| opendevreview | Iury Gregory Melo Ferreira proposed openstack/ironic master: [PoC] Unedited AI dump: asynchronous Redfish client https://review.opendev.org/c/openstack/ironic/+/970450 | 02:06 |
|---|---|---|
| opendevreview | Merged openstack/ironic master: Fail loudly when the console browser does not start https://review.opendev.org/c/openstack/ironic/+/1005360 | 02:45 |
| opendevreview | Merged openstack/ironic master: Create the console browser profile in a known directory https://review.opendev.org/c/openstack/ironic/+/1005599 | 03:00 |
| opendevreview | Jacob Anders proposed openstack/sushy master: Add per-job status API to DellJobCollection https://review.opendev.org/c/openstack/sushy/+/1005653 | 05:00 |
| opendevreview | Jacob Anders proposed openstack/ironic master: Add Dell LC job status helper for multiple JIDs https://review.opendev.org/c/openstack/ironic/+/1005656 | 05:42 |
| opendevreview | Jacob Anders proposed openstack/ironic master: Add Redfish BootProgress gate helper https://review.opendev.org/c/openstack/ironic/+/1005657 | 05:42 |
| opendevreview | Jacob Anders proposed openstack/ironic master: Verify firmware application after reboot before resuming https://review.opendev.org/c/openstack/ironic/+/1005658 | 05:42 |
| opendevreview | Jacob Anders proposed openstack/ironic master: Validate BMC stability only after BMC firmware updates https://review.opendev.org/c/openstack/ironic/+/1005659 | 05:42 |
| opendevreview | Merged openstack/ironic-ui master: Imported Translations from Zanata https://review.opendev.org/c/openstack/ironic-ui/+/1003213 | 09:10 |
| opendevreview | Merged openstack/ironic-python-agent master: tox: Drop redundant passenv https://review.opendev.org/c/openstack/ironic-python-agent/+/1004203 | 09:13 |
| opendevreview | Merged openstack/python-ironicclient master: Update master for stable/2026.2 https://review.opendev.org/c/openstack/python-ironicclient/+/1004633 | 09:18 |
| opendevreview | Roussarie proposed openstack/ironic-specs master: Add spec for rebooting into the user image via kexec https://review.opendev.org/c/openstack/ironic-specs/+/1005717 | 12:09 |
| lroussarie[m] | good morning ! | 12:19 |
| lroussarie[m] | I've proposed a spec for RFE 1757137 (kexec boot): https://review.opendev.org/c/openstack/ironic-specs/+/1005717 | 12:19 |
| lroussarie[m] | reviews appreciated. | 12:23 |
| opendevreview | Merged openstack/ironic master: Install curl in the console container images https://review.opendev.org/c/openstack/ironic/+/1005600 | 12:48 |
| opendevreview | Merged openstack/ironic master: Exit non-zero when the certificate override cannot be built https://review.opendev.org/c/openstack/ironic/+/1005629 | 12:48 |
| opendevreview | cid proposed openstack/ironic stable/2025.2: fix broken project scoping https://review.opendev.org/c/openstack/ironic/+/1004661 | 12:49 |
| opendevreview | cid proposed openstack/ironic stable/2025.1: fix broken project scoping https://review.opendev.org/c/openstack/ironic/+/1004693 | 13:23 |
| opendevreview | cid proposed openstack/ironic unmaintained/2024.1: fix broken project scoping https://review.opendev.org/c/openstack/ironic/+/1004706 | 13:40 |
| *** Unknown is now known as Mike-- | 14:51 | |
| JayF | dtantsur: look what I found https://specs.openstack.org/openstack/ironic-specs/specs/not-implemented/boot-config-api.html | 15:43 |
| * JayF is doing an audit of unimplemented crap from previous cycles for the PTG | 15:44 | |
| JayF | IDK if that is the right answer, but it's an answer we have :D | 15:44 |
| dtantsur | JayF: it's something I've been pondering, but it's only a half of a solution for Kubernetes | 15:45 |
| JayF | my thought is just we either need to use that spec or lose it | 15:45 |
| dtantsur | I think our recent idea is more promising tbh | 15:48 |
| * JayF added a section for incomplete items from previous cycles at the bottom of https://etherpad.opendev.org/p/ironic-ptg-2027.1 | 15:49 | |
