Monday, 2025-10-06

opendevreviewSteve Baker proposed openstack/ironic-tempest-plugin master: WIP test graphical console novnc connection  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/96315201:11
opendevreviewSteve Baker proposed openstack/ironic-tempest-plugin master: WIP test graphical console novnc connection  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/96315202:39
opendevreviewSteve Baker proposed openstack/ironic master: Enable fake-graphical console for all hardware types  https://review.opendev.org/c/openstack/ironic/+/96315103:04
opendevreviewSteve Baker proposed openstack/ironic master: Enable fake-graphical console for all hardware types  https://review.opendev.org/c/openstack/ironic/+/96315103:59
opendevreviewSteve Baker proposed openstack/ironic-tempest-plugin master: WIP test graphical console novnc connection  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/96315204:40
*** dtantsur_ is now known as dtantsur05:54
rpittaugood morning ironic! o/06:20
queensly[m]Good morning!07:01
rpittaualegacy_: hey was this discussed with the community https://review.opendev.org/c/openstack/requirements/+/962691 ?08:30
rpittauasking because I'm not sure is a good idea putting ironic in the upper-constraints08:30
dtantsurmy first reaction was "WAIT WUT", but we also have Neutron there, so I guess there is precedent..08:53
opendevreviewnidhi proposed openstack/sushy master: Add comprehensive PCIe function support to Sushy  https://review.opendev.org/c/openstack/sushy/+/96268008:58
opendevreviewMerged openstack/networking-baremetal stable/2025.2: Fix min version of Neutron  https://review.opendev.org/c/openstack/networking-baremetal/+/96185810:35
alegacy_rpittau: I raised this a while back when I introduced this (https://review.opendev.org/c/openstack/networking-generic-switch/+/962037).  It needs to import Ironic to access a new base class for the new driver interface.  11:29
alegacy_would have been an ideal use case for ironic-lib, but...11:29
dtantsurMaybe we need to revisit ironic-lib..11:40
dtantsurI have one idea though, I'll try to formulate it after the current meeting11:41
jrosserin openstack-ansible we have to edit upper-constraints on-the-fly in order to remove services that have been added, horizon is another example11:41
jrosserpip blows up with conflicting installation of horizon and horizon==some-specific-version in that case11:42
dtantsuryeah, bifrost will be affected too most likely11:42
rpittaujrosser: we do the same in metal3, I don't like it :)11:42
jrosseryeah, agreed i don't like it also, and the error from pip is pretty confusing if you don't know whats going on11:43
opendevreviewAllain Legacy proposed openstack/ironic master: Add support for multiple JSON-RPC configuration groups  https://review.opendev.org/c/openstack/ironic/+/96059312:14
opendevreviewAllain Legacy proposed openstack/ironic master: Add force_dhcp option for managed inspection  https://review.opendev.org/c/openstack/ironic/+/96082412:14
opendevreviewAllain Legacy proposed openstack/ironic master: WIP: Add standalone networking service for ironic  https://review.opendev.org/c/openstack/ironic/+/96249312:14
opendevreviewAllain Legacy proposed openstack/bifrost master: Add support for OVS as a virtual switch  https://review.opendev.org/c/openstack/bifrost/+/96203812:15
opendevreviewAllain Legacy proposed openstack/bifrost master: Add support for standalone ironic networking  https://review.opendev.org/c/openstack/bifrost/+/96239412:16
TheJuliagood morning13:04
dtantsuralegacy_: I haven't finalized what I was thinking about, but how terrible would be to keep a copy of SwitchDriverBase in n-g-s?13:31
dtantsurinitially, I thought we would have two separate services talking via JSON RPC.. but I guess we're moving away from this model?13:33
dtantsurOr should n-g-s (instead of Ironic) become the core of this service (with a potential to split it away entirely)?13:33
