15:01:47 <TheJulia> #startmeeting ironic 15:01:47 <opendevmeet> Meeting started Mon Jul 21 15:01:47 2025 UTC and is due to finish in 60 minutes. The chair is TheJulia. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:47 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:47 <opendevmeet> The meeting name has been set to 'ironic' 15:01:48 <TheJulia> o/ 15:01:50 <opendevmeet> rpittau: Error: Can't start another meeting, one is in progress. Use #endmeeting first. 15:01:52 <alegacy> o/ 15:01:58 <rpittau> o/ 15:02:02 <dtantsur> o/ 15:02:03 <masghar> o/ 15:02:03 <TheJulia> #chair rpittau 15:02:03 <opendevmeet> Current chairs: TheJulia rpittau 15:02:10 <rpittau> thanks :D 15:02:10 * TheJulia lets rpittau chair ;) 15:02:11 <iurygregory> o/ 15:02:21 <rpittau> Hello everyone! 15:02:21 <rpittau> Welcome to our weekly meeting! 15:02:26 <cid> o/ 15:02:30 <clif> o/ 15:02:33 <rpittau> The meeting agenda can be found here: 15:02:33 <rpittau> https://wiki.openstack.org/wiki/Meetings/Ironic#Agenda_for_July_21.2C_2025 15:02:38 <queensly[m]> o/ 15:03:03 <rpittau> lots of people today, good good :) 15:03:07 <rpittau> #topic Announcements/Reminders 15:03:22 <rpittau> Standing reminder to review patches tagged ironic-week-prio and to hashtag any patches ready for review with ironic-week-prio: 15:03:22 <rpittau> #link https://tinyurl.com/ironic-weekly-prio-dash 15:03:43 <rpittau> quite some patches with +2 missing +W 15:04:16 <rpittau> there's also an intruder patch from neutron :D 15:04:42 <TheJulia> I commented on the intruder ;) 15:04:46 <rpittau> thanks! 15:04:52 <TheJulia> specifically to encourage it 15:05:49 <rpittau> I guess we'll most of the patches there merging soon 15:05:58 <rpittau> forgot a "see" there 15:06:18 <rpittau> anyhow! 15:06:30 <rpittau> 2025.2 Flamingo Release Schedule 15:06:30 <rpittau> #link https://releases.openstack.org/flamingo/schedule.html 15:06:42 <rpittau> we're in the second half of the dev cycle 15:06:48 <rpittau> at R-10 15:06:53 <rpittau> just 10 weeks to go 15:07:11 <rpittau> I'm going to propose bugfix branches this week 15:07:45 <rpittau> anything else to announce/remind ? 15:07:47 <TheJulia> Okay, that will likely be the last bugfix/branch with eventlet if we're able to move futurist changes along 15:07:53 <rpittau> oh yeah 15:07:54 <rpittau> right 15:07:55 <rpittau> great! 15:08:08 <rpittau> (had so many issues downstream with eventlet) 15:08:15 <TheJulia> we can talk about eventlet when we get there 15:08:20 <rpittau> yeah 15:08:25 <rpittau> moving on! 15:08:33 <rpittau> #topic Working Group Updates 15:08:54 <rpittau> any update from standalone networking? 15:09:09 <alegacy> yep, making good progress here. a few things I want to raise here today. 15:09:57 <alegacy> First, I've got things roughed in from the API down to the switch. Wondering if there's any interest in a short demo to validate the approach before I start cleaning things up and submitting patches 15:10:14 <TheJulia> a quick demo would be awesome 15:10:33 <TheJulia> but at your convenience obviously 15:10:37 <cardoe> I had a question about some of the specs that seem to be tied in. 15:10:46 <alegacy> k, i'll follow-up on this outside of meeting to find a suitable time. 15:10:54 <cardoe> We've got specs for port binding that reference the standalone work. 15:11:03 <cardoe> But there's almost two competing ones. 15:11:24 <cardoe> There's adding of a "vendor" field but no clarity on how that will be used or defined. 