| JayF | dtantsur: yeah, like I said my main concern is that we either do it or unapprove/remove it :) that's the overall goal with all this old stuff | 15:50 |
| * JayF -> podcast taping, brb in a long while | 15:50 | |
| * dtantsur nods | 15:50 | |
| TheJulia | lroussarie[m]: I put some comments on that spec, some good ideas there but a few things to look at/consider | 15:50 |
| opendevreview | Guillaume Boutry proposed openstack/ironic master: Mock firmware cache refresh in completion tests https://review.opendev.org/c/openstack/ironic/+/1005758 | 16:08 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: add the node_maintenance table https://review.opendev.org/c/openstack/ironic/+/1000751 | 16:24 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: NodeMaintenance object and DB-API https://review.opendev.org/c/openstack/ironic/+/1000752 | 16:24 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: mirror the boolean into node_maintenance https://review.opendev.org/c/openstack/ironic/+/1000753 | 16:24 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: document maintenance in node history https://review.opendev.org/c/openstack/ironic/+/1000754 | 16:24 |
| opendevreview | Guillaume Boutry proposed openstack/ironic master: Mock firmware cache refresh in completion tests https://review.opendev.org/c/openstack/ironic/+/1005758 | 16:55 |
| opendevreview | cid proposed openstack/ironic unmaintained/2024.1: fix broken project scoping https://review.opendev.org/c/openstack/ironic/+/1004706 | 17:11 |
| dxterslab | hey guys, do you guys know if this OVN plugin https://docs.ovn.org/en/latest/topics/vif-plug-providers/index.html can be used to add a baremetal node to OVN network by leveraging DPUs like BF3? (By baremetal node, I mean the actual baremetal operating system be part of the OVN network just like a VM. The baremetal host would have no idea it is part of a virtual network since it only sees a virtual interface provided by the DPU and | 17:46 |
| dxterslab | has no other awareness) | 17:46 |
| opendevreview | Jay Faulkner proposed openstack/ironic master: DNM/WIP (direct from AI) implement old spec about power states https://review.opendev.org/c/openstack/ironic/+/1005776 | 17:51 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: add the node_maintenance table https://review.opendev.org/c/openstack/ironic/+/1000751 | 17:51 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: NodeMaintenance object and DB-API https://review.opendev.org/c/openstack/ironic/+/1000752 | 17:51 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: mirror the boolean into node_maintenance https://review.opendev.org/c/openstack/ironic/+/1000753 | 17:51 |
| opendevreview | cid proposed openstack/ironic master: Maintenance: document maintenance in node history https://review.opendev.org/c/openstack/ironic/+/1000754 | 17:51 |
| JayF | https://review.opendev.org/c/openstack/ironic/+/1005776 is not bad. I let deepseek v4 pro write it in opencode. I don't even know if we still want the feature anymore | 18:01 |
| JayF | but it's neat | 18:01 |
| cardoe | dxterslab: it should | 18:12 |
| JayF | dxterslab: do you work with me? That sounds like a question folks I work with would be asking :D | 18:14 |
| JayF | dxterslab: cardoe is the expert on this stuff in ironic though, he'll take care of ya | 18:14 |
| cardoe | Also "should" is doing a lot of heavy lifting here. | 18:15 |
| cardoe | If you're expecting a package install and a simple config file and it all works... I think that's a stretch. | 18:26 |
| cardoe | Is that a framework that can work with a little bit of code / tweaking / bubble gum / hopeium.... it's the right place. | 18:26 |
| dxterslab | Sounds good. I was hoping that i can just put a golang wrapper around it and orchestrate the setup. I do need to test it manually first tho | 19:52 |
| opendevreview | Doug Goldstein proposed openstack/networking-baremetal stable/2026.1: Remove binding:host_id setting for L2VNI subports https://review.opendev.org/c/openstack/networking-baremetal/+/1005793 | 20:01 |
