| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2018-01-24 | |||
| 15:16:57 | zhipeng | okey moving on | |
| 15:17:07 | zhipeng | #info Intel FPGA driver | |
| 15:17:19 | zhipeng | #link https://review.openstack.org/531129 | |
| 15:17:54 | zhipeng | Li_Liu has shaohe_feng_ addressed your comments ? | |
| 15:18:57 | Li_Liu | I havn't had the chance to look at the new patch yet | |
| 15:19:05 | Li_Liu | Doing it now | |
| 15:19:22 | zhipeng | okey | |
| 15:19:46 | zhipeng | shaohe_feng_ other than the comments you received, code-wise is your patch ready ? | |
| 15:19:52 | Dutch | I can also add some input to help keep things vendor neutral | |
| 15:20:05 | zhipeng | Dutch absolutely | |
| 15:20:39 | shaohe_feng_ | zhipeng: I have test it on our platform, it can works. | |
| 15:21:06 | shaohe_feng_ | zhipeng: but maybe we still need to introduce rootwraper | |
| 15:21:07 | zhipeng | okey then here are my suggestions: | |
| 15:21:27 | zhipeng | no rootwrapper for queens :) | |
| 15:21:33 | zhipeng | we need to discuss that in Dublin | |
| 15:21:47 | zhipeng | So here are my suggestions at the patch : | |
| 15:22:02 | shaohe_feng_ | the program need root for both intel and xilinx | |
| 15:22:05 | zhipeng | again due to the tight release window, we will have to land this patch this weekend | |
| 15:22:17 | dolpher | Li_Liu, Dutch: current implementation in fpga/intel is specific to intel FPGA, for Xilinx, we may need to create a directory of fpga/Xilinx | |
| 15:22:44 | zhipeng | as I understand the comments are mostly on vendor neutral, as dolpher just mentioned | |
| 15:22:44 | Li_Liu | dolpher, I agree | |
| 15:23:00 | dolpher | we need to make the interface generic in the end. | |
| 15:23:06 | zhipeng | so I would suggest to change the patch accordingly to reflect that this is an Intel FPGA driver | |
| 15:23:15 | Li_Liu | only issue I have is in cyborg/accelerator/drivers/fpga/util.py | |
| 15:23:26 | zhipeng | and let's try to harmonize it in the long run | |
| 15:23:40 | zhipeng | poor gerrit | |
| 15:24:22 | shaohe_feng_ | gerrit need rest :) | |
| 15:25:27 | zhipeng | anyway, shaohe_feng_ you captured my suggestion ? | |
| 15:26:14 | shaohe_feng_ | yes | |
| 15:26:24 | zhipeng | great :) | |
| 15:26:29 | zhipeng | next item | |
| 15:26:45 | zhipeng | #info Justin leftovers: conductor interaction with Placement | |
| 15:26:51 | zhipeng | not sure if zhuli is here | |
| 15:27:32 | zhipeng | but as I have discussed with zhuli and Li_Liu offline previously, we will need to quick modify and copy-paste of nova-sched report.py | |
| 15:28:00 | zhipeng | and also an extension at the agent level to implement the discover function in full | |
| 15:28:22 | zhipeng | This is also another big hole we need to fix in a short time | |
| 15:28:36 | zhipeng | shaohe_feng_ can your team help on the agent side ? | |
| 15:29:07 | zhipeng | what we need is an equal counterpart of "resource_tracker.py" in cyborg | |
| 15:29:19 | shaohe_feng_ | OK | |
| 15:29:32 | zhipeng | which I assume is the discover interface that is very basic at the moment | |
| 15:29:57 | shaohe_feng_ | yes. | |
| 15:30:33 | zhipeng | okey thx :) zhuli and Li_Liu will coordinate on the conductor side | |
| 15:30:38 | zhipeng | moving on | |
| 15:30:40 | Li_Liu | sure | |
| 15:30:44 | zhipeng | #info fix on devstack | |
| 15:31:10 | shaohe_feng_ | resource_tracker.py will check resource periodically | |
| 15:32:12 | zhipeng | shaohe_feng_ we can simplify that for queens | |
| 15:32:23 | zhipeng | and write to db instead of a provider-tree like cache | |
| 15:32:33 | zhipeng | but anyways, back to the topic | |
