03:09:43 <Li_Liu> #startmeeting openstack-cyborg
03:09:44 <openstack> Meeting started Wed Feb 13 03:09:43 2019 UTC and is due to finish in 60 minutes.  The chair is Li_Liu. Information about MeetBot at http://wiki.debian.org/MeetBot.
03:09:45 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
03:09:47 <openstack> The meeting name has been set to 'openstack_cyborg'
03:09:55 <Li_Liu> #topic Roll Call
03:10:02 <Sundar> #info Sundar
03:10:04 <Li_Liu> #info Li_Liu
03:10:13 <xinranwang> #info xinranwang
03:10:22 <Yumeng_> #info Yumeng
03:10:28 <Li_Liu> Hi guys
03:10:39 <Sundar> #help
03:10:45 <Sundar> Hi Li
03:10:49 <wangzhh> #info wangzh
03:11:07 <Li_Liu> one suggestion tho, can we move the engineering call to a different time? it's a bit close to the regular IRC
03:11:36 <Li_Liu> Say, move it to a date between 2 irc meetings
03:11:54 <Sundar> I am fine with other days. I think we did a Doodle poll and chose this
03:12:37 <Li_Liu> I feel like we just talked no long ago.. and don't have much update since then
03:12:41 <Sundar> May be we can do Thursday eve US time, Fri morn China time?
03:12:54 <Li_Liu> That sound fine to me
03:13:12 <Sundar> I see. I believe the agenda for theis call and the engineering call will be somewhat different
03:13:24 <Sundar> This call is about status updates, process etc.
03:13:36 <Sundar> Whereas the engineering call goes more into depth
03:13:56 <Li_Liu> I know, but moving them apart makes more sense
03:14:21 <Li_Liu> wangzh, do you know if coco is coming?
03:14:29 <Li_Liu> She's the super star right now :P
03:14:44 <wangzhh> Not sure.
03:14:51 <xinranwang> Friday is fine for me, better before 10:30. :)
03:15:20 <Li_Liu> #topic status updates
03:15:33 <Li_Liu> Good thing is her patch get merged
03:15:50 <Li_Liu> still waiting for the OVO one
03:16:01 <Li_Liu> Sundar, I saw you still have some comments
03:16:44 <Sundar> I noted that they are not blocking comments. We'll let the patch merge and update the details after that
03:16:44 <Yumeng_> I have a quick question about db schema, should we also change the type of device to enum instead of string?
03:17:35 <Li_Liu> Sundar, could you give it a +2 then?
03:17:55 <Sundar> Li_Liu: Sure. But it is already merged?
03:18:38 <Li_Liu> Sundar, i mean this one https://review.openstack.org/#/c/624138/
03:20:38 <Sundar> Li_Liu: Can we just address the comments that I did not mark as ignorable?
03:20:55 <Sundar> For example, the source file is called driver_handle.py instead of attach_handle.py
03:21:55 <Li_Liu> sure, I am just checking to see if the patch is ready to merge after the comments being addressed
03:22:24 <Li_Liu> that's cool
03:22:26 <Sundar> Yes, I think so. Just address the basics, and I'll give a +2
03:23:03 <Li_Liu> Looking at the stories at https://storyboard.openstack.org/#!/story/2004249
03:24:07 <Sundar> I think we should hold off on 27782.
03:24:37 <Li_Liu> why?
03:24:49 <Sundar> If we start adding stuff to placement now and change that after Stein, we have to do a reshaping of the provider tree and what not, which could be a pain for us, Nova or both
03:25:11 <Sundar> It is better to put that in the feature branch, get Nova to sign off, and then merge it
03:25:22 <Li_Liu> that one does not involve placement yet tho
03:25:39 <Li_Liu> it's just doing the update from the agent side
03:25:50 <Li_Liu> Hi Yikun
03:25:54 <Sundar> 27782: Conductor should update Placement
03:25:59 <yikun> hello
03:26:08 <yikun> soryy for late
03:26:20 <Sundar> Welcome, yikun. NP.
03:26:28 <yikun> Sundar: :)
03:26:50 <Li_Liu> Sundar, ah. my bad..  looked at a wrong one.
03:26:53 <Sundar> Good to have you join the team :)
03:27:25 <Sundar> Li_Liu: NP. The other 4 in that story are doable.
03:28:33 <Li_Liu> Sundar, can we borrow some code from your pilot branch for this?
03:28:46 <Li_Liu> I mean for  the update placement piece
03:29:52 <Li_Liu> or quickly patch it up to work with placement for some simple demo
03:29:53 <Sundar> Li_Liu: for Placement? I don;t have it. :) I am directly writing a fake device by calling placement API from a script, rather than have the conductor take discovered devices and update Placement
03:30:10 <Li_Liu> I see
03:30:30 <Sundar> Yes, once we have the Cyborg master branch code, I will fold that into the feature branch somehow for a full end-to-end flow
03:30:32 <Li_Liu> ok, let's just focus on the rest 4 stories
03:30:37 <Li_Liu> for now
03:31:32 <Li_Liu> Yumeng was asking "I have a quick question about db schema, should we also change the type of device to enum instead of string?"