15:11:25 <TheJulia> I wouldn't describe them as competing 15:11:38 <TheJulia> Its entirely differnet problem spaces and use patterns 15:11:47 <opendevreview> Merged openstack/ironic-python-agent master: doc: How hardware managers ignore certain devices https://review.opendev.org/c/openstack/ironic-python-agent/+/955469 15:11:54 <alegacy> ya, i didn't think there would be any overlap there 15:11:55 <TheJulia> For example, one pattern is nova integrated, another is "operator knows what they want" 15:12:36 <TheJulia> another is "operator is asking for baremetal and asking nova to bind ports and is explicitly using tags because they know all the secrets in between because they the operator is the user, not an end user" 15:12:45 <TheJulia> And yes, that is all very confusing and not obvious 15:13:18 * TheJulia would be happy to go rev specs with clarifying text, fwiw 15:13:22 <cardoe> So along the lines of the demo. Just wanted to get more definition on use case and also consideration of failure case. 15:13:59 <alegacy> use case for the demo specifically... or use cases for this feature in general? 15:13:59 <cardoe> But maybe I'm speaking to the wrong section. 15:15:20 <cardoe> Not for the demo specifically. 15:15:34 <cardoe> I derailed the convo. Not my intention back to being quiet. 15:15:40 <rpittau> no worries :) 15:15:40 <alegacy> :) 15:15:43 <rpittau> should we move on? 15:15:50 <alegacy> a couple more things from me... 15:15:55 <rpittau> alegacy: please :) 15:16:00 <TheJulia> heh, cardoe your pointing out an easy to confuse area, I'll sync up with you after the meeting 15:16:12 <alegacy> I've been using bifrost to do some testing. 15:16:31 <alegacy> but it doesn't have any support for virtual switches that can be controlled by ngs. 15:16:49 <alegacy> has there been any discussions in the past to add OVS into the bifrost testenv mix? 15:17:05 <alegacy> is that a good/bad idea to propose? 15:18:41 <TheJulia> It seems possible, we have some of the base structure in the ironic devstack plugin. Honestly there is no real reason not to aside from just its a bit more complexity. I think when I last approached adding neutron support in bifrost I ended up with a bunch of playbooks some of which did similar things but I also ended up building a lab locally with hardware 15:19:33 <alegacy> ok, maybe we can discuss this a bit more outside of this meeting. likely need a plan to CI this feature... 15:19:45 <alegacy> I'll take a look at what it would take. 15:19:52 <TheJulia> it would make the most sense to do such for CI, specifically 15:20:01 <alegacy> ok 15:20:08 <alegacy> 1 more topic from me. 15:20:55 <alegacy> Looking at the node transitions and the calls to add_*_network... it looks like there are states that either don't expect any network access or assume that the switch is already setup. 15:21:22 <alegacy> For instance, when the node is available it doesn't look like there are any calls to setup any network stuff. 15:21:32 <TheJulia> Correct 15:21:39 <TheJulia> it is considered detached 15:21:43 <alegacy> am I reading that right? Is network access not required/expected in that (and other) state(s). 15:21:52 <alegacy> ah 15:22:10 <alegacy> also, when it is first enrolled... same? no net access required/expected? 15:22:38 <TheJulia> correct, we don't want to leave the node attached needlessly in case it gets powered on or someone also sets up a DHCP+PXE server in a specific network (outside of our control) and tries or manages to make use of the machines 15:22:56 <TheJulia> yeah, no networking generally expected. 15:23:19 <alegacy> ok, that explains what I have been seeing. Thought maybe there were some missing hooks somewhere. 15:23:22 <alegacy> thanks 15:23:32 <TheJulia> dtantsur: you might have thoughts here given.... I'm drawing a blank on what the protocol is in the bmc layer to share an ethernet port with the bmc 15:23:39 <TheJulia> dtantsur: I blanked it from my mind in horror.... 