Thursday, 2026-09-10

opendevreviewMerged openstack/networking-generic-switch master: Add opt-in MTU management for switch ports  https://review.opendev.org/c/openstack/networking-generic-switch/+/98483203:15
opendevreviewMerged openstack/networking-generic-switch master: Add opt-in port carrier bounce on bind  https://review.opendev.org/c/openstack/networking-generic-switch/+/98483303:35
opendevreviewMerged openstack/ironic-python-agent master: Erase frozen SATA SSDs by discarding them  https://review.opendev.org/c/openstack/ironic-python-agent/+/100132305:44
dtantsurcardoe: the more I look at inventory, the more desperate I get. We somehow managed to merge the completely non-standard storage_controllers collection, while mostly neglecting standard disks...12:42
cardoeThe disk info from IPA was useless though.12:42
cardoeThe storage controller stuff includes enough details to drive RAID configs.12:43
dtantsurcardoe: I can get it, but we made it even more useless12:43
cardoeWorks for me for RAID12:43
dtantsurwhile virtually all current consumers rely on IPA format (e.g. gophercloud and thus metal3)12:43
dtantsur  "disks": [12:43
dtantsur    {12:44
dtantsur      "name": "Solid State Disk 0:0",12:44
dtantsur      "size": 40008845721612:44
dtantsur    },12:44
dtantsur    {12:44
dtantsur      "name": "Solid State Disk 1:1",12:44
dtantsur      "size": 20004964761612:44
dtantsur    }12:44
dtantsur  ],12:44
dtantsurThis doesn't even have fields available in storage_controller..12:44
dtantsurI also think we managed to boot pci_devices into inventory instead of plugin_data12:47
dtantsurwhich means the accelerators hook won't work on it. Sigh. SIGH.12:48
cardoePCI-devices should be in inventory13:01
cardoeIt’s lspci data13:01
dtantsurI don't disagree with that, but it hasn't been there13:01
dtantsurEven our own code assumes it to be in plugin_data13:02
opendevreviewDmitry Tantsur proposed openstack/ironic master: Improve compatibility between IPA and Redfish inspection  https://review.opendev.org/c/openstack/ironic/+/100501013:09
dtantsurcardoe: this should be a reasonable improvement ^^13:09
* dtantsur tests13:09
opendevreviewDmitry Tantsur proposed openstack/ironic master: Improve compatibility between IPA and Redfish inspection  https://review.opendev.org/c/openstack/ironic/+/100501013:40
cardoedtantsur: how did IPA upload to plugin_data in the first place/15:07
cardoeI guess that's allowed via the API. I thought plugin_data was private to the hooks so it wouldn't be uploaded by IPA at all.15:07
dtantsurcardoe: everything that is not under the "inventory" root key ends up in "plugin_data"15:07
cardoeoh15:07
cardoeSo I'm guessing a whoops on the IPA side for pci_devices?15:07
cardoeCause I would think before hooks run that plugin_data would be entirely empty.15:08
dtantsurcardoe: less a whooops, more of an evolution of how we see this entire topic15:08
dtantsurbut yeah. we tried to make it less of a mess, and we kind of did, but it's still a mess15:08
dtantsurbtw, I got a sample for my ancient PowerEdge FC430, if that's of interest15:09
cardoeyeah. I still need to pull from mine15:09
* dtantsur is pondering constructing /dev/disk/by-path/* from Redfish15:23
JayFhttps://bugs.launchpad.net/ironic/+bug/2166877 we need to address this ASAP16:18
dtantsurugh16:18
opendevreviewDmitry Tantsur proposed openstack/ironic master: Fix device class detection for PCI devices during Redfish inspection  https://review.opendev.org/c/openstack/ironic/+/100508316:25
dtantsurcardoe: and another fun one ^^ (I'm soooo annoyed by this enum behavior)16:25
cardoedtantsur: str(enum_thing) only sucks on Python 3.10 and older. Now that we're Python 3.11+ we can fix that.16:38
dtantsurcardoe: I'm quite sure I'm on 3.12. Or do you mean we need to do something else?16:39
* dtantsur will check back tomorrow16:44
cardoedtantsur: we need to update sushy. we're using the old enum behavior16:47
cardoeWe need to tell it to use new enum behavior16:48
cardoedtantsur: btw... https://review.opendev.org/c/openstack/ironic-specs/+/1004909 if ya get time20:29
cardoecid: I think that might be related to something you looked at too?20:29
cid++, will look at tomorrow20:30
JayFhttps://bugs.launchpad.net/ironic/+bug/2166877 was on 29.0.5 which implies to me the bug was already fixed20:37
JayFI'm digging but I think we're OK20:37
JayFmarking 2166877 as invalid; 29.0.6 has teh OSSA-2026-026 fix in it which removes that issue20:42
JayFzigo: ^ JFYI; that was a debian user, if you wanna reach out20:42
zigoWho?20:43
JayFthe linked ironic bug 20:45
JayFthe reporter is a debian user, in case you wanted to reach out20:46
JayFI closed as invalid as they have 29.0.5 and the bug they reported is fixed in 29.0.620:46
opendevreviewDoug Goldstein proposed openstack/ironic-specs master: Add spec for BMC settings management  https://review.opendev.org/c/openstack/ironic-specs/+/100490920:55
zigoJayF: I do have 1:29.1.0-1~bpo13+1 packaged, but in osbpo.debian.net, not in Debian proper, I'll let him know in the bug.21:24
zigoJayF: For Debian proper, maybe I can backport the patch if you can tell me where to find it?21:25
zigoI replied in the launchpad bug.21:28
JayFthe advisory is linked there in my comment, which links to the patches21:28
zigoBTW, is 38.0.0 the final version for Hibiscus, or should I hold my breath for a new one ?21:29
JayFnew one will be incoming21:29
JayFdeadline is week of 9/20ish21:29
zigoOk.21:29
zigoThe version he's talking about should have the fix already, normally.21:30
zigoIt's green here: https://wiki.debian.org/OpenStack/CVE-2026?pow_referer=21:30
JayFI know it's happy in the branch; IDK what that user's got going on :)21:30
* zigo goes to bed, it's late here21:31
zigoNight night !21:31
TheJuliaGood morning21:55

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!