Friday, 2025-09-19

cardoeiurygregory: yeah nice find.00:03
cardoeThe job is there and it shows up.00:03
cardoeJayF: https://docs.openstack.org/hacking/latest/user/hacking.html H306 doesn’t seem to00:04
cardoeHave the special OpenStack but not us group. Which means I bet I can configure ruff to comply with H30600:04
JayFooh03:14
cardoeJust saying... I love sorting imports in 0.10 seconds vs 15 seconds.03:27
opendevreviewMorten Stephansen proposed openstack/ironic-python-agent stable/2024.1: Fix for motherboards where efibootmgr returns UTF-8.  https://review.opendev.org/c/openstack/ironic-python-agent/+/95794808:12
*** kubajj_ is now known as kubajj10:44
zigoWho's taking care of networking-baremetal ? Is this under the Ironic team ?10:50
zigoBecause version 7.0.0 of networking-baremetal doesn't work with Neutron 26.0.0 (ie: epoxy), and its requirement must be 27.0.0~rc1 (or maybe the b1, I didn't check for it).10:51
opendevreviewStephen Finucane proposed openstack/ironic master: Migrate setup configuration to pyproject.toml  https://review.opendev.org/c/openstack/ironic/+/96178811:23
iurygregorygood morning ironic11:45
iurygregorydtantsur, TheJulia idrac10 https://paste.opendev.org/show/bvrvdimREnVRlM4VeADv/12:12
iurygregoryI think I'm almost there...12:12
dtantsurYou sure the image is actually good and reachable?12:12
iurygregorydtantsur, yup, I can access it in other machines http://<ip>:8080/centos9-image.qcow212:13
dtantsurand the GPT-vs-MBR is also as expected?12:14
iurygregoryinspection finished without problems, now it's a problem when i trigger ./bifrost-cli deploy -e @baremetal-install-env.json12:14
iurygregorynot 100% sure, going to start debugging now, need more coffee :D12:15
iurygregorybut at least I was able to enroll the node and it went to available https://paste.opendev.org/show/bFaLNVvDRvXIVNouG0pg/12:15
iurygregoryI think is a good sign at least...12:16
dtantsurindeed12:20
iurygregorythis was with an ironic hack for slot1 + TheJulia patch https://review.opendev.org/c/openstack/sushy/+/961730 12:21
iurygregoryand a fresh bifrost install with stable 2025.112:22
opendevreviewTakashi Kajinami proposed openstack/sushy master: Restore flake8-import-order  https://review.opendev.org/c/openstack/sushy/+/95312412:54
TheJuliaiurygregory: that ain't your BMC controller, that is the host has a 4k raid controller and you likely used a 512 byte block aligned image13:31
TheJuliathey are mutually exclusive layouts13:31
iurygregoryTheJulia, ohhhhhhhhh13:32
iurygregoryso I need to do some magic for it to work? 13:32
TheJuliaWe hit some of this on the openstack side with newer raid controllers based on the latest generation of Broadcom LSI Fusion MPT SAS controllers13:33
iurygregorydtantsur, in ocp do we handle such cases? <eyes>13:33
TheJuliayou literally just have to use a 4k block aligned image13:33
TheJuliayour not rolling an image out13:33
TheJuliain OCP, quite like this13:33
TheJuliauhh... stevebaker did some work on DIB to be able to convert images13:33
TheJuliabut you can also pass DIB_BLOCK_SIZE=4096 to a diskimage-builder command as an environment variable and it will align the blocks properly for that hardware13:34
TheJuliaThe root issue, unfortunately, is the way the GPT partitioning standard is written13:34
TheJuliaIt doesn't say strucutres are positionally aligned by byte location, but by logical blocks based upon the device13:34
TheJuliaSo for 4k, your partitioning data starts at like block 2, for 512 byte blocks its like block 16 or 32 (offhand, I don't remember exactly), but if you stop and do the math, you'd realize that it leaves a ton of extra space which structurally makes it impossible to have a universal image13:36
TheJuliait sort of cascades from there, but hopefully you get the basic idea13:36
* TheJulia sips coffee in a vain attempt to become human13:39
* iurygregory searches for DIB_BLOCK_SIZE13:45
TheJuliazigo: it is ours... 7.0.0 is flamingo, not epoxy though, so I'm a little confused what the issue is13:46