| opendevreview | Doug Goldstein proposed openstack/networking-baremetal stable/2026.1: Guard rabbit config access in _get_notification_transport_url https://review.opendev.org/c/openstack/networking-baremetal/+/1005794 | 20:03 |
| opendevreview | Doug Goldstein proposed openstack/networking-baremetal stable/2026.1: Remove invalid requested-chassis opt for localnet ports https://review.opendev.org/c/openstack/networking-baremetal/+/1005795 | 20:04 |
| cardoe | hjensas: apparently those are the 3 patches I've pulled into our 2026.1 usage | 20:08 |
| TheJulia | cardoe: Regarding dxterslab's question that requires consistency across the cloud and the ovn participation/engagement by the software on the device, which.... ahem... as I understand it was rejected because nvidia decided to go in the direction of a ovs fork instead. | 20:35 |
| TheJulia | Which is why I've sort of mused, maybe a interface to go do on-card configuration remotely via it's management interface might be a way to bridge the gap | 20:35 |
| TheJulia | One day I'll feel like a humanoid carbon unit again | 20:44 |
| TheJulia | (jetlag) | 20:44 |
| cardoe | So I mean the OVN vif plugin providers IMHO is the best place in a perfect world where you tried to make a single pane of glass networking thingy | 20:46 |
| cardoe | But you and I both know that OVN doesn't necessarily store all the networking data with conforms with the world. | 20:47 |
| TheJulia | Well, so there are a couple distinct issues, including just the card being powered on | 20:47 |
| TheJulia | because that itself is model dependent and the neutron integrated model is actually to treat it as a compute node as well... | 20:47 |
| TheJulia | well, as networking services on a compute node | 20:47 |
| cardoe | oh yes good point | 21:00 |
| TheJulia | So, that is all rooted where the is_smartnic flag came from. I know some vendors are doing dynamic core liveliness as well with some special firmware as well, i.e. dropping the runtime os only to cpu0 on the card, but its not universlal either | 21:01 |
| dxterslab | my idea was not to integrate with BF3s OVN/OVS implementation but run my own OVN/OVS implementation in a container similar to how ovn-kubernetes does it but implemented in kube-ovn. I still have to see if running another implementation of ovn/ovs is possible on BF3 | 21:01 |
| TheJulia | strictly speaking, I'm unsure and you definitely won't get the same wire speed behavior if it even works, I'm super curious if it does though | 21:03 |
| JayF | dxterslab: Are you using integrated openstack w/neutron included? Just verifying. | 21:04 |
| dxterslab | mhmmm... good point... i would loose offload capabilities. I will look at how ovn-kubernetes has integrated with it | 21:04 |
| TheJulia | a clean way would be to leverage that OVN/OVS agent with the neutron-ovs-agent *might* work, or at least get you closer | 21:04 |
| TheJulia | dxterslab: yeah, the downside of the ovs on the base OS being an effective fork at this point :( | 21:04 |
| dxterslab | No Openstack or neutron here lol. Just metal3 with Ironic and looking to orchestrate network services on baremetal nodes as if they were VMs | 21:04 |
| TheJulia | heh, so, yeah, that won't work with the neutron agent stuffs | 21:05 |
| TheJulia | back to why I was thinking we build a plugin which logs into the card and treats it like a "switch" | 21:05 |
| cardoe | yep | 21:05 |
| cardoe | I might have switches already enrolled as ironic nodes | 21:05 |
| TheJulia | s/might// ;) | 21:06 |
| cardoe | it's my noop driver right now. | 21:06 |
| cardoe | but inspect is coming | 21:06 |
| JayF | you can do anything you can make openstack do cardoe | 21:06 |
| JayF | I'll only get annoyed if you document it for others ;) | 21:07 |
| JayF | lol | 21:07 |
| cardoe | Well I did want to sketch it out for a group of you and see if it made sense. | 21:07 |
| cardoe | And if it did I would sketch something for the ironic blog. | 21:07 |