TheJuliacentral driver code and exceptions are sort of the issue13:35
dtantsurI guess I had a different mental model of all this13:35
TheJuliaI sort of did as well13:35
dtantsurI thought that everything Ironic-related stays in Ironic (its network interface)13:35
dtantsurand n-g-s or $something turns into a JSON RPC service that knows nothing about nodes and ports but operates on a lower level13:36
TheJulia$something ended up in ironic, and n-g-s part is pluggable13:36
TheJuliabut then another part is in ngs13:36
TheJuliabecause it exposes it as a driver13:36
dtantsurso, not JSON RPC, just ironic+ngs?13:36
TheJuliano, with json rpc13:37
dtantsurwut13:37
dtantsurthe main point of JSON RPC was to avoid problem we're discussing now: coupling n-g-s and ironic in code13:37
dtantsurif we don't do that, why have an RPC?13:37
TheJuliaironic network interface -> jsonrpc -> ironic-networking service -> ironic-networking-service drivers -> ngs-driver -> switch driver -> switch13:37
dtantsurwhy not: ironic network interface -> ironic-networking-service drivers -> ngs-driver -> switch driver -> switch13:38
TheJuliabecause there seems to be a desire to make it pluggable to use a different plugin  for the switch config13:38
TheJulia(if anyone says networking-ansible, I'm going to just go in to the corner and cry)13:38
dtantsurwe can do it with entrypoints as it's already done in alegacy_'s patches13:38
TheJuliadtantsur: parent driver class is in ironic13:39
TheJuliaalong with the exceptions he wants to raise13:39
dtantsurI guess I had the same picture as you in mind, but I assumed that ironic-networking service is disjointed from ironic and its concepts13:39
TheJuliait is, minus the desire to have type and exact same class matching13:39
opendevreviewAllain Legacy proposed openstack/bifrost master: Add support for standalone ironic networking  https://review.opendev.org/c/openstack/bifrost/+/96239413:40
dtantsuroh, I understand what is odd to me. I thought that JSON RPC will take the role of the plug-in point.13:40
dtantsuressentially, ironic network interface -> jsonrpc -> ngs-service -> switch driver -> switch13:40
dtantsuror, ironic network interface -> jsonrpc -> something else -> switch driver -> switch13:41
TheJuliaYeah, I was always thinking, no matter how we end up doing it, we didn't try for eact matching and just figured it what was going on via jsonrpc, not explicit exceptions. Granted, there are huge downsides to that crazy idea.13:45
dtantsurMaybe, for the sake of keeping alegacy_'s sanity, we table the whole JSON RPC idea for now and just make it all direct Python calls...13:46
dtantsurI'd still probably just copy over the base class (and avoid using isinstance in Ironic)13:46
dtantsurmaintaing Ironic version compatibility may be fun13:46
opendevreviewnidhi proposed openstack/ironic master: Add PCIe function fields to redfish inspection  https://review.opendev.org/c/openstack/ironic/+/96317913:47
TheJuliadtantsur: that doesn't make any of it any better though13:49
TheJuliathe issue is the explicit pluggable driver interface being added below13:49
dtantsurit reduces the operational complexity quite a bit13:49
dtantsurI mean, we're not getting ironic/n-g-s separation, so why complicate everything with an RPC bus between ironic and ironic?13:49
TheJuliaIt eliminates the ability to build trust, but that is not his product concern13:50
alegacy_i thought the point of the new service was to gain the ability to give that to another department and let them run and manage it.13:50
TheJuliaI feel like, at this point, that is change for change's sake and not addressing the root challenge13:50
TheJuliaalegacy_: yeah, exactly, allow for a barrier to exist13:51
dtantsurthat's a pure theory13:51
dtantsurand if we want the barrier, we cannot be coupling the code so tightly13:51
alegacy_IMO, that new service does need to have the ability to pull in drivers as a pluggable interface and that's where that base class came from.  i.e., not just tightly coupled to ngs.13:52
