| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2026-03-25 | |||
| 16:31:20 | elodilles | Uggla: just a note, that i haven't created a nova RC2 patch as there're only doc change fixes on stable/2026.1 | |
| 16:31:41 | Uggla | elodilles ok | |
| 16:36:39 | Uggla | sean-k-mooney, if you have time, could you make a first pass review of 981557: Reproduce bug #2108974: keypairs lost during cross-cell resize | https://review.opendev.org/c/openstack/nova/+/981557 and 981558: Fix keypairs lost during cross-cell resize | https://review.opendev.org/c/openstack/nova/+/981558 | |
| 16:45:49 | Uggla | gibi, I know you are busy, there is no hurry. But if you can have https://review.opendev.org/c/openstack/nova/+/946945/2 and https://review.opendev.org/c/openstack/nova/+/970337/5 on your re view list that will be great ! | |
| 17:33:11 | clif | o/ | |
| 17:33:53 | JayF | sean-k-mooney: clif is a developer from GR-OSS, I've asked him to help review+cleanup your ironic speedup patches | |
| 17:34:06 | JayF | sean-k-mooney: you'll be better off with clif than me for that kinda stuff; he's more developer coded (to my ops-coded nature) | |
| 17:34:20 | sean-k-mooney | cool | |
| 17:34:33 | sean-k-mooney | clif: o/ | |
| 17:35:03 | cardoe | sean-k-mooney: I've passed https://bugs.launchpad.net/ironic/+bug/2076265 back to Nova. I've left Ironic on it but Nova has to make a change first. | |
| 17:35:06 | clif | nice to meet you, hopefully I can help some of these over the line at least | |
| 17:36:16 | sean-k-mooney | clif: so as i said in the last irc meeting its easy to make things go fast if correctness is not a requirement so that actully goign to part fo the challnge here making sure that these optmistaion are both valid and correct | |
| 17:36:37 | clif | fair enough | |
| 17:37:06 | clif | I also prefer valid and correct | |
| 17:37:14 | clif | fast and incorrect is not very valuable imo | |
| 17:37:15 | JayF | how fast we talkin'? /s | |
| 17:37:46 | sean-k-mooney | JayF: so i think 2 of the 4 optimistiaon are likely correct and of high value | |
| 17:37:50 | sean-k-mooney | the other 2 im unsure | |
| 17:38:07 | sean-k-mooney | so using the threadpool to process nodes in paralle and the deepcopy optimisation | |
| 17:38:38 | sean-k-mooney | but i would love if we coudl create a devstack with the fake ironic driver and try and replciate this in a more real env to see if they actullly help | |
| 17:38:53 | sean-k-mooney | i.e. to make 100% sure that its not an artifcat of the simulator | |
| 17:39:02 | sean-k-mooney | i dont think they are but that still an open | |
| 17:39:46 | sean-k-mooney | for the start up case we dont even really need a full ironic just enouch to hae the api +db so we can add the nodes that nova cna then discover/enmerate | |
| 17:40:03 | JayF | > create a devstack with the fake ironic <-- this is #3 on my list r/n | |
| 17:40:29 | sean-k-mooney | are there docs that cover that or example local.confs | |
| 17:40:38 | JayF | there will be when I get there on my list | |
| 17:41:09 | sean-k-mooney | cardoe: ack for what its worth nova only supprot prot with vnic_tyupe normal for --network or --subnet | |
| 17:41:36 | sean-k-mooney | cardoe: that why we have alwasy said to use ironic you must precreate teh neutron port with vnic-type=baremetal | |
| 17:41:58 | cardoe | I understand that. Ironic has never supported VNIC_NORMAL and this has been the behavior since Nova added Ironic support on day 1. | |
| 17:41:59 | JayF | sean-k-mooney: what's the path to changing that? | |
| 17:41:59 | sean-k-mooney | we dont block that using --network and --subnet with ironci flavor today however | |
| 17:42:13 | JayF | sean-k-mooney: changing the support for --network/--subnet to include baremetal vnic | |
| 17:42:26 | cardoe | It's invalid for Nova to ever create the port and send it to the hypervisor backend. | |
| 17:42:26 | sean-k-mooney | JayF: that was rejected in the past as wont do | |
| 17:42:32 | JayF | why? | |
