Thursday, 2026-02-05

cardoeThat’s why I was asking what I can do. Not asking ya to do anything with it. You’re just the last person to merge something.00:15
JayFI emailed the team there00:18
JayFwhich AIUI was reassigned and maybe merged my patch as a courtesy? TBH I'm not sure I realized it ever merged, my downstream hard-forked this a while back00:18
JayFany saltiness about the issue mainly boils down to the logic being in a third party library that we don't hold the keys for and was not well maintained 00:20
JayFbut as usual, Dr Seuss and his red fish are saving the day :D 00:20
JayFoh BTW; clif volunteered today to ensure we get a patch merged to retire shellinabox; we ran across a todo today to remove it this cycle (score)00:21
rpittaugood morning ironic! o/07:41
dtantsurcardoe: these are the people with ACL on proliantutils: https://review.opendev.org/admin/groups/70a7661dca62f34a214925d41655fd35c53a465c,members09:57
cardoeAh. Thanks!11:53
opendevreviewAllain Legacy proposed openstack/ironic master: Address remaining review comments for rpc methods  https://review.opendev.org/c/openstack/ironic/+/97118412:50
opendevreviewAllain Legacy proposed openstack/ironic master: Add standalone networking service installation guide  https://review.opendev.org/c/openstack/ironic/+/96647112:50
kubajjhello everyone, this might be a stupid question, when I inspect hardware (after getting rid of inspector) is it the conductor telling IPA the commands to inspect the HW with or does IPA know them? (trying to debug some formatting in the inventory data)13:37
dtantsurkubajj: it's initiated by IPA based on kernel params13:38
dtantsur(we definitely lack some sort of a "how stuff works" documentation)13:38
kubajjdtantsur: ok, thanks13:39
kubajjdtantsur: my issue is that the cpu name in inventory data is "cpu_name": "Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz\n  name: Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz"13:40
kubajjand then nova complains 😢13:41
kubajjdtantsur: found  some docs :D https://docs.openstack.org/ironic-python-agent/latest/admin/how_it_works.html#inspection (had to search by the kernel params name first)13:44
dtantsurheh, we're clearly parsing something incorrectly13:44
kubajjdtantsur: maybe, if I figure it out, I'll patch it13:48
opendevreviewJulia Kreger proposed openstack/ironic master: Add a `reattach_networking` service step  https://review.opendev.org/c/openstack/ironic/+/97341314:25
kubajjdtantsur: sorry, it was a false alarm, mea culpa14:34
dtantsurMmm? Now I'm curious :)14:35
kubajjIt took me a while to figure out how the inspection works in IPA and I just couldn't find cpu_name anywhere in the repo - this is because we have a custom HW manager which adds some more functions and this one does some parsing incorrectly14:39
dtantsurAhhh, I see.14:39
kubajjbtw, I will implement the default inspect interface soon, should it just set the inspect_interface to the config param on new nodes?14:40
dtantsurkubajj: hmm, check how other defaults are implemented, it's very strange that inspect is not one of them14:40
dtantsurit's probably somewhere in driver_factory14:40
kubajjok, will do14:40
kubajjdtantsur: do you know if `openstack baremetal introspection data save` command should still work after disabling inspector?15:32
kubajj(getting service unavailable)15:33
kubajjagain, sorry to bother you, it change to inventory15:40
dtantsurkubajj: 'baremetal introspection' are all inspector-specific, yeah16:29
dtantsur'node inventory save' is the right one16:29
kubajjdtantsur: Today is my lucky day, found another bug16:30
dtantsur\o/16:30
dtantsurHonestly, sometimes I feel like I find bugs every time I dive into some feature. I guess it's just life.16:30
kubajjyeah, can't wait until this is done. Kinda wanna close the chapter of inspector and not have to think about it (my prod regions are still running Caracal because of this)16:31
dtantsurOh, I hear you. I spent waaaay too much time on the new implementation, now folks spend way to much time on migrating to it :D16:32
kubajjok, got it to work - is it normal that if I do `inventory save`,  most of the double quotes are escaped?16:34
opendevreviewJakub Jelinek proposed openstack/ironic master: Fix retrieval of inventory data from Swift  https://review.opendev.org/c/openstack/ironic/+/97582417:11