| JayF | put it on the PTG? | 21:07 |
| JayF | that's the perfect place to get that kinda consensus | 21:07 |
| cardoe | yeah I can do that. | 21:08 |
| dxterslab | Just to give you an idea of what i am planning to do. We use kubernetes today to orchestrate baremetal nodes with metal3 and ironic. We are looking to orchestrate the network for the baremetal nodes and we want, as a requirement, for baremetal nodes to exist on the same L2 as VMs and odd cases kubernetes pods. We use kube-ovn as a CNI today and it allow us to deliver very similar cloud services on our virtual stuff. I am musing on | 21:08 |
| dxterslab | how to integrate BM into that virtual network and whether it is possible cause it would bring all the benefits of OVN to the baremetal node and provide the same capabiliities we can achieve virtually | 21:08 |
| cardoe | Not necessarily consensus but more like "is this super offensive and I need to stop?" or "that's cool. write about it." | 21:08 |
| JayF | dxterslab: OK now I believe we don't work together ;) | 21:08 |
| TheJulia | cardoe: Remember, the self hugging jackets are the warmest jackets :) | 21:09 |
| cardoe | speaking of offensive things... TheJulia that BMC spec/api patch | 21:09 |
| cardoe | Let's say I needed to do a yolo implementation against current Ironic and don't wanna risk DB migrations and microversion incompat... | 21:10 |
| cardoe | if I wire something behind vendor passthru... that's probably the least offensive? | 21:11 |
| TheJulia | I already told you that wasn't offensive | 21:11 |
| cardoe | I just need to send 2 different blerps of BMC settings against a $VENDOR box | 21:11 |
| JayF | cardoe: is this that redfish proxy spec? | 21:12 |
| JayF | cardoe: I read that today when sorting through "incomplete specs to discuss at ptg" and that one gave me heartburn lol | 21:12 |
| cardoe | https://review.opendev.org/c/openstack/ironic-specs/+/1004909 | 21:12 |
| cardoe | That's first pass at doing it correctly. | 21:13 |
| JayF | okay that is much better than the old spec I refer to | 21:13 |
| cardoe | It's to let us manage the BMC settings much like the BIOS settings are managed. | 21:13 |
| cardoe | With validation and all. | 21:13 |
| JayF | cardoe: https://specs.openstack.org/openstack/ironic-specs/specs/not-implemented/ironic-redfish-proxy.html made me go BWAH like Hank Hill | 21:13 |
| cardoe | oh yes. Said $VENDOR linked that to me and did their best Blue Steel "Do It" | 21:14 |
| cardoe | Or maybe Emperor Palpatine? | 21:14 |
| JayF | I have that listed in the PTG doc under the list of stuff to evaluate or rm | 21:14 |
| JayF | and that one is .... holy hell I don't want it | 21:14 |
| JayF | lol | 21:14 |
| cardoe | yeah me neither it seems like a dangerous slope. | 21:15 |
| JayF | But I also know that the security bugs have me looking for scary stuff | 21:15 |
| JayF | and that's a level of attack surface I want no part of | 21:15 |
| JayF | auth bypass in Ironic meaning you can run arbitrary redfish commands would be horrid | 21:15 |
| cardoe | So I've wrangled the $VENDOR into effectively 2 blobs of config data that need to be sent against the BMC | 21:15 |
| cardoe | Like here's a list of options and their values. | 21:15 |
| cardoe | So my thought is a runbook | 21:16 |
| cardoe | Cause at least I control loading that runbook. | 21:16 |
| cardoe | Long term I get that BMC settings spec done and the runbook calls that. | 21:16 |
| cardoe | But in the interim, I was thinking of making vendor passthru do it and have the runbook call that. | 21:17 |
| * TheJulia is loosing the reference and blames jetlag | 21:18 | |
| TheJulia | cardoe: I'm all for the Emperor laughing evilly, fwiw. | 21:18 |
| JayF | cardoe: tl;dr (to make sure I grok this) | 21:18 |
| JayF | you need to set some arbitrary BMC setting, in a vendor/OEM space, with specific timing | 21:19 |
| JayF | it's not a bios setting so you can't use that interface | 21:19 |
