03:05:38 <Li_Liu> #startmeeting openstack-cyborg 03:05:39 <openstack> Meeting started Wed Dec 12 03:05:38 2018 UTC and is due to finish in 60 minutes. The chair is Li_Liu. Information about MeetBot at http://wiki.debian.org/MeetBot. 03:05:40 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 03:05:42 <openstack> The meeting name has been set to 'openstack_cyborg' 03:05:48 <Li_Liu> #topic Roll Call 03:05:56 <Li_Liu> #info Li_Liu 03:06:03 <Li_Liu> Let's get started 03:06:10 <xinranwang> #info xinranwang 03:06:14 <Sundar> #info Sundar 03:06:20 <Li_Liu> I hope the new time slot works better for most of the people 03:07:08 <Li_Liu> #topic status updates 03:07:23 <Yumeng_> # info Yumeng_ 03:07:30 <Li_Liu> Sundar, what's new on your side? 03:08:35 <Sundar> The feature branches for CYborg and its client have been created. Will be submitting code for a prorotype that shows how Nova could call into CYborg 03:09:13 <Li_Liu> that's great. So would help push the Nova-Cyborg interaction spec right? 03:09:16 <Sundar> This is intended to show developers how the interaction in the spec will play out in practice 03:09:22 <Sundar> I hope so :) 03:10:06 <Li_Liu> Once created, let's find some time to talk with Nova folks directly 03:10:11 <Li_Liu> let's push together 03:10:42 <Li_Liu> xinrangwang, anything you wanna share? 03:10:55 <Sundar> I also need to develop and push the Nova patches that do the actual invocation 03:10:56 <Li_Liu> https://review.openstack.org/#/c/624340/ 03:11:44 <Li_Liu> xinranwang, is that one ready to be reviewed? 03:12:24 <Li_Liu> Sundar, maybe we can show them how it should work first? 03:12:26 <xinranwang> Li_Liu: nope, it depends on the DB schemas the ovo object 03:13:03 <xinranwang> it's just the initial version 03:13:31 <Li_Liu> xinranwang, the spec should be ready anytime, does it also depends on Coco/zhenghao's work? 03:14:39 <Sundar> Li_Liu: Yes, the pilot code for Cyborg and its client will be submitted first. 03:15:16 <Li_Liu> great 03:16:24 <xinranwang> yes, because the input param for conductor is ovo object 03:16:41 <Li_Liu> Sundar, I think https://review.openstack.org/#/c/615462/ can be merged 03:17:03 <Li_Liu> xinranwang, ok\ 03:17:08 <xinranwang> Li_Liu: Btw, I have submitted our POC code with workflow -1 03:17:49 <Li_Liu> xinranwang, right, understand 03:17:58 <Li_Liu> thanks a lot for the work 03:18:14 <Sundar> Li_Liu: The db spec has progressed a lot. Only a couple of points left, which you wanted to address offline. I replied by email about half hour ago. 03:18:52 <Coco_gao> I received you email 03:19:00 <Li_Liu> Sundar, I saw that one, I think i address them in my patch already 03:19:01 <Yumeng_> Sundar, for tasks in Client and DB https://storyboard.openstack.org/#!/story/2004248. which one has a higher priority currently? 03:19:13 <Coco_gao> I will update the OVO object patch as well as submmit New DB patch. 03:19:34 <Li_Liu> Coco_gao, are you gonna work on the DB object based on the new db spec? 03:20:42 <xinranwang> Coco_gao: I can take over the diff part in conductor if you don't have enough time :) 03:20:59 <Coco_gao> Yes. 03:21:01 <Coco_gao> Li 03:21:06 <Sundar> The table currently named ARQ actually has ExtARQ data, since it includes Cyborg-private fields. That's why I said the table is better named as ExtARQ. ExtARQ is a superset of ARQ 03:21:26 <Li_Liu> that seems a lot of work, you might wanna offload them 03:22:01 <Coco_gao> xinran, I have the diff function in the old ovo patch, so it won't take long if the db object is ready. 03:22:26 <Li_Liu> Sundar, you mean there will not be a ARQ table then. only ExtARQ 03:22:51 <Sundar> Li_Liu: Yes 03:22:56 <Coco_gao> Li, I think the ExtARQ combines old ARQ and ExtARQ 03:23:02 <Li_Liu> ok, I will make the change right now 03:23:11 <Sundar> Li_Liu: Thanks 03:24:20 <Sundar> For the question of whether attach handles go with deployables or devices, one could argue both ways but I thought it is easier if it goes at the device level. if we want to keep it in the current form, it may require an upgrade later. As long we are all ok with that, I am fine 03:24:24 <Coco_gao> Li_Liu, I will implement table device and deployable and attribute. Other DB work might owned by zhenghao or anyone else? 03:24:57 <Sundar> Coco_gao: It may make sense to split the db work, since there are so many. Thanks for suggesting that. 03:24:59 <Li_Liu> Coco_gao, I will talk with zheng hao and xinran for the work load 03:25:07 <Yumeng_> coco I have can help 03:25:22 <Coco_gao> OK, you can discuss with zhenghao about other DB works. 