| 17:42:41 | JayF | It's the future now! | |
| 17:42:46 | sean-k-mooney | JayF: i tried to get that change approved as part fo the get me a newtrork work | |
| 17:42:54 | cardoe | In fact there's a couple of hypervisor backends that don't support VNIC_NORMAL and Nova shouldn't do that for them. | |
| 17:43:04 | sean-k-mooney | but tldr was a mix of the no more proxy adn no more orchstration policy | |
| 17:43:05 | cardoe | JayF: https://review.opendev.org/c/openstack/nova/+/979354 I wrote the fix already. | |
| 17:43:33 | sean-k-mooney | cardoe: ironic is the only one im aware of tha tdoes nto work with normal | |
| 17:45:00 | sean-k-mooney | JayF: im not against it but we basiclly nova didnt want to keep adding new proxy comamdn for port creation aspect and standardise on you alwasy create a port ahead of tiem | |
| 17:45:18 | sean-k-mooney | we obviously cant remove the exisitng fucntionaltiy but that is the history of why it was rejected before | |
| 17:45:34 | JayF | Might be interesting to revisit, if for nothign else than backwards compat with existing broken Ironic flows (Which have been in use, even if "broken" for a decade+) | |
| 17:46:11 | sean-k-mooney | one suggetion that i didnt really by was this can be automated in the cli/sdk | |
| 17:46:24 | sean-k-mooney | technaily anythign can be but that felt like a non answer | |
| 17:46:44 | JayF | I just need better reasons for my downstream about making their flows more complex | |
| 17:46:48 | sean-k-mooney | we have the exact same gap for neutron sriov ports | |
| 17:47:00 | JayF | precreating ports in Nova makes us inch closer and closer to deployments w/nova being as complex as deployments w/o nova | |
| 17:47:11 | cardoe | Yep. | |
| 17:47:52 | sean-k-mooney | that is true but that is offically the only thing nova ahs ever supproted | |
| 17:48:08 | sean-k-mooney | the changing of the vnic type by ironic after the fact is techniclly not supproted | |
| 17:48:16 | sean-k-mooney | even thought at what everyone relies on | |
| 17:48:21 | JayF | okay; it's been working great and not supported for a decade | |
| 17:48:28 | JayF | so that's sorta an oroboros-y thing to say at some point | |
| 17:48:38 | cardoe | I'm really trying hard to contribute to Nova. To tell others to contribute to Nova. But I'm just constant told no. | |
| 17:48:45 | sean-k-mooney | well it violating nueton api contract too | |
| 17:48:57 | sean-k-mooney | serivce are not ment to modify the vnice tyep and it not supproted to change it on a bound port | |
| 17:49:09 | cardoe | Literally the first commit adding support for Ironic into Nova relied on that behavior. So I'm at a loss how it's unsupported. | |
| 17:49:33 | JayF | ^^ this is what I'm saying, I would love real reasoning other than "because it's always been that way" | |
| 17:49:36 | cardoe | sean-k-mooney: you can change the type of an unbound port. The port isn't yet bound. It's valid by Neutron spec and API. | |
| 17:49:57 | cardoe | That was confirmed during the last PTG. | |
| 17:50:08 | sean-k-mooney | you can but a service shoudl not the vnice tyep is a request form the end user | |
| 17:50:25 | sean-k-mooney | and a port with a nova isntane atach shoudlnot be modifed by ironic but it will be | |
| 17:50:33 | sean-k-mooney | as i said this is how it works today | |
| 17:50:51 | sean-k-mooney | but it has been a long standing pain point that ironic is not workign liek every other virt dirver | |
| 17:50:52 | cardoe | s/today/has always worked/ | |
| 17:51:04 | sean-k-mooney | right and it has alwasy been a mild security issue | |
| 17:51:22 | sean-k-mooney | nova actully did have a security issue which we have fix related to vnic types beign changed | |
| 17:51:54 | sean-k-mooney | anywya im not agains supprotign specifying a vnic type but other are | |
| 17:52:11 | sean-k-mooney | so this would be an api change and one requiring a spec | |
| 17:52:34 | sean-k-mooney | valid to dicuss just ohters woudl prefer not too change this | |