| JayF | and it has to be a step because you need it to happen at a specific point in the ironic flow | 21:19 |
| TheJulia | correct, and it is also manager settings which is a separate base object as well | 21:20 |
| cardoe | https://www.youtube.com/watch?v=7AnoagDv8Qo | 21:20 |
| TheJulia | and even under manager, there is an oem object in $evil_vendor_of_doom | 21:20 |
| cardoe | JayF: exactly. | 21:20 |
| cardoe | So this is actually one place I don't disagree with the OEM object. | 21:20 |
| JayF | that step is pretty scary | 21:20 |
| JayF | makes me want to require TheJulia's "limit step usage in rbac" stuff first | 21:21 |
| JayF | which may already have landed? | 21:21 |
| cardoe | It's a feature specific to that $VENDOR and DMTF has decidedly not wanted to standardize the behavior. | 21:21 |
| TheJulia | yeah, Oem even sort of makes sense | 21:21 |
| TheJulia | JayF: I believe so | 21:21 |
| cardoe | So I think it's correct to be in Oem. | 21:22 |
| TheJulia | https://www.youtube.com/watch?v=w1cn8Q1161M or https://www.youtube.com/watch?v=xUGqbk_tTT4 | 21:23 |
| JayF | I mean, it's fine, it's not a feature I love | 21:23 |
| JayF | in some ways I almost wonder if it should be in the blocked steps by default | 21:23 |
| JayF | because of that | 21:23 |
| JayF | but we have all the knobs to let people accept the risk if they want, so have fun :D | 21:23 |
| cardoe | So I don't need it to be in a deployment or anything. | 21:24 |
| opendevreview | Julia Kreger proposed openstack/ironic master: Fix file descriptor leak in neutron client connections https://review.opendev.org/c/openstack/ironic/+/1005807 | 21:24 |
| cardoe | My thought was a runbook that runs as a servicing step | 21:24 |
| TheJulia | I need to double check the bug tracker to see if ^^^ has already been filed | 21:24 |
| * TheJulia is quickly reaching no-sleep-no-spoons-left-for-day state | 21:25 | |
| * JayF talked dxterslab into considering making an appearance at the PTG | 21:26 | |
| JayF | the trap has sprung! | 21:26 |
| cardoe | TheJulia: I think you can make that much easier to follow with with | 21:28 |
| TheJulia | All good as long as we're not summoning powers combined. | 21:29 |
| TheJulia | oh yeah, I was thinking originally what you wanted to do would only be viable as a servicing or cleaning step, not a deploy step | 21:30 |
| * TheJulia fully expects the next response as a starwars movie clip | 21:31 | |
| JayF | TheJulia: don't worry, I made sure there's no Heart, that pesky Captain Planet won't bother us ;) | 21:32 |
| TheJulia | lol | 21:32 |
| cardoe | just FYI, I'm away from work tomorrow through Sunday. I'm bringing my personal laptop with me. It's just a break from work trip. | 21:35 |
| cardoe | But ya know me... OSS is my hobby | 21:35 |
| TheJulia | regarding lroussarie[m]'s kexec spec, cardoe and JayF I think you both might want to take a glance: https://review.opendev.org/c/openstack/ironic-specs/+/1005717 I think the broad idea is to make it drivable via metal3 as well an dto leverage the default happy path on deploy with an alternate sub-path. Since you can apparently do kexec without violating secureboot state now, seems more reasonable to take on as a thing in general. | 21:35 |
| cardoe | yeah I can take a peek | 21:36 |
| TheJulia | cardoe: Your sounding like me.... | 21:36 |
| TheJulia | cardoe: seriously though, enjoy the time off | 21:45 |
| TheJulia | I expect the next few days to be heavy code review and follow-up since last week was a blur | 21:45 |
| opendevreview | Julia Kreger proposed openstack/ironic master: Fix file descriptor leak in neutron client connections https://review.opendev.org/c/openstack/ironic/+/1005807 | 21:58 |
| opendevreview | Julia Kreger proposed openstack/ironic master: Fix file descriptor leak in neutron client connections https://review.opendev.org/c/openstack/ironic/+/1005807 | 21:59 |
| TheJulia | Okay, I'm mentally calling it a day. | 22:02 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!