15:23:48 * dtantsur reads scrollback 15:24:09 <alegacy> that's it from me for standalone networking 15:24:18 <rpittau> ack, thanks alegacy 15:24:30 <dtantsur> attachment in available depends on fast track 15:24:37 <dtantsur> but this should be handled by Ironic already 15:24:54 <alegacy> k, i'll double check and follow-up with you offline 15:25:10 <cardoe> You need some kind of "safe" state. 15:25:13 <rpittau> anythign else for the standalone networking ? 15:25:18 <jrosser> TheJulia: nc-si ? 15:25:23 <TheJulia> jrosser: that! 15:25:26 * TheJulia hides in terror 15:25:40 <rpittau> lol 15:25:44 <cardoe> But the question for "safe" state is do you want it configured so that LLDP works so that inspection works to find the linkage? 15:26:15 <alegacy> inspecting doesn't apply a network config 15:26:21 <dtantsur> jrosser: oh noes 15:26:22 <alegacy> sorry, does 15:26:23 <TheJulia> cardoe: inspection is an entirely separate state in the state machine, and also doesn't really apply networking config 15:26:31 <alegacy> it does apply network config 15:26:42 <dtantsur> yeah, there is inspection_network and all the business 15:27:07 <TheJulia> Yeah, exactly 15:27:19 <TheJulia> So onward to eventlet removal? 15:27:23 <rpittau> yep 15:28:25 <TheJulia> So, regarding eventlet removal, we're super close, but also have a little bit of a ways. We're in the phase of finding the things out we didn't know or expect, and overall we seem to be heading in a good direction. dtantsur has been awesome and driving forward a change in futurist which we'll end up using. The profile will be a little different but I don't think we're going to have to do much more aside put some warnings 15:28:25 <TheJulia> about increased memory usage 15:28:41 * dtantsur nods 15:28:44 <TheJulia> and by usage, I mean more around allocations. I fully expect we might find some things we never expected either, but one step at a time. 15:29:19 <TheJulia> (not the allocations feature, but thread launch reserved stack size, which I'm thinking of setting to the smallest available value) 15:30:01 <TheJulia> Any questions regarding eventlet removal? 15:31:34 <rpittau> looks like we're good 15:32:03 <rpittau> #topic Discussions topics 15:32:12 <rpittau> we have 2 topics today 15:32:22 <rpittau> first one from queensly[m] :) 15:33:06 <queensly[m]> Hi everyone, I’m 8 weeks into the Outreachy internship and would like to give an update on my project. I’m working on setting a bare metal host's clock remotely using Redfish.... (full message at <https://matrix.org/oftc/media/v1/media/download/Adi0_wK6ZUl3pIczSde2O0zEYgSL0rwg6DoU5WekkLulsOjibVB4GC349Js5zzCHpLBD9jl8oFfTS2fQQOWqHIZCeYdbcJbgAG1hdHJpeC5vcmcvcUhjSlpJekpDaVFIbVpIUGxKd05qY1Jy>) 15:34:37 <TheJulia> queensly[m]: cool, how can we be of assistance? 15:34:37 <queensly[m]> So far, I have done the following:... (full message at <https://matrix.org/oftc/media/v1/media/download/AWJV0NBsoIoK-tlCz01-HwjVZCVDAfWfgBH2GdpoKl0eOj5AalTRt1-SmWwwyhEvOVXYvfCkI12L6XC1DkmzZ9BCeYdbhtrQAG1hdHJpeC5vcmcvTFloUk5MT1VpSUxkV2xqU01sbmxxWlNM>) 15:35:05 * cid I just blinked and we moved on already D. I was just about to ask, the checkpointed task approach, that's till on the table(?). I'm getting around a PoC this week. TheJulia, dtantsur. I'm effectively playing catching currently with all things eventlet as a result of doing non-eventlet removal things the past week. 15:36:47 <queensly[m]> TheJulia thanks. I need reviews on my current patches. I think that would help me know what to add or remove. 