*** takashin has joined #openstack-placement | 02:26 | |
*** e0ne has joined #openstack-placement | 03:47 | |
*** takashin has left #openstack-placement | 03:53 | |
*** e0ne has quit IRC | 04:49 | |
*** e0ne has joined #openstack-placement | 04:50 | |
*** e0ne has quit IRC | 04:51 | |
*** e0ne has joined #openstack-placement | 05:08 | |
*** e0ne has quit IRC | 05:16 | |
*** e0ne has joined #openstack-placement | 06:00 | |
*** e0ne has quit IRC | 06:03 | |
*** e0ne has joined #openstack-placement | 06:12 | |
*** e0ne has quit IRC | 06:20 | |
*** e0ne has joined #openstack-placement | 08:09 | |
*** cdent has joined #openstack-placement | 09:17 | |
openstackgerrit | Chris Dent proposed openstack/placement master: Add consumer_types migration, database and object changes https://review.opendev.org/669170 | 09:44 |
---|---|---|
openstackgerrit | Chris Dent proposed openstack/placement master: Microversion 1.37: API support for consumer types https://review.opendev.org/679441 | 09:44 |
openstackgerrit | Chris Dent proposed openstack/placement master: Switch ConsumerType to use an AttributeCache https://review.opendev.org/679486 | 09:44 |
cdent | maybe that will finally work | 09:44 |
*** artom has joined #openstack-placement | 10:13 | |
cdent | gibi or stephenfin : if you have a chance to look at this request logging fix up, that would be grand: https://review.opendev.org/#/c/676982/ | 10:25 |
gibi | cdent: ack | 10:28 |
cdent | thanks | 10:29 |
* cdent finds lunch | 10:50 | |
*** cdent has quit IRC | 10:50 | |
*** cdent has joined #openstack-placement | 11:26 | |
cdent | thanks gibi | 11:27 |
sean-k-mooney | cdent: by the way ill try to get to https://storyboard.openstack.org/#!/story/2006456 today | 11:33 |
cdent | great, thanks | 11:34 |
sean-k-mooney | my preference as i said last week woudl be HW_CPU_X86_INTEL_SST_* at least initally as COMPUTE_ implies t | 11:34 |
sean-k-mooney | somting else but i also need to read the comments | 11:35 |
cdent | there seems to be a bit of confusion over whether the trait is useful in isolation | 11:35 |
cdent | and if not, if the other bits are going to happe | 11:35 |
cdent | n | 11:35 |
sean-k-mooney | there is a use case with the provider.yaml | 11:37 |
sean-k-mooney | actully maybe not | 11:37 |
sean-k-mooney | the provider.yaml is currenly restrited to CUSTOM_ traits | 11:37 |
sean-k-mooney | with the cpu pinning work there will be two pools of cpus on the host | 11:37 |
* cdent nods | 11:37 | |
sean-k-mooney | the cpu_dedicated_set and cpu_share_set | 11:38 |
sean-k-mooney | the proposal was on host that support SST_BF | 11:38 |
sean-k-mooney | to use teh dedicated set for the highr priority cores | 11:38 |
sean-k-mooney | and the shared set for the rest | 11:38 |
sean-k-mooney | that would then allow you to request the traits + pinning | 11:38 |
sean-k-mooney | to land on a hos twih high priority cores | 11:38 |
sean-k-mooney | or the traits + shared to land on the low pirority cores | 11:39 |
sean-k-mooney | that said due to the CUSTOM_ constarit on provider.yaml you would have to use a non standard trait to do that | 11:40 |
sean-k-mooney | so in its current form i guess this would only be useful if the virt dirver(e.g. libvirt) reported supprot for this standard trait | 11:41 |
sean-k-mooney | efried: does ^ makes sense to you? | 11:41 |
*** cdent has quit IRC | 12:50 | |
*** david-lyle has quit IRC | 13:02 | |
*** dklyle has joined #openstack-placement | 13:02 | |
*** spatel has joined #openstack-placement | 13:10 | |
*** cdent has joined #openstack-placement | 13:15 | |
*** mriedem has joined #openstack-placement | 13:29 | |
efried | sean-k-mooney: even if we allowed standard traits in provider config, the part that doesn't make sense is: | 13:34 |
efried | => one trait would indicate the system is capable of SST-BF | 13:35 |
efried | => a different trait indicates that we've taken SST-BF + PCPU/VCPU and made PCPUs high-priority. | 13:35 |
efried | => The latter requires the former | 13:35 |
efried | => There's no use case for the former in isolation. | 13:35 |
sean-k-mooney | well we dont actully want to supprot the latter | 13:36 |
sean-k-mooney | the pcpu/vcpu trick is a hack for no formal support | 13:36 |
sean-k-mooney | it was only intended for use until there as actully support nova/cyborg/rmd for SST-BF | 13:37 |
sean-k-mooney | so the former haing a trait to indicate hardware support is the only trait that i think makes sense | 13:37 |
efried | Oh, I did not get the impression that there would ever be native support for CPU priority. | 13:39 |
efried | in any case, just the SST-BF capability would never be enough to do anything with. | 13:39 |
efried | You would always have to say *something* else indicating how the frequencies are to be used. | 13:40 |
cdent | dansmith: actually, I have a real request for you. you're still grenade-oriented, yes? this seems to be working, but I'm not fullyy confident it is _right_: https://review.opendev.org/679655 | 14:10 |
dansmith | cdent: I'm three deep right now.. remind me in 60 minutes? | 14:11 |
cdent | dansmith: can do america | 14:12 |
openstackgerrit | Merged openstack/placement master: Merge request log and request id middlewares https://review.opendev.org/676982 | 14:20 |
openstackgerrit | Merged openstack/placement master: Clarify the NOTE associated with ordering of middleware https://review.opendev.org/673335 | 14:20 |
*** cdent has quit IRC | 14:26 | |
*** spatel has quit IRC | 14:41 | |
*** spatel has joined #openstack-placement | 14:50 | |
*** spatel has quit IRC | 15:13 | |
*** cdent has joined #openstack-placement | 15:25 | |
cdent | dansmith: is it is 1 18 minutes later, here is your late reminder | 15:28 |
cdent | s/1/1h/ | 15:28 |
dansmith | cdent: yeah I came looking for you... I can't look at that and say for sure it's right without looking at some of the other things that are totally top-level like you need to be | 15:28 |
dansmith | so i think dean probably needs to have a look, | 15:28 |
cdent | i copied it from keystone, and then fixed what was broken, so I reckon it is fairly okayish | 15:29 |
dansmith | but in general, i think you're doing the right thing, I just don't have enough of the bootstrap stuff of other projects in my head to say for sure without digging really deep (which I imagine you've done) | 15:29 |
dansmith | yeah | 15:29 |
dansmith | figured | 15:29 |
cdent | it was inspired by the needed-by, which now works as a result | 15:29 |
cdent | cool. at least nothing obvious leaps out as dumb. thank you. | 15:30 |
*** e0ne has quit IRC | 16:00 | |
*** e0ne has joined #openstack-placement | 17:12 | |
*** e0ne has quit IRC | 17:23 | |
*** e0ne has joined #openstack-placement | 17:23 | |
*** e0ne has quit IRC | 17:30 | |
*** artom has quit IRC | 17:31 | |
*** artom has joined #openstack-placement | 17:31 | |
*** cdent has quit IRC | 17:38 | |
*** e0ne has joined #openstack-placement | 18:15 | |
*** e0ne_ has joined #openstack-placement | 19:24 | |
mriedem | melwitt: any chance you can update https://review.opendev.org/#/c/675520/ this week so we can try to get it in the next release for train? | 19:25 |
*** e0ne_ has quit IRC | 19:26 | |
*** e0ne has quit IRC | 19:27 | |
melwitt | mriedem: yes, will update today, thanks | 20:12 |
*** artom has quit IRC | 20:30 | |
*** artom has joined #openstack-placement | 20:36 | |
*** mriedem has quit IRC | 22:10 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!