clifanyone run into this when trying to run unit tests on latest master locally? a bunch of unit tests under ironic/tests/unit/drivers/modules/network/ started failing with ironic.common.exception.DriverNotFoundInEntrypoint: Could not find the following items in the 'ironic.hardware.interfaces.network' entrypoint: ironic-networking.17:12
clifI believe my tree is clean but maybe I screwed something up17:12
clifI've also recreated tox env17:13
clifdang it must be my tree being unclean somehow, I cloned completely fresh and things passed17:29
*** dmellado42 is now known as dmellado417:39
clifThere's a section in the docs about Node web console that seems to rely on shellinabox... is there an alternative to use there or does that mean Node web console is also abandonded by us?17:39
clifits located in docs/source/admin/console.rst:5417:43
dtantsurkubajj: I haven't noticed that, it does not sound normal17:48
dtantsurclif: I had the same problem with console some time ago. I haven't dived into it, but there seems to be some difference with the CI in how drivers are loaded.17:48
kubajjdtantsur: I will try to debug the client tomorrow then17:49
kubajjhttps://www.irccloud.com/pastebin/he1Otkka/17:50
dtantsuryeah, it's total nonsense, and it was not the case previously17:51
dtantsurI hope it's a client breakage and not API....17:52
JayFclif: there's like a wheel or .wheelinfo or something like that17:58
JayFclif: some versions of setuptools leave crap around that can break tests17:58
JayFI've experienced that, it was always that there was an opcode cache hiding somewhere17:59
opendevreviewcid proposed openstack/networking-baremetal master: Add conductor group sharding support  https://review.opendev.org/c/openstack/networking-baremetal/+/94843218:01
cardoeclif: node web console stuff is dead. The VNC stuff stevebaker[m] did is the only functional replacement.19:33
clifthanks, that helps19:59
TheJuliaPacking up my laptop, I'll try to glance at things and maybe do some reviews from the hotel bar tomorrow night, but otherwise I'll be in ireland next week.21:05
opendevreviewClif Houck proposed openstack/ironic master: Remove shellinabox  https://review.opendev.org/c/openstack/ironic/+/97586421:17
opendevreviewcid proposed openstack/ironic master: Redfish VirtualMedia NFS and SMB/CIFS Support  https://review.opendev.org/c/openstack/ironic/+/95850521:25
opendevreviewMerged openstack/ironic master: Fast track shouldn't happen when firmware update is in progress  https://review.opendev.org/c/openstack/ironic/+/97426521:29
cidTheJulia, safe travels.21:48
cidTheJulia, cardoe, I updated -> https://review.opendev.org/c/openstack/networking-baremetal/+/94843221:48
rm_work[m]So I'm considering writing a different inspector interface23:02
rm_work[m]https://github.com/openstack/ironic/tree/master/ironic/drivers/modules/inspector23:02
rm_work[m]that isn't "agent" / inband -- is that sane? The idea being instead of inspecting a live host it'd just get all the info it needs from our inventory management system23:02
rm_work[m]or is this a fool's errand and I'd be better off just writing a cron job to hit inventory management, generate an enrollment yaml, and run the standard node enroll flow23:04
JayFI question if having ironic "pull" the data in that situation is better than having it pushed in by the management software23:05
JayFbecause it's essentially a one-time sync, every second after that "inspection" the data will go outta sync from the source of truth23:06
rm_work[m]yeah I mean with an agent I think it WOULD have to be "receiving an event" from something, because that's the inspect architecture23:06
JayFso I'd approach it as event driven from the source of truth: something changes; force the API update in Ironic23:06
JayFbut do it via a third party agent polling both or through integration with your internal CMDB23:06
rm_work[m]yeah right now it seems like it'd be simpler to just do a systemd timed service that checks CMDB, diffs against ironic, generates an enrollment yaml, and does the standard enroll23:07
rm_work[m]because honestly once a day even would be fine likely23:07
rm_work[m]trying to finagle this inspector autoenrollment thing to do out-of-band seems overcomplicated23:08
rm_work[m]but just wanted to make sure that is the way you see it too23:08
rm_work[m](technically the ask of me was "see if we could make autoenrollment work")23:09
JayFI suspect the simplest answer will be "BYO-glue"23:12
JayFif your CMDB is in-house, having it event-driven instead of poll would be awesome23:12

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