15:37:27 <dtantsur> cid: I'm not sure what you're referring to (re checkpointed tasks) 15:37:35 <cid> You're doing great, queensly[m], o/ 15:37:37 <TheJulia> cid: I think it is something we might want to look at and consider, but I don't think that is something we need to do right now(). It does solve a lot of things later on and with some more complex flows thought 15:37:48 <dtantsur> ahh, remembered 15:38:19 <queensly[m]> cid: Thank you 15:38:21 <dtantsur> I agree with Julia, we need this, but it's not a blocker priority 15:38:28 <opendevreview> Merged openstack/ironic-tempest-plugin master: Add Tempest API tests for inspection rules https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/951761 15:38:31 <TheJulia> queensly[m]: Yes, your doing great. I'll try to look at your ironic change later today 15:38:46 <cid> Oh, good thing I asked. 15:39:08 <queensly[m]> Alright thank you. 15:39:24 <TheJulia> cid: great you asked, but yeah, ironic works surprisingly well at the core, and oddly even better in some ways, but the big issue we've spotted is the power state sync 15:39:37 <TheJulia> And... it is distinctly possible we'll find *more* issues as time moves on 15:39:39 <rpittau> thanks queensly[m] for your contributions :) 15:39:49 <TheJulia> but they might not be "in the next 2 months" sort of efforts 15:40:12 <TheJulia> but they might be "oh wow, this is bad, we need to backport this thing which feels like a feature but it fixes a bug" 15:40:42 <queensly[m]> rpittau: thanks :) 15:40:50 <cid> TheJulia, that 'll already be a heavy lift, :) 15:41:16 <TheJulia> Indeed, but we're definitely heading into a better place overall by removing eventlet. 15:41:34 <rpittau> I agree! :D 15:41:37 <queensly[m]> I have the following ongoing:... (full message at <https://matrix.org/oftc/media/v1/media/download/Adh3F3hIIZ2UUE3bxIOlu2cpI4-PMPyZsf6BaqsG_dcJKYZ4D515tjJxM-aCUwE-LC44ND3epGxyAfymzIXxhaxCeYdb7TcAAG1hdHJpeC5vcmcvQ3lBVXZuQ21QSFFNcVNYWnpPcWFFekt2>) 15:41:50 <dtantsur> meanwhile, I've added https://review.opendev.org/c/openstack/futurist/+/955217 to our list, and more eyes are welcome 15:42:04 <cid> Kudos, TheJulia, dtantsur, just ping me if you think there's anywhere I could help out with. 15:42:32 <cardoe> speeaking of bifrost... what do we want to do with https://review.opendev.org/c/openstack/bifrost/+/946724 ? it's sitting in the weekly queue for a bit. 15:42:53 <TheJulia> dtantsur: I'll spin it up after the meeting 15:43:11 <rpittau> shoud lwe move on? 15:43:17 <queensly[m]> I would like to thank my mentors rpittau and masghar for their support so far and also to the community. I'm happy to answer any questions or suggestions you may have. 15:43:39 <rpittau> :) 15:43:41 <queensly[m]> rpittau: Yeah, I'm done with my update. 15:43:50 <rpittau> thanks! 15:43:57 <cid> tks 15:44:12 <masghar> thanks for your contributions queensly :) 15:44:14 <queensly[m]> You're welcome :) 15:45:03 <rpittau> next topic is from me, it's just yet another headsup that we're going to remove support for python 3.9 soon 15:45:03 <rpittau> ideally after the next bugfix branches are cut 15:45:03 <rpittau> what's missing now is just moving cs9 jobs in bifrost to non-voting 15:45:48 <TheJulia> Are we pinning the CS10 usage to specific providers? 15:45:56 <rpittau> related to that, the ipa-builder patch to build ipa ramdisk with CS10 is now ok with CI, the jobs are pinned to el10 hardware compatible nodes https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/950965 15:46:09 <TheJulia> cool cool 15:47:37 <rpittau> a reminder that we're using python 3.12 in the current CS9 ipa ramdisks 15:48:12 <rpittau> so migrating to CS10 is not super urgent 15:48:47 <rpittau> I think that's all 15:48:47 <rpittau> any questions/doubts/concerns? 15:48:56 <TheJulia> none on my end 15:49:40 <rpittau> ok, moving on! 15:49:55 <rpittau> anything for BUG updates or RFEs ? 15:51:25 <rpittau> alright! 15:51:31 <rpittau> any Open Discussion topics? 15:52:19 <TheJulia> I don't have anything right now 15:52:41 <rpittau> then I'm declaring the meeting over! 15:52:41 <rpittau> Thanks everyone! 15:52:47 <rpittau> #endmeeting