| 17:52:41 | cardoe | Well I've been trying to get a spec approved for 2 years and now ya want me to write another? | |
| 17:52:58 | sean-k-mooney | all api changes require specs that unconditional | |
| 17:53:01 | JayF | maybe we can find what needs to change in your spec to be what sean-k-mooney needs? | |
| 17:53:06 | sean-k-mooney | so this woudl never be conidered a bug in nova | |
| 17:53:06 | JayF | can that be put in review? | |
| 17:53:38 | sean-k-mooney | JayF: im +2 on the ohter spec by the way and have been for 3 cycles :) | |
| 17:53:49 | cardoe | It's absolutely a Nova bug. The Nova Ironic driver creates a Port that cannot work with Ironic. | |
| 17:55:16 | sean-k-mooney | no the compute manger does | |
| 17:55:38 | sean-k-mooney | vnic_type=normal is the only thing that nova can create | |
| 17:55:50 | sean-k-mooney | and we have alwasy say that to use ironic you must pass a port | |
| 17:55:59 | sean-k-mooney | as a reasult of the requrieemnt for vnic tyep baremental | |
| 17:57:30 | cardoe | Can you point me to the docs which say to use Ironic with Nova you must pass your own Port? | |
| 17:57:36 | cardoe | https://docs.openstack.org/nova/2025.2/admin/configuration/hypervisor-ironic.html doesn't show it | |
| 17:59:14 | cardoe | https://docs.openstack.org/nova/2025.2/user/support-matrix.html even shows the operation where Nova is creating the Port itself and as supported. | |
| 18:02:12 | sean-k-mooney | cardoe: its partly a sideffe of nova only supprot vnic_type=normal prot when it creats them and https://review.opendev.org/c/openstack/nova-specs/+/672400 was the last tiem we propsed this as a spec | |
| 18:02:26 | sean-k-mooney | cardoe: ill see fi we have an explcit doc or not | |
| 18:21:06 | sean-k-mooney | cardoe: ironic has https://cdn.maingear.com/docs/ironic/2025.1/admin/networking.html#vif-attachment-flow the srong recoemdation to alwasy create the prot in advnce. one thing to note is while teh current flow was orgianly approved in https://github.com/openstack/nova-specs/blob/master/specs/mitaka/approved/ironic-networks-support.rst#tenant-networking-support-for-ironic-driver | |
| 18:21:08 | sean-k-mooney | with the modification of the vnic type out of band by ironic that spec was never reappovoed and accpated for a later release but the code wa smerged for some unknow reason | |
| 18:21:53 | sean-k-mooney | at the time this was approve the only usecase for a non defautl vnic type was neutron sriov prot which requried the port to be precreated. | |
| 18:21:59 | cardoe | Let's just abandon my proposed change. I'll go away. | |
| 18:22:20 | sean-k-mooney | cardoe: i think the idea of a hypervior depenent defualt vnic type is not a bad one | |
| 18:22:50 | sean-k-mooney | so im not against this just provideign the wider context | |
| 19:00:52 | JayF | Yeah I think we have the fix in line with cardoe's change, just need to figure out the right way to market it to at least one other Nova core and if needed update the paperwork to be in line with whatever is expected if they consider that a feature and not a bugfix | |
| 20:07:24 | sean-k-mooney | JayF: next step would be nova irc meeting. it would be better in the long term if we also had a path to growign more nova cores with ironic expertiese. in the past john garbut was one but they have less time for nova these days | |
| 20:07:39 | sean-k-mooney | JayF: im +1 on the patch over all. | |
| 20:08:39 | JayF | sean-k-mooney: similarly on the Ironic side; I've tried to take the mantle of "expert in the driver", which I think I've "hit" but many of the things here are lower level than that. Between the group of us we can find a solution I'm sure :) | |
| 20:10:58 | opendevreview | Antonin Ruan proposed openstack/nova-specs master: Temp URL download spec https://review.opendev.org/c/openstack/nova-specs/+/975883 | |
| 20:22:00 | opendevreview | Erlon R. Cruz proposed openstack/nova master: Add RBD XML update functionality for migration process https://review.opendev.org/c/openstack/nova/+/974032 | |