03:25:31 <xinranwang> Coco_gao: ok, this patch is an initial version of conductor update, https://review.openstack.org/#/c/624340/, plz review it to see if it fits your requirements. It need more functions to be implemented. 03:25:31 <Coco_gao> Thanks, Yumeng_ 03:25:43 <Li_Liu> Sundar, I kinda lean towards let attachHandle reference to Device now 03:25:51 <Coco_gao> #info Coco_gao 03:26:42 <Sundar> Li_Liu: OK :) It is all a bit hypothetical because the devices in Stein train frame will work fine either way. 03:27:38 <Sundar> BTW, even if we put in device level, we have to figure out how to handle a PCI card wih 2 FPGAs, if that ever becomes a thing 03:27:54 <Li_Liu> sure 03:28:32 <Coco_gao> xinranwang, I will check your patch, thank you. 03:28:37 <Li_Liu> ok, I think it's great we can finalize the spec and can push things further 03:29:28 <Li_Liu> #topic AoB 03:29:57 <Li_Liu> On top of what we have discussed, do you guys have anything to add? 03:30:11 <xinranwang> Sundar: do you have any concern about the POC code I put in the community? 03:30:51 <Coco_gao> Sundar, You mean that cyborg client code is ready? 03:31:04 <Sundar> xinranwang: My main concern is this is an outdated poc, and it may cause confusion. 03:31:35 <Sundar> Coco_gao: I have made some changes to Cyborg client too. I will put that on a feature branch too. But it is still WIP 03:31:35 <Coco_gao> or we still need other one to implement that? 03:32:38 <xinranwang> Sundar: I understand your concern, it should be clarified to nova that these patches is for reference only 03:32:52 <Sundar> BTW, clarification: The prototype is not the final code. There will be many TODOs. Our developers are free to re-implement, or fill the gaps while leveraging the prototype 03:33:34 <Coco_gao> Yes , I think so. 03:34:03 <Sundar> Coco_gao, Zhenghao and possibly Yumeng_ are working on the db. Xinran is working on the agent/conductor API and the diff. Is that correct? 03:34:47 <Li_Liu> Sundar I think so 03:35:10 <Yumeng_> Sundar: yes, next time please also email me(bao.yumeng@zte.com.cn) if there is any update about the specificaiton :) 03:35:30 <Coco_gao> Or add yumeng to the reviewer. 03:35:33 <Sundar> Yumeng_: Sure :) Thanks for helping! 03:35:40 <Li_Liu> We did not have you before? Yumeng_ 03:35:48 <Li_Liu> We will in the future 03:36:12 <Coco_gao> That's great 03:36:13 <Yumeng_> not yet before. great! :) 03:36:19 <Li_Liu> :) 03:36:32 <Li_Liu> ok, I think we can wrap up for today 03:36:39 <Sundar> We have tons more things to do. Like new v2 APIs, checking that v1 APIs srtill work, adding tempest tests, .... 03:36:58 <Li_Liu> oh yes, gonna be a fun 2019 03:37:02 <Sundar> Can we identify owners for that? I propose that Xinran and I can work on the APIs? 03:37:12 <Li_Liu> sure 03:37:27 <Sundar> xinranwang: I think you have worked with pecan layer before. Are you fine with that? 03:38:00 <Coco_gao> OK, we can identify the owner now. 03:39:10 <Sundar> Do we have some developers to check rolling upgrades from Rocky to Stein? The v1 APIs should return null result in Stein, I think 03:39:23 <Sundar> For e.g. list deployables will return [] 03:41:52 <Sundar> Silence ... Shall we take this up next week? 03:42:02 <Yumeng_> Sundar: did you notice this patch https://review.openstack.org/#/c/619997/ 03:42:15 <Coco_gao> I can list deployables in my previous POC. 03:42:26 <Li_Liu> Sundar, I guess we can come up with a complete list first 03:42:31 <xinranwang> Sundar: Yes, I can work with you on Cyborg-Nova API side 03:43:00 <Coco_gao> Have you got authenticated before asking. 03:44:10 <Sundar> Yumeng_: We are dropping the accelerators table in Stein. So this API should be modified to return [] all the time in Stein? 03:44:20 <Yumeng_> I made this change in v1 APIs to avoid returns null for the situation where we actually have accelerators 03:45:03 <Sundar> Yumeng_: is anybody using Cyborg today, and therefore relying on tv1 he API? 03:46:23 <Sundar> *the v1 API 03:46:49 <Li_Liu> I don't think anyone is using cyborg 03:46:53 <Sundar> Li_Liu: OK, we can form that task list by next meeting 03:47:19 <Li_Liu> Sundar, we can talk on zoom in case you need it 03:48:11 <Sundar> Sure 03:48:11 <Coco_gao> I can get the deployable list, I also have the test code if you need that. 03:48:32 <Yumeng_> Sundar,Li liu: ok, maybe next meeting. I will dig more to discuss with you guys. 03:48:35 <Coco_gao> Anyway, let's arrange other meething. 03:49:39 <Li_Liu> ok 03:50:11 <Li_Liu> can we wrap up? I think it's almost lunch time for china folks :P 03:50:43 <Coco_gao> Yes, It's time for lunch, bye. 03:50:50 <Li_Liu> #endmeeting