03:31:39 <Li_Liu> Sundar, could you comment on that?
03:32:33 <Sundar> Yumeng_, Li_Liu: Sure, makes sense to me.
03:32:35 <Li_Liu> Hi Shaohe
03:32:38 <Sundar> However, ...
03:32:51 <shaohe_feng_> hi Li_Liu
03:33:26 <Sundar> There could be cards in the neat future that may involve multiple RCs. The device type wil translate into one of those RCs, e.g. CUSTOM_ACCELERATOR_FPGA. Others will be custom RCs
03:33:33 <Sundar> *near
03:34:35 <Sundar> yikun: I replied to your comments in https://review.openstack.org/#/c/630427/
03:34:49 <Yumeng_> Sundar, Li_Liu: Thx. Got it. Then I will do the change later.
03:34:50 <Li_Liu> in that case, string makes more sense
03:35:23 <yikun> Sundar: yep, I see, I was just a little wired why we should keep a arq obj in cyborg
03:35:24 <Li_Liu> Hi CoCo
03:36:03 <Coco_gao> Hi  all
03:36:04 <yikun> IMO, if it's something like pcireq, it should be an obj in nova.
03:36:10 <Coco_gao> #info Coco_gao
03:36:38 <Sundar> yikun: The ARQ is a combination of a Neutron port and vif. It is used for binding and attaching.
03:36:57 <Sundar> It is defined between Nova and Cybprg
03:37:06 <yikun> and the other thing, about ARQ or AcceleratorRequst, It's okay to me in db or obj level, but I think we should spell completely in API at least.
03:37:43 <Li_Liu> For external APIs, sure
03:38:45 <Sundar> Ok
03:39:00 <Coco_gao> Spell completely is more common in OpenStack Projects, I agree
03:39:29 <Kevin_Zheng> like in most projects, we only use abbreviations for well-kown concepts like IPs, DNS
03:40:00 <Kevin_Zheng> we don't use this kind of abbreviations for some thing like this
03:40:21 <Li_Liu> thanks for the comments
03:40:42 <Kevin_Zheng> like we don't call GET /AC in placement, we callGET /allocation_candidates
03:41:03 <Coco_gao> Thanks for all this advice
03:41:40 <Sundar> So, we are talking only of URLs, not method names?
03:41:41 <Li_Liu> Coco, you can address them in a future patch when you have time
03:41:51 <Kevin_Zheng> and ARQ is a well-known abbreviation for Automatic  Repeat-reQuest
03:42:05 <Kevin_Zheng> it could be really easy to get confused
03:42:12 <Li_Liu> lol
03:42:33 <yikun> :D
03:42:58 <Li_Liu> alright, move on
03:43:04 <Sundar> That stuff about Automatic Repeat-.... has been brought up before. but it is in embedded systems, etc., not in Openstack . Most people agreed it is not confusing
03:43:31 <Sundar> So, URLs will spell accelerator_request in full. Based on previous comments, db/object layer need not do that
03:43:48 <xinranwang> btw, can you guys have a look at https://review.openstack.org/#/c/630579/
03:44:10 <Sundar> APi layer code -- like get_obj_arq?
03:44:44 <xinranwang> let's get the object layer code merged as well.
03:45:02 <Li_Liu> xinranwang, you might wanna rebase it
03:46:02 <Li_Liu> #topic new member intro
03:46:08 <xinranwang> Li_Liu: yes sure
03:46:44 <Li_Liu> Yikun is my colleague in Huawei. He will be joining cyborg.
03:47:08 <Li_Liu> I just wanna make it more official, yikun
03:47:15 <Coco_gao> Welcome, Yikun
03:47:45 <Li_Liu> Do you wanna introduce yourself a little bit?
03:47:57 <yikun> thanks, and I will pay more attention on nova-cyborg intergration. :)
03:48:03 <xinranwang> welcome
03:48:29 <Li_Liu> great, that's exactly what we need :P
03:48:40 <Yumeng_> Yikun: welcome :)
03:48:57 <Li_Liu> #topic AoB
03:49:10 <yikun> Coco_gao xinranwang Yumeng_  thanks, :)
03:49:33 <Li_Liu> if there is no other stuff to discuss, let's call the meeting
03:49:44 <Li_Liu> have a good day/night guys
03:49:52 <Coco_gao> have a good day
03:49:58 <Coco_gao> bye
03:50:04 <Yumeng_> bye
03:50:11 <Li_Liu> bye
03:50:26 <Li_Liu> #endmeeting