02:38:23 <xinranwang> #startmeeting openstack-cyborg
02:38:24 <openstack> Meeting started Thu Feb 25 02:38:23 2021 UTC and is due to finish in 60 minutes.  The chair is xinranwang. Information about MeetBot at http://wiki.debian.org/MeetBot.
02:38:25 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
02:38:27 <openstack> The meeting name has been set to 'openstack_cyborg'
02:38:28 <brinzhang> xinranwang:As I know, may people at outsaid play snow until 0:00 or later
02:38:45 <xinranwang> #topic roll call
02:38:57 <xinranwang> brinzhang:  wow, amazing.
02:39:00 <chenke> #info chenke
02:39:08 <xinranwang> #info xinranwang
02:39:14 <brinzhang> #info brinzhang
02:39:16 <songwenping_> #info songwenping_
02:39:21 <xinranwang> #topic vgpu status
02:39:51 <xinranwang> today, I think the main topic is vgpu. let's start by this one.
02:40:24 <songwenping_> i saw your new comment
02:40:53 <xinranwang> I prefer to config all pgpu explictly in config file. what do you think?
02:41:54 <songwenping_> it's complex i think
02:42:28 <songwenping_> if we have too many gpus
02:42:53 <chenke> if we didn't config, we default report it as pgpu. agree with wenping.
02:43:29 <xinranwang> what if admin wanna reserve some pgpus for host
02:43:38 <chenke> but also, if we explictly config pgpu in cyborg.conf, it will be more clearly for user
02:44:31 <songwenping_> xirnanwang: some other accelerators have the same problems as you say.
02:44:52 <chenke> maybe this should add a single conf in cyborg.conf.
02:45:45 <brinzhang> IMO, if there is more than one pGPUs on the host, we should support enable them at the same time
02:46:29 <brinzhang> may now they is less that scenario, but there need to be done, or in the future
02:48:08 <xinranwang> chenke:  +1, we need to add something like blacklist in cyborg conf  to disable the devices;
02:48:52 <chenke> I think xinrang and brin 's are all we need to implement .
02:49:20 <chenke> 1、enable/disable devcie 2、a blacklist
02:50:19 <songwenping_> so we should config different blacklist for different compute nodes.
02:50:31 <brinzhang> of course, that can be as an optimization, if we dont want to change in this patch ,we should have a todo
02:51:26 <brinzhang> more than one blacklist exist in the cluster that will be complex for users
02:52:00 <songwenping_> this patch cannot do it, better to do it in other change.
02:55:06 <brinzhang> songwenping_: agree, these things is important, and we can enable to config one at fist step
02:55:39 <songwenping_> i remember we plan to disable devices for reserve, admin can reserve the devices althrough they are reported.
02:55:58 <xinranwang> sorry. just lose the connection
02:56:18 <brinzhang> sorry, guys, I hsot to join and host another meeting
02:56:30 <xinranwang> brinzhang:  ok, bye
02:56:34 <chenke> bye
02:56:52 <songwenping_> bye
02:56:52 <brinzhang> I will return back to read the irc log after end the meeting
02:59:02 <songwenping_> xinranwang: can you see the record from 10:50 to 10:55
03:00:54 <xinranwang> yes, i can
03:00:55 <songwenping_> lost again?
03:00:58 <xinranwang> can we just config the pgpu explictly for now, and after we implements the disable interface, we can improve it to easier configuration?
03:01:13 <xinranwang> I have change the network. this time will be ok
03:01:21 <songwenping_> ok
03:01:31 <chenke> I agree. xinran.
03:01:42 <xinranwang> ok, great
03:01:47 <songwenping_> emm
03:02:24 <xinranwang> songwenping_:  or you prefer keep the current config?
03:02:41 <songwenping_> same question, do we config different pgpus for different compute nodes.
03:03:42 <xinranwang> it depends on the device. now we config at pci addr level, so the configuration should be different between compute node
03:05:20 <songwenping_> what about vgpu configure?
03:05:47 <xinranwang> what do you mean by vgpu configuraion?
03:07:22 <songwenping_> i cannot understand, does vgpu configure needn't to change?
03:07:54 <songwenping_> we just add a single config for pgpus?
03:09:02 <songwenping_> how we decide which gpus to virtualize?
03:10:58 <xinranwang> no, just forget the single config now. we use the config file in your patch, but just config all pgpu explictly
03:14:46 <songwenping_> ok, you can write down in the comment.
03:15:07 <chenke> hi xinran. wenping. suggest to talk in wechat's after our meeting.
03:15:23 <xinranwang> songwenping_:  ok, will do.
03:15:27 <xinranwang> #topic
03:15:34 <xinranwang> #topic tags
03:15:56 <xinranwang> small question aboout tags in gitignore
03:16:02 <xinranwang> what ctags used for
03:16:43 <songwenping_> it uses to jump function in vim tool.
03:16:53 <chenke> yes.
03:17:45 <xinranwang> oh, I don't know it before.  It seems useful ;)
03:18:17 <songwenping_> https://review.opendev.org/c/openstack/nova/+/237371
03:19:07 <xinranwang> got it
03:19:16 <xinranwang> I will give +2 to the patch
03:19:30 <xinranwang> ok, that's all from my side
03:19:44 <songwenping_> thanks
03:19:50 <xinranwang> thanks songwenping_  for the efforts on vgpu.
03:20:13 <xinranwang> do you have other things to bring up?
03:20:33 <songwenping_> seems yumeng is busy for the new work.
03:20:47 <xinranwang> Yes
03:21:33 <songwenping_> i have the the ssd driver to review https://review.opendev.org/c/openstack/cyborg/+/723310, thanks
03:21:49 <xinranwang> songwenping_:  ok, will do it by today
03:22:17 <songwenping_> nothing from myside
03:22:47 <chenke> +2 already.
03:22:48 <xinranwang> so let's wrap up the meeting. thanks guys.
03:22:57 <chenke> thanks all haha
03:23:10 <songwenping_> thanks
03:23:16 <xinranwang> #endmeeting