| opendevreview | melanie witt proposed openstack/nova master: Add config option [quota]strict_unified_limits https://review.opendev.org/c/openstack/nova/+/924025 | 02:21 |
|---|---|---|
| opendevreview | melanie witt proposed openstack/nova master: Add config option [quota]strict_unified_limits https://review.opendev.org/c/openstack/nova/+/924025 | 07:32 |
| melwitt | sean-k-mooney: ^ POC attempt at the different logic we chatted about the other day | 07:35 |
| opendevreview | Balazs Gibizer proposed openstack/nova stable/2023.2: fix qemu-img version dependent tests https://review.opendev.org/c/openstack/nova/+/923907 | 08:04 |
| opendevreview | Balazs Gibizer proposed openstack/nova stable/2023.2: Stabilize iso format unit tests https://review.opendev.org/c/openstack/nova/+/923936 | 08:04 |
| opendevreview | Balazs Gibizer proposed openstack/nova stable/2023.1: fix qemu-img version dependent tests https://review.opendev.org/c/openstack/nova/+/923908 | 08:09 |
| opendevreview | Balazs Gibizer proposed openstack/nova stable/2023.1: Stabilize iso format unit tests https://review.opendev.org/c/openstack/nova/+/923937 | 08:09 |
| sean-k-mooney | melwitt: oh cool. after creating that how do you feel about one approch vs the other. no need to reply now but let me know after you get some sleep | 09:38 |
| gibi | sean-k-mooney: could you approve https://review.opendev.org/c/openstack/nova/+/923935 to continue landing the series? | 10:09 |
| gibi | most of the 2023.2 is also on the gate now thanks to elodilles | 10:09 |
| sean-k-mooney | i for got to clice the buttons and only left a commet ... | 10:13 |
| sean-k-mooney | but done now | 10:14 |
| gibi | sean-k-mooney: thanks, could you review the same patch on the older branches? Then I can just ping elodilles when we need the approve | 10:15 |
| sean-k-mooney | already am | 10:15 |
| sean-k-mooney | ok i think im +2 on all the ones i didnt push | 10:17 |
| opendevreview | Merged openstack/nova stable/2023.2: port format inspector tests from glance https://review.opendev.org/c/openstack/nova/+/923727 | 10:27 |
| opendevreview | Merged openstack/nova stable/2023.2: Reproduce iso regression with deep format inspection https://review.opendev.org/c/openstack/nova/+/923728 | 10:27 |
| frickler | gibi: sean-k-mooney: gate failure in openstacksdk-functional-devstack on https://review.opendev.org/c/openstack/nova/+/923729 . if you can double check that it is unrelated, I'd submit it right back into gate again | 11:18 |
| opendevreview | Merged openstack/nova stable/2024.1: Stabilize iso format unit tests https://review.opendev.org/c/openstack/nova/+/923935 | 11:20 |
| opendevreview | Merged openstack/nova stable/2023.1: port format inspector tests from glance https://review.opendev.org/c/openstack/nova/+/923731 | 11:20 |
| opendevreview | Merged openstack/nova stable/2023.1: Reproduce iso regression with deep format inspection https://review.opendev.org/c/openstack/nova/+/923732 | 11:21 |
| vicent | Good morning! I am using sriov, I have two PFs with two VFs each. The PFs are connected to different networks, so when I create a neutron port with type direct, I need a way to map that port so it uses one of the VFs of a specific PF. I found this document https://specs.openstack.org/openstack/nova-specs/specs/zed/approved/pci-device-tracking-in-placement.html which uses traits, but I need to set | 11:41 |
| vicent | the trait on the port, not on the flavor. Basically, how can I configure the VFs of different PF in pools, so I have control to which VF pool is used for which network? I managed to get my PFs as resource providers on placement and I my ports get correctly bound to the VFs, but I can't control the mapping to a specific custom trait. How does nova request a specific resource provider for the port? | 11:41 |
| vicent | Is there a way to do what I am trying to do? | 11:42 |
