Thursday, 2026-02-19

TheJuliacardoe: I'm going to try and pull the code state into devstack and start with a fresh look tomororw, it looks like my CI job attempts tickeled just a different enough config that I'll need to sit down and figure out what is really going on in not CI00:22
opendevreviewHarald Jensås proposed openstack/sushy-tools master: novadriver: rescue-based net-boot support  https://review.opendev.org/c/openstack/sushy-tools/+/97728403:00
opendevreviewHarald Jensås proposed openstack/sushy-tools master: novadriver: rescue-based net-boot support  https://review.opendev.org/c/openstack/sushy-tools/+/97728403:14
opendevreviewHarald Jensås proposed openstack/sushy-tools master: novadriver: rescue-based net-boot support  https://review.opendev.org/c/openstack/sushy-tools/+/97728403:41
kubajjgood morning Ironic! o/07:25
rpittauhey kubajj :)08:44
rpittaudking: are you using a stable release of metal3/BMO/ironic-image? to you can try checking the driver_info to see if the values are available for the node, you shoould also see something from the metal3 pod logs08:49
rpittaudking: one more thing, for all metal3 related stuff better start asking in the metal3 slack channel, just because there's more people working principally with metal3 there :)10:02
*** tkajinam_ is now known as tkajinam10:13
opendevreviewRiccardo Pittau proposed openstack/ironic bugfix/34.0: [bugfix only] remove metal3 integration job  https://review.opendev.org/c/openstack/ironic/+/97722410:32
kubajjrpittau: wait, the metal3-integration is already fixed?12:54
kubajjhttps://review.opendev.org/c/openstack/ironic/+/97700612:54
TheJuliagood morning12:54
TheJuliakubajj: that was a courtesy "have the change staged if we don't need to do anything locally" change. 12:56
opendevreviewcid proposed openstack/ironic stable/2025.1: ci: Pin `setuptools` to a range that still ships  https://review.opendev.org/c/openstack/ironic/+/97731713:04
* cid my dev environment is messed up :D13:06
TheJulia:(13:07
opendevreviewcid proposed openstack/ironic master: ci: Pin `setuptools` to a range that still ships  https://review.opendev.org/c/openstack/ironic/+/97731813:10
opendevreviewcid proposed openstack/ironic master: ci: Pin `setuptools` to a range that still ships  https://review.opendev.org/c/openstack/ironic/+/97731813:14
opendevreviewcid proposed openstack/ironic master: ci: Pin `setuptools` to a range that still ships  https://review.opendev.org/c/openstack/ironic/+/97731813:19
rpittaukubajj: yes it';s fixed in metal3-dev-env :)13:22
kubajjthat was quick13:23
rpittauwell it took us a couple of days :D13:23
rpittaufreaking docker....13:23
rpittaucid: re setuptools pin: do we need that for master too?13:26
dkingrpittau: I found my problem. The image sets up the configuration with a script (/bin/configure-ironic.sh). In there, it does a check to see whether the ramdisk and kernel actually exist as files before setting the variable. This is a race condition for me as I have an external container that downloads and manages those images, and that configuration only runs once.13:41
rpittaudking: ok, I see13:42
dkingrpittau: I'm using IrSO, and I tried to set that with my ironics' .spec.extraConfig. It creates an environment variable OS_CONTROLLER__DEPLOY_KERNEL, but I'm not sure that's getting used anywhere.13:45
rpittaudking: IrSO maps the exrta config to env variables, but you will find yourself in the same condition if the files are not there13:48
opendevreviewJulia Kreger proposed openstack/networking-generic-switch master: WIP: Crazy idea: NDFC support in NGS?  https://review.opendev.org/c/openstack/networking-generic-switch/+/96848413:51
cidrpittau, I believe master has an indirect dependency as well (?)14:02
rpittaudking: yeah14:08
dkingrpittau: It would seem to me that it would be useful to allow an override there that forces the variables, even if the files are not currently there. There would be an error eventually if they never do exist, but as is the case with any overrides, the end user would be aware of that risk. Now that I'm looking at it, it looks like the Ironic image itself would handle it, as there just needs  to be a DEPLOY_KERNEL_URL and DEPLO14:09
dkingRAMDISK_URL passed in, so maybe I need to to pass in those variables directly, rather than using extraConfig.14:09
rpittaudking: you should be able to do that in IrSO though, using extraConfig in the ironic custom resource14:12
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: be specific on exception checks  https://review.opendev.org/c/openstack/networking-baremetal/+/97697714:23
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: trunk-reconciler: Dial back logging  https://review.opendev.org/c/openstack/networking-baremetal/+/97707414:28
*** dmellado47 is now known as dmellado14:40
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: fix orphanded localnet binding  https://review.opendev.org/c/openstack/networking-baremetal/+/97732915:32
TheJuliacardoe: stevebaker[m] ^^^ would be appreciated.15:32
TheJuliaI'm hit it locally and was like "oh, yeah, that makes sense... doh!"15:33
cardoeboy I thought you read my mind on a launchpad bug I asked someone to file.15:33
cardoeWe've got an issue with the console stuff stevebaker[m]. When the user's browser connection times out or errors in anyway... on the backend the console still stays logged in and the user cannot reconnect cause it says there's another session and it prompts the original session to allow sharing but nobody can click yes.15:34