TheJuliaIts operational reality based upon numerous operators providing feedback13:52
TheJuliaThat is where the extra complexity seems rooted, fwiw13:53
TheJuliaat least, the complexity which we're finding challenging and going back and forth on13:54
dtantsurWell, I guess Metal3 is a bit specific case in this sense13:54
dtantsurAnyway13:54
dtantsurJust let's really not import Ironic :)13:54
* TheJulia sighs13:55
dtantsurSimilarly, importing Ironic is not a problem for Metal3 (we have one image anyway, and everything is run together), but it seems to be a problem for other folks13:57
alegacy_ironic-lib would have been perfect.  alternatively, i could look into moving the new entry point from ngs back into ironic and have it import ngs.  13:57
alegacy_It would still be a pluggable driver, but unfortunately it would never be possible to have a driver live purely outside of Ironic.13:57
alegacy_IMO, the current way all of the drivers inside of Ironic works is a bit limiting in that way... not possible to externalize them without Ironic.13:58
TheJuliawho is saying its a problem?13:58
dtantsuryeah, it's also not a clean solution: then we need to add n-g-s to requirements13:58
alegacy_exactly13:58
alegacy_not saying its a problem perse, but it is definitely a limitation... anyone wanting to implement their own driver needs to get it into the Ironic repo.13:59
TheJuliawell, in the code I looked at last week, the base is in the ironic repo14:00
dtantsurTheJulia: jrosser for openstack-ansible14:00
dtantsureveryone will have to hack upper-constraints to be able to install ironic14:00
TheJuliaoh, ugh14:00
TheJuliayeah, if they are trying to enforce it, yeah14:00
TheJuliathat is a good point14:01
TheJuliaAnyway, lets stop trying to push the problem around the plate14:01
TheJuliawhat is the cleanest path?14:01
TheJuliaa small library?14:01
dtantsuragain, why cannot we copy the base class?14:02
TheJulia... I'm fine with that14:02
* dtantsur is in a meeting and might have missed some answers14:02
opendevreviewnidhi proposed openstack/sushy master: Add comprehensive PCIe function support to Sushy  https://review.opendev.org/c/openstack/sushy/+/96268014:02
TheJuliaI floated that as well, but... yeah14:02
jrosserexample for horizon mess is here https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/77028314:02
JayFJFYI I'm on a sick day today 14:10
dtantsurJayF: get better :(14:11
kubajjdtantsur: or any IPA pro, is there any simple way of passing information between cleaning steps? (other than global variables - we are trying to avoid this)14:20
TheJuliakubajj: trying to save what the root device should be according to a prior step for a later step to pickup?14:21
kubajjTheJulia: yes, exactly14:22
TheJuliauhhhh nothing in the steps themselves, but you can reset/save it if memory serves14:23
* TheJulia looks14:23
TheJuliahttps://github.com/openstack/ironic-python-agent/blob/master/ironic_python_agent/hardware.py#L1840-L1902 14:25
TheJuliaThat is the block of code you want to influence, you, could.. in theory modify the cached node, but that is overwritable. I think the right thing would be to create a new flag of "this is it" which could just return quickly if set.14:25
TheJuliaand a prior step could save that value on the hardware manager.14:26
TheJuliaSeems like that ?might? work14:26
kubajjTheJulia: yes, so we will just move the global variable to the HW manager, sounds like a plan, thanks14:27
TheJuliakubajj: the cached node is a similar concept, fwiw, but it can be refreshed in some steps14:30
dtantsurkubajj: I don't think so..14:32
* dtantsur -> short walk, may miss the meeting14:32
opendevreviewAllain Legacy proposed openstack/bifrost master: Add support for standalone ironic networking  https://review.opendev.org/c/openstack/bifrost/+/96239414:46
clifo/15:00
alegacy_o/15:00
