| s3rj1k | sean-k-mooney: Hi, where you able to tackle that change you proposed about distributed discovery? | 10:40 |
|---|---|---|
| sean-k-mooney | i didnt actully start on a real implemantation beyond the psudo code we discussed. i dont think it would be particallary hard to do im just not sure ill have time to do it | 10:43 |
| s3rj1k | I am interested to test it out, just in case :) | 10:43 |
| sean-k-mooney | if i can find an hour or two this week i may throw up a poc and you can let me know if it works for you, im going on pto for the rest of the year wednesday next week so there are a few things i need to do before then | 10:45 |
| s3rj1k | sounds good, thanks | 10:46 |
| opendevreview | Balazs Gibizer proposed openstack/placement master: WIP: Round-robin candidate generation https://review.opendev.org/c/openstack/placement/+/936832 | 13:05 |
| opendevreview | Merged openstack/nova master: Deprecate [wsgi] secure_proxy_ssl_header https://review.opendev.org/c/openstack/nova/+/817927 | 13:14 |
| *** haleyb|out is now known as haleyb | 14:03 | |
| sean-k-mooney | bauzas: Uggla: comments left on https://review.opendev.org/c/openstack/nova-specs/+/936407, im kind of furstrated that many of the things i requested durign the ptg dicsssion where not done. with that said reading back over the ptg note we did not recored all the thigns we agreed and some of the things that were "agreed" are wrong so i can understand where some of the | 14:36 |
| sean-k-mooney | confution comes form | 14:36 |
| sean-k-mooney | if i recall correctly i joined that topic about 10 mins in and wehn i did the direction changed significanlty | 14:37 |
| sean-k-mooney | it looks like we didnt remove the previosu agrrements that happend before i was involved | 14:37 |
| bauzas | sean-k-mooney: thanks for your comments, as a reminder, this was a WIP but once the spec is updated, we'll tell you | 15:02 |
| bauzas | sean-k-mooney: if you think we missed a specific action item on the PTG etherpad, please tell me | 15:02 |
| sean-k-mooney | there are 3 points that seam to have been missed. 1.) i asked that the spec not be GPU specific, 2 that the live migration and general supprot be called out spereatly in tow differnt sections, 3.) that pci in placment should be optional | 15:04 |
| opendevreview | Merged openstack/nova master: api: Add response body schemas for host aggregate APIs https://review.opendev.org/c/openstack/nova/+/924585 | 15:04 |
| sean-k-mooney | the spec does not yet cover the live mgiration changes, yet the usecase refence it so i was expecting the current draft to cover the non live migration usecases/constrains and then for the live migration aspect to be added leater | 15:05 |
| sean-k-mooney | im fine with discussing the generic enabling and live migration at the same time but given i expect those to be delvier potentiall in diffent relases i want to make sure we are clear on what is required for each case | 15:07 |
| sean-k-mooney | i would hope we can do both in epoxy | 15:07 |
| sean-k-mooney | but i dont want to require both to merge anything | 15:07 |
| sean-k-mooney | so what im hopign to see is a clear plan for how to supprot pci passthough with vendor specific vfio-pci variant drievers, and addtionally how to support live migratable VF when supproted by the vendor | 15:09 |
| bauzas | about 1), this was a mistake from rené, the trait was needing CUSTOM_ and it was just an example | 15:14 |
| bauzas | 2/ agree with you | 15:14 |
| bauzas | 3/ agree with you as well, but we definitely need to test the POC before, in order to be sure we're not missing anything | 15:15 |
| opendevreview | Merged openstack/nova master: api: Add response body schemas for extensions API https://review.opendev.org/c/openstack/nova/+/924586 | 15:31 |
| sean-k-mooney | bauzas: well using CUSTOM_ woudl still be wrong as we use Resouce Classes to modle the device vendor id and product id when using pci in placment | 17:20 |
| sean-k-mooney | so "traits": "CUSTOM_NVIDIA_A16_16A" woudl still be incorect in that case | 17:20 |
| sean-k-mooney | CUSTOM_MIGRATABLE is more accpable but we can standarize that trait as its generic | 17:21 |
| sean-k-mooney | if you have differend VFs with the same vendor_id and product_id bug are configured for diffent nvida vgpu types then you should use spereate resouce classes to modle that not traits if we want this to be compatiable with how pci in placmeent is intended to work | 17:23 |
| -opendevstatus- NOTICE: Gerrit will have a short outage while we update to the latest 3.9 release in preparation for our 3.10 upgrade on Friday | 21:31 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!