| sean-k-mooney | vicent: the only way to do that today is to use a differnt neutron phsynet for each PF | 11:43 |
| sean-k-mooney | there is no support for anti affintiy or vf bonding in openstack today | 11:44 |
| sean-k-mooney | some have tried to hack it in without enableing it in nova or neutron | 11:44 |
| sean-k-mooney | melonox have some vf lag supprot | 11:44 |
| sean-k-mooney | but its not officaly supproted in nova or neutorn | 11:44 |
| vicent | I can't use physnet in my case :S | 11:45 |
| vicent | Do you have a link to the mellanox resource? | 11:45 |
| sean-k-mooney | then unfortunetly you cant do this in openstack | 11:45 |
| sean-k-mooney | the proibelme with the melxons approch si there is no awarenss of it in nova ans no schduler supprot as a result | 11:46 |
| sean-k-mooney | there is alos no way to request it as an end user | 11:46 |
| sean-k-mooney | it done out fo band bay howe you confirg the host hardware | 11:46 |
| vicent | ok, then it won't work for me either | 11:46 |
| sean-k-mooney | https://www.stackhpc.com/vflag-kayobe.html | 11:47 |
| vicent | does nova have any plugin/extension mechanism that could be used to control that? | 11:49 |
| sean-k-mooney | no | 11:49 |
| vicent | so if I wanted to get that, I would need to modify nova directly... Ok | 11:50 |
| sean-k-mooney | as i said the only way to do anti affinity at the pf/vf level is with physnets | 11:50 |
| sean-k-mooney | yes and form an upstream poitn of view we do not plann to supprot this until neutron has a bonding api similer to trunk ports | 11:51 |
| sean-k-mooney | which was discuss a few years ago but no one is currently working on that | 11:51 |
| vicent | My use case is not for bonding | 11:51 |
| sean-k-mooney | well we also don thave any kind of port grouping or anti affintiy api in neutron that can be used moddel the relationship | 11:52 |
| sean-k-mooney | that really the gap | 11:52 |
| vicent | It is more about scheduling and resource selection, not grouping or anti-affinity | 11:53 |
| sean-k-mooney | right so pci in palcement does nto supprot neutorn prots righ tnow | 11:53 |
| sean-k-mooney | that is planned for a future release | 11:53 |
| vicent | I want to be able to enforce a specific PF on my port without physnet. | 11:53 |
| sean-k-mooney | vicent: well that is somethign we explictly dont support | 11:54 |
| sean-k-mooney | its a disging goal ot not allow you to directly choose the PF on a host | 11:54 |
| sean-k-mooney | you can poticaly abuse traits to support that | 11:54 |
| vicent | well, not the PF, but the any VFs of that PF | 11:55 |
| sean-k-mooney | but we dont intent to allwo that for the same reason we dont support selecting numa nodes or cpus | 11:55 |
| sean-k-mooney | that is simpler to do and can be acive once we supprot neutron sriov prots in placemnet | 11:55 |
| sean-k-mooney | this is the planned desgin https://specs.openstack.org/openstack/nova-specs/specs/2023.1/implemented/pci-device-tracking-in-placement.html#neutron-sr-iov-ports-out-of-scope | 11:57 |
| sean-k-mooney | neutron port can have resouce requests. and if we support triat request via that on the port level it would enable your usecase in a not terrible way | 11:58 |
| sean-k-mooney | however i expect it to be a year or more before we implement that | 11:59 |
| vicent | Ok, that clarifies | 11:59 |
| vicent | I guess I will try with physnet and without placement | 12:00 |
| vicent | Thanks sean-k-mooney | 12:00 |
| sean-k-mooney | the hrrible hack that telcos have done for years is ot have 2 phsnets on the same phsyical netowrk(something your not allowed to do technialy) | 12:01 |
| sean-k-mooney | and create a netron neetwork with the same vlan on both | 12:01 |
| sean-k-mooney | and then one port on each network | 12:01 |
| sean-k-mooney | that allwos them to have one port form each numa node or on diffent phsyical nic | 12:02 |
| sean-k-mooney | depending on how they created the phsynets | 12:02 |
