendev.org/c/openstack/nova-specs/+/750116
03:06:49 <songwenping_> xinranwang, brinzhang_: what's your opinion?
03:07:57 <xinranwang> I haven't look into the spec. I am afraid that I can't have any output now.
03:08:46 <xinranwang> I will look into the spec. and we can discuss it offline. Please involve me in your discussion if possible.
03:10:07 <xinranwang> songwenping_:  does it works for you ?
03:11:18 <songwenping_> yeah
03:11:22 <brinzhang_> songwenping_: IMO, follow bauzas's opinion, he doesnot want us to use his interface, we can
03:11:53 <brinzhang_> that just an implementation, it doesnot contained in the spec
03:12:07 <brinzhang_> s/doesnot/does
03:12:18 <songwenping_> ok
03:13:09 <songwenping_> i will update and merge the spec first.
03:14:12 <xinranwang> thanks songwenping_
03:14:26 <xinranwang> ok, we are running out of time, let's wrap up the meeting.
03:14:34 <xinranwang> Thanks for your supports :)
03:14:46 <songwenping_> thanks bye
03:14:50 <xinranwang> byebye
03:15:53 <xinranwang> #endmeeting