| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2018-05-16 | |||
| 14:28:25 | shaohe_feng | hopeful next time, they can join. | |
| 14:28:40 | Li_Liu | no response.. | |
| 14:28:54 | Li_Liu | is Dutch still the contact? | |
| 14:29:20 | shaohe_feng | zhipeng, http://eavesdrop.openstack.org/meetings/openstack_cyborg_driver/2018/openstack_cyborg_driver.2018-05-14-14.01.log.html | |
| 14:29:25 | shaohe_feng | it is here. | |
| 14:29:32 | zhipeng | thx man :) | |
| 14:29:45 | zhipeng | Li_Liu no he left Xilinx | |
| 14:29:51 | zhipeng | Chuck Song is the contact | |
| 14:30:02 | Li_Liu | hmm. ok | |
| 14:32:39 | Sundar | FYI, I am readying a patch for driver-agent API update. It would be good to know who is working on what area | |
| 14:32:42 | zhipeng | Li_Liu any update on the doc side ? | |
| 14:33:09 | zhipeng | SUndar i think lenovo team could help on that | |
| 14:33:15 | Li_Liu | zhipeng, not yet, I will start org it after the summit | |
| 14:33:54 | zhipeng | okey | |
| 14:34:56 | zhipeng | #topic critical rocky spec review | |
| 14:38:22 | zhipeng | #info python-cyborgclient update | |
| 14:38:25 | zhipeng | #link https://review.openstack.org/#/c/565023/ | |
| 14:38:44 | zhipeng | and #link https://review.openstack.org/568799 | |
| 14:38:50 | shaohe_feng | zhipeng, as we discussed | |
| 14:38:57 | shaohe_feng | I have change the command to: | |
| 14:38:57 | zhipeng | this took large chunk of time last meeting | |
| 14:39:08 | zhipeng | so plz review it and if ok let's land it now | |
| 14:39:13 | shaohe_feng | openstack accelerator list | |
| 14:39:20 | shaohe_feng | ^ Sundar | |
| 14:40:03 | shaohe_feng | and change the service type to accelerator | |
| 14:40:37 | Sundar | shaohe: Sure. I did +1 just now | |
| 14:41:40 | shaohe_feng | thanks | |
| 14:41:48 | Li_Liu | I will give a +2 | |
| 14:42:36 | shaohe_feng | we can add more command base on this patch | |
| 14:42:50 | shaohe_feng | then users can easy to use cyborg | |
| 14:42:57 | zhipeng | yes :) | |
| 14:43:01 | zhipeng | this is a big help | |
| 14:43:15 | Li_Liu | been waiting for this. | |
| 14:43:19 | Li_Liu | this is great | |
| 14:43:25 | Sundar | shaohe: https://review.openstack.org/#/c/565023/ looks good. There are still some references to deprecation. I'll try to understand that and post comments. | |
| 14:43:50 | Sundar | May be we can remove the deprecated stuff later too | |
| 14:44:08 | shaohe_feng | yes. let it work first. | |
| 14:44:14 | shaohe_feng | then others can works on it. | |
| 14:44:27 | shaohe_feng | it is a base patch for other commands. | |
| 14:44:53 | shaohe_feng | I notice NokMikeR need cyborg client. | |
| 14:45:09 | zhipeng | okey :) | |
| 14:45:52 | zhipeng | #info quota specs | |
| 14:46:04 | zhipeng | xinran__ do you have any further updates ? | |
| 14:48:06 | xinran__ | I have update the spec | |
| 14:49:10 | xinran__ | work | |
| 14:49:10 | xinran__ | Now I think quota usage can | |
| 14:49:10 | xinran__ | And also submitted the patch | |
| 14:49:24 | xinran__ | When users create or delete one accelerator quota usage will also be updated | |
| 14:50:04 | shaohe_feng | great. | |
| 14:50:05 | zhipeng | sounds great | |
| 14:51:25 | xinran__ | Need more review now | |
| 14:53:10 | Sundar | OK, will review it soon. | |
| 14:53:42 | xinran__ | Sundar: thanks | |
| 14:54:16 | shaohe_feng | thanks xinran__ , will review it. | |
| 14:54:26 | zhipeng | #info fpga programming spec | |