TheJuliaI guess it is time15:02
TheJulia#startmeeting ironic15:02
opendevmeetMeeting started Mon Oct  6 15:02:50 2025 UTC and is due to finish in 60 minutes.  The chair is TheJulia. Information about MeetBot at http://wiki.debian.org/MeetBot.15:02
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.15:02
opendevmeetThe meeting name has been set to 'ironic'15:02
TheJuliaSo, this will be our last meeting until after the PTG!15:03
kubajjo/15:03
TheJuliaThe PTG is the week of the 27th. Our next meeting will be November 3rd.15:03
cido/15:04
TheJulia#topic Announcements / Reminders15:04
TheJuliaA general reminder to review items tagged with the hashtag "ironic-week-prio"15:04
TheJulia#link https://tinyurl.com/ironic-weekly-prio-dash15:04
rpittauo/15:05
TheJuliaFor this next release, gazpacho, the schedule has been posted.15:05
TheJulia#link https://releases.openstack.org/gazpacho/schedule.html15:05
TheJuliaGenerally, in terms of the release schedule, the PTG is a little later into the cycle than we're typically used to based upon the past history. Also, Ironic will likely need to cut it's final release in R-2 which is the third week of March.15:06
TheJuliaAny questions or concerns regarding the schedule?15:07
TheJulia#topic Working Group Updates15:08
TheJuliaAny working group updates for us this week?15:08
alegacy_networking... 15:08
alegacy_First, thanks to those that have provided review comments so far.  I've done my best at addressing most of those.  Still a couple to go...but making progress15:08
TheJuliaThanks alegacy_!15:09
alegacy_I seem to have hit a snag with the whole Ironic import business though.15:09
TheJuliaYeah, some discussion earlier regarding that15:09
alegacy_Should I maybe organize a separate call to kick that around a bit?  I feel like there are some questions about the direction...15:09
TheJuliaI think that will likely be a good idea15:09
TheJuliaI also suspect we're at the low end of critical mass for that topic here at the moment.15:10
TheJuliaOnward?15:10
alegacy_ya... i'll try to find a suitable timeslot and post it after the meeting.  15:11
TheJuliaOkay15:11
TheJulia#topic Discussion Topics15:11
TheJuliarpittau has a single topic this week.15:12
TheJuliarpittau: would you like to bring it up or do you want me to?15:12
rpittauyep, just a reminder that I'm going to count the number for the ironic dinner at OIS tomorrow at the latest15:12
TheJulia#link https://etherpad.opendev.org/p/Ironic_OIS_community_dinner_in_Paris15:12
rpittauand then put a reservation :)15:12
rpittauwe're 12 at the moment15:13
TheJulia13 by my count15:13
rpittauI can't count :D15:13
rpittauthat's all15:14
TheJuliaIt happens :)15:14
TheJulia#topic Bug Deputy Updates15:14
TheJuliacid: I think you've been quietly updating the agenda (thanks!)15:14
cidYep15:15
cidI was bug deputy for the week.15:15
cidAlready updated https://wiki.openstack.org/wiki/Meetings/Ironic. 15:15
cidThere were 4 newly filed bugs, and 1 RFE15:15
cidThe RFE is about allowing setting BOOTIF whether or not there is a VIF15:15
cidhttps://bugs.launchpad.net/ironic/+bug/212664415:15
cidI think TheJulia already commented on it on LaunchPad.15:15
TheJuliaYeah, I think it makes sense to do15:16
TheJuliaI think we likely *ought* to rename some of the internals15:16
opendevreviewMerged openstack/ironic-python-agent-builder master: ci: Remove ironic-inspector related ci job  https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/96299715:16
TheJuliai.e. "pxe_interface"15:16
TheJuliaSince PXE, it is not.15:16
TheJuliaAny questions or dicsussion regarding the RFE?15:18
* TheJulia suspects the coffee maker needs to be turned on and coffee distributed15:18
TheJuliaSince we're on hold the next few weeks, the expectation is that contributors do take time to sporadically look at bugs and triage as time permits. 15:19
TheJuliaAs such, no need to name a bug deputy for the next few weeks.15:19
TheJulia#topic Open Discussion15:19