TheJuliaiurygregory: downstream, we have a kb, step 3 is along the lines of what you could do to convert an image; https://access.redhat.com/articles/709692413:52
iurygregoryTheJulia, tks!13:52
TheJuliaiurygregory: so, does that mean the target changes were not necessary for sushy?13:59
TheJuliaI guess we need to get a change landed w/r/t slotting13:59
iurygregoryTheJulia, they were necessary, without https://review.opendev.org/c/openstack/sushy/+/950694 it would fail 14:00
opendevreviewMerged openstack/sushy master: Improve Dell Asynchronous task handling  https://review.opendev.org/c/openstack/sushy/+/96173014:01
TheJuliawith or without the retry logic change for async handling14:01
TheJulia... ^ that one14:01
TheJulia(that was great timing!)14:01
iurygregoryI've tested with it =)14:07
iurygregoryTheJulia, by any chance you have the block-device-default.yaml ?  seems like my centos9 machine is unable to find openstack-tripleo-image-elements14:08
iurygregoryor can I just use https://opendev.org/openstack/diskimage-builder/src/branch/master/diskimage_builder/elements/block-device-mbr/block-device-default.yaml ?14:11
TheJuliaiurygregory: uhh14:15
TheJuliayou can use a default one from DIB14:16
TheJuliauhh14:16
TheJuliayou *do not* want to use mbr14:16
TheJuliayou want EFI14:16
iurygregoryok14:16
iurygregoryhttps://opendev.org/openstack/diskimage-builder/src/branch/master/diskimage_builder/elements/block-device-efi-4k/block-device-default.yaml14:16
iurygregorythis one14:16
TheJuliaSo, technically MBR booting might work on that kind of gear, but I think it would be insane14:17
iurygregorynow I will map things for disk-image-create -x -a x86_64 -o overcloud-hardened-uefi-full-4k.qcow2 --image-size 6GiB replace-partition-redhat :D14:17
TheJuliaI don't know what your source is, but should likely work14:20
TheJuliaDIB_BLOCK_SIZE does work the same on a standard image build, though14:20
iurygregoryimage-size 6GiB seems a lot14:21
iurygregoryO.o14:21
iurygregoryI think I'm doing something wrong https://paste.opendev.org/show/bixHZsFDw4nAjOHGuKJT/ O.o15:12
dtantsuriurygregory: OCP does not use images15:14
dtantsurcoreos-installer uses the content of the ISO15:14
iurygregoryso we won't have the 4k problem? nice15:15
dtantsuryeah, I don't think we will15:15
dtantsur(except for BMaaS, of course)15:15
iurygregoryI'm trying to spin an OCP cluster with some machines15:16
iurygregorywill likely be able to test with the custom changes to be sure things work begin next week15:16
iurygregoryfingers crossed15:16
cardoeJust wanna poke folks and say the ironic-week-prio is filling up (my definition is over 25) and there's a number of patches with one (sometimes three) +2 reviews.15:47
cardoeWe can use the weekend lull for all the gate jobs.15:47
TheJuliaiurygregory: that is the base image size to expect16:09
TheJuliadtantsur: yeah, and in that case, all you can really do is just make folks aware and point to tools/options to make a 4k image16:10
* dtantsur nods16:13
iurygregorycardoe, I will take a look at some today16:17
opendevreviewJulia Kreger proposed openstack/sushy stable/2025.2: Improve Dell Asynchronous task handling  https://review.opendev.org/c/openstack/sushy/+/96181816:41
-opendevstatus- NOTICE: All hosted mailing lists are undergoing maintenance for the next hour: https://lists.opendev.org/archives/list/service-announce@lists.opendev.org/message/UTMXRWWTE5WA3IF6WS3BIEJAORI2D62V/20:01
zigoTheJulia: Its unit tests are failing if I don't put python3-neutron (>= 2:27.0.0~rc1) in the build-depends (ie: it fails with the epoxy version). I'm telling you so you should add such thing in the requirements of networking-baremetal for Flamingo, IMO.21:52
*** dmellado9 is now known as dmellado22:01
JayFIdeally, yes, but upper-constraints.txt enforces the right version, too.22:36
JayFIf you put in the merge request I'd review it22:37

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