| 14:54:30 | Sundar | We need to understand what it means to reserve an accelerator in Cyborg without informing Nova | |
| 14:54:38 | zhipeng | Li_Liu you think this is in good shape ? | |
| 14:55:16 | Li_Liu | I think so | |
| 14:55:28 | Li_Liu | Sundar, do you have further comments? | |
| 14:55:39 | Li_Liu | I addressed your comments in my last patch | |
| 14:55:46 | Sundar | Yea, what does it mean for the driver to reserve an accelerator? | |
| 14:56:30 | Sundar | I will post comments on the patch | |
| 14:56:33 | zhipeng | Sundar prepare for future usage ? | |
| 14:57:24 | Li_Liu | are you talking about the programming spec? | |
| 14:58:15 | Sundar | zhipeng: I will post comments after thinking some more | |
| 14:58:29 | Sundar | Li_Liu: I think we are talking of Xinran's quota spec | |
| 14:58:55 | xinran__ | Sundar: you mean quota reserve or ... | |
| 14:59:05 | zhipeng | Sundar regarding the fpga programming spec ? | |
| 14:59:26 | shaohe_feng | :) | |
| 14:59:50 | Sundar | xinran: yes, in your patch: self._driver.reserve(...) | |
| 15:00:19 | Sundar | zhipeng: I will review all the specs and post comments :) | |
| 15:00:35 | zhipeng | okey :) | |
| 15:01:05 | shaohe_feng | I think admin can set the reserve. if reserve, cyborg should not report to placement? | |
| 15:01:53 | edleafe | shaohe_feng: you can report total=1 and reserved=1 | |
| 15:02:10 | edleafe | at least with the bug fix that's currently being worked on | |
| 15:02:25 | shaohe_feng | edleafe, got it. thanks. | |
| 15:02:46 | shaohe_feng | edleafe, which bug? | |
| 15:03:10 | xinran__ | Sundar: we need to reserve quota usage first, check delta+current usage < 0 or not, but quota usage will not be updated at this time. and after create accelerator successfully, we need call commit to update quota usage. | |
| 15:03:33 | edleafe | shaohe_feng: https://review.openstack.org/#/c/568613/ | |
| 15:03:59 | edleafe | shaohe_feng: it's still being debated whether this is a bug fix or a behavior change :) | |
| 15:04:11 | xinran__ | Did I get your point? | |
| 15:05:33 | shaohe_feng | edleafe, thanks. | |
| 15:06:56 | Sundar | xinran: As edleafe said, the better way is to increment the reserved count in Placement. Not all drivers may support the concept of reservation. | |
| 15:09:58 | zhipeng | #info cyborg-nova interaction | |
| 15:10:10 | zhipeng | Sundar have you address ed's comment on the spec ? | |
| 15:10:32 | xinran__ | Sundar: okay let’s discuss after meeting | |
| 15:10:37 | Sundar | zhipeng, I will do it today. | |
| 15:10:39 | Li_Liu | have to drop. see you guys in the summit :) | |
| 15:10:57 | zhipeng | okey :) | |
| 15:11:18 | Sundar | I think there is value in having one RC for FPGAs, one for GPUS, etc. I need to articulate it in the response. :) | |
| 15:11:19 | shaohe_feng | Sundar, Do you have changed to multi resource class? | |
| 15:11:22 | zhipeng | so let's try to review as many as possible by the end of the week | |
| 15:11:42 | Sundar | shaohe: yes, the spec has been updated to have one RC for each device type (GPU, FPGA, etc.) | |
| 15:13:19 | shaohe_feng | good. Thanks. | |
| 15:13:58 | zhipeng | #topic AoB | |
| 15:14:04 | zhipeng | any other business :) | |
| 15:15:55 | zhipeng | okey let's conclude the meeting now, thx everyone for showing up :) | |
| 15:15:59 | zhipeng | safe travels | |
| 15:16:02 | zhipeng | #endmeeting | |
| 15:16:05 | openstack | Meeting ended Wed May 16 15:16:02 2018 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | |