| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-08-23 | |||
| 15:17:46 | sean-k-mooney | i obvioulsy missed somthing while debuging my devstack setup | |
| 15:17:54 | edmondsw | efried you can probably blame me there | |
| 15:18:21 | edmondsw | I don't know that either of us were suggesting it should be done, just wondering | |
| 15:18:25 | sean-k-mooney | edmondsw: placement does not store metrics no | |
| 15:18:33 | sean-k-mooney | edmondsw: also it likely never will | |
| 15:18:47 | edmondsw | I don't really know why nova ever used metrics for scheduling | |
| 15:18:55 | mriedem | it was an ibm feature | |
| 15:18:58 | edmondsw | ha | |
| 15:19:02 | mriedem | you probably should remmeber this | |
| 15:19:06 | mriedem | i'm pretty sure it was powervc relatd | |
| 15:19:08 | mriedem | *related | |
| 15:19:13 | mriedem | sudipta biswas | |
| 15:19:14 | sean-k-mooney | edmondsw: but bandwidth based schduling is different then metric based suchduling that was keed off network usage | |
| 15:19:21 | mriedem | it was for metrics weighers | |
| 15:19:57 | jaypipes | hold up... somebody said metrics and placement in the same breath... /me perks up | |
| 15:20:09 | sean-k-mooney | edmondsw: for gibi's bandwith based schduling work we are tracking bandwidth as a consumable resouce which is exactly what placement was designed for | |
| 15:20:24 | edmondsw | sean-k-mooney yep, understood | |
| 15:20:39 | sean-k-mooney | jaypipes: yep the comment i made was that was not a thing | |
| 15:20:54 | gibi | ohh I see the relation between nw bw and metrics. If neutron would report physically used nw bw to placement that would be the metric | |
| 15:20:59 | efried | jaypipes: I'm going to tag providers with dynamic traits like CUSTOM_CPU_CYCLES_SINCE_LAST_RESET_<NNNN>, updated every periodic via update_provider_tree. | |
| 15:21:38 | sean-k-mooney | gibi: ya but we dont want to contiunally update allocation to model real usage. at would be an abuse of the datamodel in placement | |
| 15:21:48 | gibi | sean-k-mooney: totally agree | |
| 15:22:03 | gibi | sean-k-mooney: I never wanted to do that | |
| 15:23:08 | mriedem | do they sell those? | |
| 15:23:10 | sean-k-mooney | gibi: i have seen a really old poc for bandwidth based schduling using celomiter metrics in the past | |
| 15:23:21 | efried | I was expecting something along the lines of "over my cooling corpse" | |
| 15:23:44 | jaypipes | mriedem: https://www.amazon.com/Posey-Straight-Jacket-MEDIUM/dp/B001ANOR6E | |
| 15:23:44 | efried | mriedem: https://www.amazon.com/Freebily-Leather-Straitjacket-Straight-Costume/dp/B07B8KV2NQ/ref=sr_1_2?ie=UTF8&qid=1535037811&sr=8-2&keywords=straightjacket | |
| 15:23:50 | jaypipes | efried: :P | |
| 15:24:02 | sean-k-mooney | efried: what you do in the privacy of your out of tree virt driver is fine just so long as that never happens in the nova tree :) | |
| 15:24:17 | jaypipes | "Customers also bought..." section is awesome. | |
| 15:24:39 | mriedem | i suggested to my wife last night that while in berlin we should shop for something like that | |
| 15:24:48 | jaypipes | heh | |
| 15:24:49 | mriedem | b/c my german stereotype includes 80s synth music and leather | |
| 15:25:08 | efried | I like how it's other-legit-restraint other-legit-restraint kinky-s&m-thing other-legit-restraint... | |
| 15:31:08 | openstackgerrit | Chen proposed openstack/nova master: Fix create_resource_provider docstring https://review.openstack.org/595453 | |
| 15:43:46 | mriedem | so anyone that cares about this placement extract / governance thing should probably be in the tc channel as of about an hour ago | |
| 16:08:29 | openstackgerrit | Merged openstack/nova master: placement: use single-shot INSERT/DELETE agg https://review.openstack.org/592654 | |
| 16:32:24 | sean-k-mooney | mriedem: am im going to drop for dinner. i might be online later. was there anything else you wanted me to take a look at for RC3? | |
| 16:35:29 | mriedem | sean-k-mooney: no i think we're good, thanks again | |
| 16:35:54 | sean-k-mooney | mriedem: no worries. | |
| 16:35:56 | sean-k-mooney | o/ | |
| 16:55:37 | mriedem | melwitt: questions in https://review.openstack.org/#/c/595455/ | |
| 16:56:19 | melwitt | looking | |
| 17:02:26 | melwitt | mriedem: replied. I think you answered my question though, I need to call out cells v1 in both places | |
| 17:02:58 | melwitt | I was worried about people confusing it with cells v2. maybe I'll say cells v1, legacy cells | |
| 17:04:57 | openstackgerrit | Dan Smith proposed openstack/nova master: Make instance_list perform per-cell batching https://review.openstack.org/593131 | |
| 17:04:58 | openstackgerrit | Dan Smith proposed openstack/nova master: Record cell success/failure/timeout in CrossCellLister https://review.openstack.org/594265 | |
| 17:04:59 | openstackgerrit | Dan Smith proposed openstack/nova master: Optimize global marker re-lookup in multi_cell_list https://review.openstack.org/594577 | |
| 17:13:59 | mriedem | at this point, anyone using cells v1 knows what we mean when we say v1 | |
| 17:14:21 | mriedem | https://docs.openstack.org/nova/latest/user/cells.html | |
| 17:14:29 | mriedem | for people that don't know the difference, ^ i hope they can read | |
| 17:16:26 | melwitt | maybe I'll link to that too | |
| 17:16:56 | melwitt | well yeah, I was worried about the inverse. non cells v1 people thinking they have cells v1 because they have "cells" | |
| 17:24:31 | mriedem | i think specifically calling out cells "v1" makes that clear | |
| 17:24:45 | mriedem | if i don't know, i'm likely going to ask, "there are multiple versions?" | |
| 17:35:59 | mriedem | jaypipes: i just want to say, before i go "meet the teacher", that if you hadn't birthed placement from your loins it wouldn't exist, so, you know, thanks and stuff | |
| 17:36:07 | mriedem | that seems to be lost a bit in this | |
| 17:37:11 | jaypipes | mriedem_away: there's lots of contributors to placement. but thank you. | |
| 17:37:56 | openstackgerrit | Ken'ichi Ohmichi proposed openstack/nova master: tests: Create functional libvirt test base class https://review.openstack.org/407055 | |
| 18:29:27 | mriedem | melwitt: so i'm not strong on this vif type / virtio stuff in https://review.openstack.org/#/c/595592/ | |
| 18:29:51 | mriedem | i can take a crack at a reno, which based on sahid's comment would be, "rx/tx queues only work with vif_model=virtio" | |
| 18:30:21 | mriedem | which is (1) configurable per image with the hw_vif_model property or globally in nova.conf via the (2) [libvirt]/use_virtio_for_bridges option | |
| 18:31:20 | melwitt | mriedem: ack. did stephenfin give any ideas too or? | |
| 18:31:21 | mriedem | but without someone around that understands this to actually ack my wording it's tough | |
| 18:31:29 | mriedem | not in the patch | |
| 18:31:32 | mriedem | or ic | |
| 18:31:33 | mriedem | *irc | |
| 18:31:34 | mriedem | so no | |
| 18:32:04 | melwitt | dangit | |
| 18:33:52 | dansmith | mriedem: I think that's legit | |
| 18:34:09 | dansmith | setting the queues only works with certain versions of qemu, libvirt, and on certain vif types | |
| 18:34:11 | mriedem | from moshe's paste, | |
| 18:34:22 | mriedem | he's not using virtio vif model b/c it's not in the interface xml | |
| 18:34:35 | mriedem | so he must have use_virtio_for_bridges=False in nova.conf | |
| 18:34:49 | dansmith | I dunno about the interacton with that config though | |
| 18:35:08 | mriedem | getting things to work is basically the hellraiser cube | |
| 18:36:30 | mriedem | the release note for the feature says, "Only supported for vhost/vhostuser interfaces" | |
| 18:36:58 | mriedem | so, | |
| 18:37:09 | mriedem | i think we're probably covered enough here | |
| 18:37:17 | mriedem | i can't really wordsmith this better than what's already in the reno | |
| 18:37:20 | mriedem | or config option help | |
| 18:37:48 | mriedem | and given these options don't have default values, you have to opt into it | |
| 18:37:51 | mriedem | so buyer beware | |
| 18:38:15 | mriedem | https://github.com/openstack/nova/blob/e731813796764a1034e7e16a6c154c8831984c9b/releasenotes/notes/bp-libvirt-virtio-set-queue-sizes-6c54a2ce3dc30d18.yaml#L6 | |
| 18:38:19 | mriedem | https://github.com/openstack/nova/blob/34956bea4beb8e5ba474b42ba777eb88a5eadd76/nova/conf/libvirt.py#L1213 | |
| 18:41:47 | mriedem | melwitt: so punt on that for rc3 is what i'm saying | |
| 18:42:15 | melwitt | that is ideal. thanks | |
| 18:44:48 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/rocky: libvirt: Don't react to VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED events https://review.openstack.org/595864 | |
| 18:44:49 | mriedem | ^ is the other one you needed | |
| 18:52:40 | mriedem | dansmith: melwitt: btw, our public cloud guy was fine with the option to 500 if users are listing and have servers in down cells | |
| 18:52:50 | dansmith | sweet | |
| 18:53:16 | melwitt | *thumbs up* | |
| 19:21:40 | openstackgerrit | melanie witt proposed openstack/nova master: Correct the release notes related to nova-consoleauth https://review.openstack.org/595455 | |
| 19:24:02 | openstackgerrit | melanie witt proposed openstack/nova master: Correct the release notes related to nova-consoleauth https://review.openstack.org/595455 | |
| 19:28:36 | melwitt | mriedem: ^ updated the release note patch | |
| 19:29:25 | mriedem | so the config option help doesn't need to be updated? | |
| 19:30:40 | melwitt | oh, sorry. I missed that | |
| 19:32:45 | mriedem | looks ok otherwise | |
| 19:32:50 | mriedem | just update the option help and i think we're good | |