TheJuliaMy port binds, I've got dhcp from OVN which means the bridge across is working, but routing connectivity is unhappy. Its further than I'm getting upstream but a starting point on my resurrected devstack15:43
TheJuliacardoe: session sharing on the BMC?15:44
TheJulia*likely* what needs to happen on a new connection request, we need to shutdown the prior15:44
TheJuliahmm, feels like that other ovn bug even though my ports are up. WUT.15:47
cardoeTheJulia: well not trying to session share the BMC. Just re-connect after getting dropped.17:30
cardoeSo yes. Need to shutdown on a new connection request.17:31
cardoeAt the doctor’s office with the kid. So many balls in the air.17:31
opendevreviewcid proposed openstack/sushy master: Handle missing collection members gracefully  https://review.opendev.org/c/openstack/sushy/+/97734517:35
TheJuliaoh, so logout needs to be an explicit thing then17:47
TheJuliayour kid is obviously more important17:47
TheJulia:)17:47
TheJuliaand another bug found18:29
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: fix orphanded localnet binding  https://review.opendev.org/c/openstack/networking-baremetal/+/97732918:56
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: Fix port stickiness  https://review.opendev.org/c/openstack/networking-baremetal/+/97735318:56
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: Fix mech looping and supported type declaration  https://review.opendev.org/c/openstack/networking-baremetal/+/97721819:53
cardoecid: you're a dup of https://review.opendev.org/c/openstack/sushy/+/976329 would you prefer your implementation?22:15
opendevreviewDoug Goldstein proposed openstack/sushy master: skip missing collection members instead of failing  https://review.opendev.org/c/openstack/sushy/+/97632922:16
cardoeHe didn't put Closes-Bug in the commit message. I just added that. I hate that when you correct something in the commit message it drops previous reviews.22:17
cidApproved. I will abandon mine.22:20
cardoeTheJulia: for segment in context.segments_to_bind is the bug from my original push.22:22
cardoeThat's my mistake22:22
TheJuliacardoe: doing it or the lack of doing it?22:22
cardoeNo I was just reviewing your changes22:22
TheJuliaahh, okay22:23
cardoeAnd I was looking at our code. That's what we run and when I pushed the original to neutron, I wrote it wrong.22:23
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: vxlan: Fix mech looping and supported type declaration  https://review.opendev.org/c/openstack/networking-baremetal/+/97721822:23
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: WIP: CI: implement vxlan job  https://review.opendev.org/c/openstack/networking-baremetal/+/97708122:23
cardoeCause I was splitting functionality apart in our code since we've got 1 monolithic mech that needs to get broken down22:23
TheJuliaahh, yeah22:23
TheJuliano worries22:24
TheJuliaI just want to try to get it all to work in CI22:24
TheJuliaso far though! I've got it working locally (attachment wise)22:24
TheJuliajust... yeah22:24
cardoealright I think you've convinced me to move sushy inside of ironic.22:32
TheJuliahuh, what? how?22:32
cardoe*BUT* I wanna see the hardware workaround stuff move out of the scattered all over ironic to ironic/sushy22:32
TheJuliawhaaaat?!22:32
TheJuliabartender, one whiskey please.22:33
cardoeLet's just say I spent way too long just now on a call about 2 machines which are apparently the same.22:33
cardoeThey came from someone's dorm room in Texas.22:35
TheJuliabut they have "slightly different" internals?22:35
TheJuliaThat is... so very standard22:35
cardoeOh I wish.22:35
TheJulia*really* actually identical?22:35
cardoeRedfish thinks they're identical22:36
* TheJulia raises an eyebrow22:36
cardoeThey're like identical twins that nobody can tell apart except their mother can.22:36
TheJuliaeven the serials?22:36
cardoeThey're different down to the PCI revisions22:36
cardoeer the same down to the PCI revisions22:36
cardoeBut they're different22:37
* TheJulia pours cardoe a beverage22:37
cardoeOne needs to have NIC.Slot.1-1-1 enabled and one needs NIC.Slot.1-122:38
* TheJulia adds more to the glass22:38
TheJulia"Momma says we're the saaame!"22:39
cardoeOne is running OEM firmware from the NIC manufacturer and one is running firmware from the Texas hardware vendor.22:39
TheJuliaserious question, same *even* if you wipe out all the stored config on the bmc and force it to rebild it all?22:39
TheJuliaOH!22:39
TheJuliayeah22:39
TheJuliathats totally a thing22:39
cardoeThey usually don't reuse the same sub-vendor and sub-device IDs though22:40
TheJuliabecause the card speaks PLDM differently22:40
cardoeor the same PCI revision22:40
TheJuliaand gets represented differently as a result22:40
TheJulianah, its the special vendor builds22:41
TheJuliasorry, I haven't had braincells to look at any of that, heads down on vxlan stuffs23:09
TheJuliawrong window!23:10
opendevreviewJulia Kreger proposed openstack/networking-baremetal master: WIP: CI: implement vxlan job  https://review.opendev.org/c/openstack/networking-baremetal/+/97708123:32
TheJuliahttps://paste.opendev.org/show/bqClqY8lBgffbsfv6q2E/ <-- neutron hates me23:38
TheJuliaI think I know what is going on!23:51

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