| 15:32:43 | zhipeng | #link https://review.openstack.org/536251 | |
| 15:32:56 | zhipeng | another good patch from Jinghan to make devstack working | |
| 15:33:16 | zhipeng | crushil if you thinks it ok you could just land it :) | |
| 15:34:55 | crushil | zhipeng, Ok. Will look at it later today | |
| 15:35:15 | zhipeng | gr8t thx :) | |
| 15:35:50 | zhipeng | okey folks that is all we need to do for Queens, I see Nova is still panicking on landing patches, so we are not alone :) | |
| 15:36:05 | zhipeng | let's try to wrap it up before next Mon | |
| 15:36:18 | dolpher | :) | |
| 15:36:49 | zhipeng | release is always driving people insane | |
| 15:37:53 | zhipeng | #action crushil to update the generic driver patch, and work with Howard to land all the remaining patches | |
| 15:38:21 | crushil | I thought the release data was feb 4th? | |
| 15:38:24 | zhipeng | #action shaohe_feng_ to update the patch to reflect the patch is Intel FPGA driver | |
| 15:38:44 | zhipeng | crushil that is like hard cutof, we have to run through the CIs and other stuff | |
| 15:38:47 | zhipeng | docs ... | |
| 15:38:53 | crushil | Hmmm | |
| 15:39:08 | zhipeng | so I left out a week of time for more panicking stuff lol | |
| 15:39:11 | crushil | Ok, so are we cutting RC1, RC2, etc. | |
| 15:39:12 | crushil | ? | |
| 15:39:18 | zhipeng | not for this release | |
| 15:39:22 | crushil | Ok | |
| 15:39:29 | crushil | We need to start doing that from Rocky | |
| 15:39:30 | zhipeng | but for Rocky we are gonna use the MS1,MS2,... | |
| 15:39:37 | crushil | Ok, makes sense | |
| 15:39:42 | zhipeng | yes, so that things are not so in chaos | |
| 15:40:39 | zhipeng | #action zhuli Li_Liu and shaohe_feng_ works on the Placement interaction patch (report and discovery) | |
| 15:41:09 | crushil | I can help review all the patches | |
| 15:41:30 | zhipeng | crushil counting on you man lol | |
| 15:41:55 | zhipeng | okey now moving away from the crazy real world problem | |
| 15:42:07 | zhipeng | #topic Dublin Rocky PTG Planning | |
| 15:42:30 | crushil | zhipeng, I won't be there most probably | |
| 15:43:06 | zhipeng | crushil waaaaat ! | |
| 15:43:12 | zhipeng | #link https://etherpad.openstack.org/p/cyborg-ptg-rocky | |
| 15:43:29 | crushil | zhipeng, Ya. We have MWC the same week | |
| 15:43:36 | zhipeng | ah shit | |
| 15:43:44 | zhipeng | forget about MWC | |
| 15:44:14 | zhipeng | so Dutch Amol Vipparthy will you guys be present at Dublin ? | |
| 15:44:30 | Li_Liu | In case I can not work out the VISA.. do you guys have video conference in PTGs? | |
| 15:44:41 | zhipeng | Li_Liu I could set that up | |
| 15:44:43 | Dutch | I'll see if I can make it | |
| 15:45:18 | Li_Liu | ok | |
| 15:45:50 | zhipeng | so in any case, please propose the topic you would like to discuss on the etherpad | |
| 15:46:46 | zhipeng | other than the specific topics | |
| 15:47:01 | zhipeng | the important general topic we will cover is the planning for Rocky | |
| 15:47:29 | zhipeng | and the items are 1. finishing up Nova interaction bits 2. Burn fucntionality 3. Quota support | |
| 15:47:44 | zhipeng | these are the major general items I could think of for the next release | |
| 15:47:58 | zhipeng | any other suggestions ? | |
| 15:49:30 | shaohe_feng_ | does | |
| 15:50:17 | shaohe_feng_ | cyborg support multi tenant？ | |
| 15:51:02 | zhipeng | shaohe_feng_ that questions should be divides in two | |
| 15:51:28 | zhipeng | functionality wise, cyborg should definitely support multi-tenancy, hence the quota feature planning | |
| 15:51:55 | zhipeng | but client wise, I haven't see strong use case for a tenant consuming client necessary for cyborg now | |