cid++15:20
TheJuliaDo we have anything else to discuss today? Aside from making more coffee!15:21
esimoneHi, first time here... there is a backport in stand by due to a CI issue (https://review.opendev.org/c/openstack/ironic/+/962993)15:21
esimoneI tried to investigate the issue (web server restart fault on devstack setup) and commented15:21
rpittauesimone: looks ok to me15:22
TheJuliayeah, I'll approve it once my browser is signed in15:23
TheJuliaDone!15:23
TheJuliaAnything else?15:23
* TheJulia takes silence as time for everyone to have more coffee.... ;)15:25
TheJuliaThanks everyone!15:26
TheJulia#endmeeting15:26
opendevmeetMeeting ended Mon Oct  6 15:26:59 2025 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)15:26
opendevmeetMinutes:        https://meetings.opendev.org/meetings/ironic/2025/ironic.2025-10-06-15.02.html15:26
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/ironic/2025/ironic.2025-10-06-15.02.txt15:26
opendevmeetLog:            https://meetings.opendev.org/meetings/ironic/2025/ironic.2025-10-06-15.02.log.html15:26
opendevreviewVerification of a change to openstack/ironic master failed: ci: remove tinyipa job alias usage  https://review.opendev.org/c/openstack/ironic/+/96300115:28
alegacy_how about 1500UTC tomorrow for an opportunity to discuss the Ironic--NGS import issues and related topics?15:31
TheJuliaThat time is clear for me15:32
opendevreviewJakub Jelinek proposed openstack/ironic-python-agent master: WIP: Implement functionality for the is_root_volume RAID config  https://review.opendev.org/c/openstack/ironic-python-agent/+/96320015:33
kubajjTheJulia: does ^ this look any sensible? (wanted to push a draft before we prepare any tests)15:34
kubajj(we did not test it yet, but will do tomorrow)15:35
alegacy_Ok, tentatively let's go with 1500UTC then... This is the link: https://meet.google.com/mzm-xtky-jbr  ... i'll re-post tomorrow an hour or so beforehand as a reminder.15:36
TheJuliakubajj: yeah, along those lines is what I was thinking15:37
kubajjTheJulia: nice, will tests and get it ready in the coming days15:40
opendevreviewMerged openstack/ironic stable/2025.1: Allow multiple HTTP redirects for image source  https://review.opendev.org/c/openstack/ironic/+/96299316:37
opendevreviewJulia Kreger proposed openstack/ironic master: CI: disable multi-n-cpu on ironic multinode job  https://review.opendev.org/c/openstack/ironic/+/96298017:19
TheJuliaI hope that sorts out the intermittent failure issues for that job17:31
opendevreviewJulia Kreger proposed openstack/ironic master: ci: remove snmp ci job  https://review.opendev.org/c/openstack/ironic/+/96321517:44
opendevreviewJulia Kreger proposed openstack/ironic master: ci: remove metalsmith ci job invocation  https://review.opendev.org/c/openstack/ironic/+/96321617:50
TheJuliastevebaker[m]: doesn't https://review.opendev.org/c/openstack/ironic/+/963151/3/ironic/drivers/fake_hardware.py seem like the opposite of what is wanted?18:07
stevebaker[m]TheJulia: It has moved to a parent class. Uh which isn't in that change :P I'll refresh it in a bit18:12
TheJuliaokay! :)18:12
TheJuliaI was like 'wait, what?!"18:13
TheJuliaThanks!18:13
opendevreviewSteve Baker proposed openstack/ironic master: Enable fake-graphical console for all hardware types  https://review.opendev.org/c/openstack/ironic/+/96315119:03
TheJuliaSo, an observation, it looks like the ovn jobs seem to be fail happy, like dhcp is not working. I'll check some more logs tomorrow, it is certinly weird20:43
opendevreviewSteve Baker proposed openstack/ironic-tempest-plugin master: WIP test graphical console novnc connection  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/96315221:01
opendevreviewSteve Baker proposed openstack/ironic master: Enable fake-graphical console for all hardware types  https://review.opendev.org/c/openstack/ironic/+/96315122:55

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!