| sean-k-mooney | that breaks tenant isolation as vlans on diffent phsynets are not ment to be able to interact like that | 12:03 |
| sean-k-mooney | but in a private cloud where the network are created by an admin that work for them | 12:03 |
| gibi | frickler: I agree that was an unrealted failure | 12:30 |
| gibi | thanks for monitoring it | 12:31 |
| opendevreview | Balazs Gibizer proposed openstack/nova master: DNM: test with post-copy enabled https://review.opendev.org/c/openstack/nova/+/924075 | 14:48 |
| opendevreview | Merged openstack/nova stable/2023.2: Add iso file format inspector https://review.opendev.org/c/openstack/nova/+/923729 | 15:31 |
| opendevreview | Merged openstack/nova stable/2023.2: fix qemu-img version dependent tests https://review.opendev.org/c/openstack/nova/+/923907 | 15:31 |
| melwitt | sean-k-mooney: I think I like the second approach more. it turned out to simpler than I expected | 17:22 |
| opendevreview | Merged openstack/nova stable/2023.2: Stabilize iso format unit tests https://review.opendev.org/c/openstack/nova/+/923936 | 18:16 |
| opendevreview | Merged openstack/nova stable/2023.1: Add iso file format inspector https://review.opendev.org/c/openstack/nova/+/923733 | 18:16 |
| opendevreview | Merged openstack/nova stable/2023.1: fix qemu-img version dependent tests https://review.opendev.org/c/openstack/nova/+/923908 | 18:16 |
| sean-k-mooney | melwitt: you have encapulstaed the logic pretty cleanly into the should_enforce function | 18:39 |
| sean-k-mooney | melwitt: im not entirly sure how i feel about caching the keystone client for the limt enfroceer | 18:40 |
| sean-k-mooney | its proably fine if it internally handels updating the auth | 18:41 |
| sean-k-mooney | when a token expires | 18:41 |
| sean-k-mooney | we can have diffent limties per project howver right so w eneed to ensure we are usign the users token to get teh registered limits so we might need to concruct it each time | 18:43 |
| melwitt | sean-k-mooney: yeah, I wasn't totally sure about that but I think it does. we cache most of the clients like placement and neutron IIRC | 18:43 |
| melwitt | sean-k-mooney: oh.. yeah good point. 😆 | 18:44 |
| sean-k-mooney | dont we normally passs a diffent context to the client wehn making a call | 18:44 |
| sean-k-mooney | with the user token | 18:44 |
| melwitt | yeah you're right | 18:44 |
| melwitt | I'll need to change that | 18:44 |
| sean-k-mooney | so in stead of needign a new keyston section in the config i thin kwe can use the user token + our service user tooken | 18:44 |
| sean-k-mooney | ignoing th eclinet the patch looks pretty nice to me in general | 18:46 |
| sean-k-mooney | elodilles: thanks for all the work you have done on the iso fromat backport | 18:48 |
| melwitt | I'll revise the spec to the second approach | 18:49 |
| sean-k-mooney | if your happy with that it works for me. ill try and find time to review the spec and patch properly on monday so | 18:50 |
| melwitt | ok I'll have it updated by then. thanks | 18:52 |
| sean-k-mooney | well dont kill your self to do it over the weekend but if we want to do it this cycle we need to get it landed before spec freze which is thrusday | 18:56 |
| melwitt | I did this to myself 😝 but it also won't kill me. it should be easy (tm) | 18:57 |
| elodilles | sean-k-mooney: no problem :) thanks for the fixes :] | 20:21 |
| opendevreview | Elod Illes proposed openstack/nova master: [tools] Ignore bot generated patches https://review.opendev.org/c/openstack/nova/+/924103 | 21:19 |
| opendevreview | Elod Illes proposed openstack/nova master: [tools] Backport validator: handle unmaintained https://review.opendev.org/c/openstack/nova/+/924104 | 21:19 |
| opendevreview | Merged openstack/nova stable/2023.1: Stabilize iso format unit tests https://review.opendev.org/c/openstack/nova/+/923937 | 21:32 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!