| clarkb | JayF: I'm not sure what yaml you are referring to | 00:12 |
|---|---|---|
| clarkb | nodepool builds images with dib using a yaml specification and dib uses yaml to describe disk layouts | 00:12 |
| clarkb | but there isn't really a yaml specification for dib I don't think | 00:12 |
| JayF | Basically just the list of keys that can go in the top level of a build definition yaml file. We have the disk partitioning portion extensively documented, but not the general usage portion. Or at least I couldn't find it | 02:58 |
| JayF | Those top level keys are all listed by calling diskikage-builder --help | 02:58 |
| opendevreview | Gwen Dawes proposed openstack/diskimage-builder master: Force grub2-install to bypass secureboot complaints. https://review.opendev.org/c/openstack/diskimage-builder/+/937442 | 09:11 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/diskimage-builder master: Add support for DNF5-based systems https://review.opendev.org/c/openstack/diskimage-builder/+/934332 | 12:41 |
| Petraea | Hi folks. Is it just me, or is the OpenEuler 22.03 test suite failing a bunch of stuff? I have a proposed fix to a bug I ran into yesterday, and Zuul's failing over those tests using it. | 14:36 |
| clarkb | its not just you it need to be fixed or turned off but I've been too busy to follow up on it | 15:10 |
| clarkb | the problem is we flipped our mirror over to 24.xy which removed the 22.03 content. We thought we could fallback to using whatever they default it but it seems that we need an explicit mirror. So the fix would be to point the tests at some explicit location that isn't our mirror or remove the tests / make them non voting until they get updated to 24.xy | 15:11 |
| clarkb | JayF: I didn't realize there was a build definition yaml file | 15:12 |
| opendevreview | Gwen Dawes proposed openstack/diskimage-builder master: Force grub2-install to bypass secureboot complaints. https://review.opendev.org/c/openstack/diskimage-builder/+/937442 | 15:56 |
| Petraea | clarkb: I hope that's OK, little new to the Gerrit process. | 15:57 |
| clarkb | Petraea: thats perfect | 15:58 |
| clarkb | if you open up the change link just above our messages you'll see the latest push created a second version (patchset 2) on the existing change | 15:58 |
| clarkb | the top level change captures the logical work as it iterates through review and updating it the way you did is exactly what we want so that we don't lose track of the prior review comments | 15:58 |
| clarkb | I've got a morning full of meetings so unlikely I'll dig into the openeuler testing for a bit but it is on my list. Happy fopr others to poke at it too | 15:59 |
| Petraea | Aha. I'm used to squash-after-review, this squash-before is a little confusing to me! But I'm glad. | 15:59 |
| clarkb | I find it makes it a lot easier to see the context of "what do we actualyl want to change" while also capturing iterative improvements | 16:01 |
| clarkb | whereas squash after review tends to come with a branch with many commits without any real organization that you have to wade through | 16:01 |
| JayF | clarkb: diskimage-builder devstack-noble.yaml | 16:04 |
| JayF | https://www.irccloud.com/pastebin/BIORVKRa/devstack-noble.yaml | 16:04 |
| JayF | that's what I was wanting to do (and successfully did) | 16:04 |
| clarkb | fungi: you've been working with the openeuler stuff do you think simply removing the test / making it non voting is a big issue? | 16:11 |
| clarkb | I suspect that is the most straight forward option right now | 16:12 |
| fungi | clarkb: sounds fine to me. honestly my only real involvement was reviewing the change from the openeuler community folks to update the mirror source and then manual sysadmin side steps associated with that. i don't really have any background on their plans with dib testing | 16:15 |
| fungi | i can push up a change to remove the job from check and gate for now | 16:15 |
| fungi | er, i guess it's just specific tests in a larger job | 16:16 |
| clarkb | fungi: there will be a specific openeuler functional test that should be removed/non voting then also the openeuler test case in the func tests | 16:18 |
| clarkb | I think the functests have a mechanism for treating a failure as non voting too but I'm not positive of that | 16:18 |
| opendevreview | Jeremy Stanley proposed openstack/diskimage-builder master: Temporarily disable OpenEuler functional testing https://review.opendev.org/c/openstack/diskimage-builder/+/937466 | 16:22 |
| clarkb | oh we don't have a build image and boot it job for openeuler just the functests | 16:26 |
| clarkb | I've approved ^ so that we can recheck a few of the outstanding changes out there and then hopefully do a release soon | 16:26 |
| JayF | clarkb: https://jay.jvf.cc/posts/diskimage-builder-yaml/ I wrote it up, since I figured if you weren't aware probably a lot of people aren't. I'll put it on my "eventually" list to add it to actual-docs, but that takes thinking about how to design it :) | 16:38 |
| clarkb | I suspect it was added and because I never use it just never registered it was possible | 16:39 |
| clarkb | it is nice how descriptive that ends up ebing for your image build though | 16:39 |
| JayF | I believe you can add more list entires in the same yaml | 16:39 |
| JayF | and build more images | 16:40 |
| JayF | which means you could take full advatantage of yaml templating | 16:40 |
| Petraea | On that note, Zuul seems to be failing out on the ARM64 pipeline too. Is that one new? It looks separate from the main chain. | 16:53 |
| clarkb | that looks like a case of debootstrap is too old to know how to build noble on bookworm for the deb job. I feel like we fixed something similar with a symlink to another release maybe? | 16:56 |
| clarkb | https://opendev.org/zuul/nodepool/commit/5ff7dd8289e0171d627f9f6e4b27c51df21671d8 this is what I was thinking of. How does it work for x86 in dib though | 16:58 |
| clarkb | the rpm arm job is failing on openeuler too | 16:59 |
| clarkb | those jobs don't vote already so we can fix them separately though | 16:59 |
| JayF | https://github.com/openstack/diskimage-builder/blob/master/diskimage_builder/elements/base/install.d/00-baseline-environment#L11 is there a reason we install iscsi on *everything*? | 18:50 |
| JayF | this install happening before the reboot is causing breakages on ubuntu-noble (basically, it needs to upgrade the kernel before apt installs start working, for reasons I can't fully grok yet) | 18:51 |
| JayF | but it seems weird generally to have a hard requirement on iscsi utils | 18:51 |
| JayF | > linux-image-6.8.0-50-generic : Depends: linux-modules-6.8.0-50-generic but it is not installable | 18:52 |
| JayF | bah, even the update is busted :( | 18:52 |
| JayF | this is literally what was working yesterday | 18:52 |
| JayF | yeah I think ubuntu might just be busted or something? | 18:54 |
| clarkb | well we build ubuntu noble happily and I'm pretty sure our images use the base element and that package intall has probably been there for 12 years so I'm guessing ti isn't iscsi at fault | 18:55 |
| clarkb | as for why you install iscsi I suspect it is an artifact from when dib was born in tripleo and they were booting baremetal and wanting to mount volumes | 18:56 |
| JayF | have you tried running that build **today** | 18:57 |
| clarkb | I dunno I'd have to go look at the logs | 18:57 |
| JayF | I think there's an issue in the ubuntu base image and/or their repos causing fresh breakage today | 18:57 |
| JayF | if someone else could verify the failure it'd at least justify me going deeper into this hole | 18:57 |
| clarkb | so two things we build using ubuntu-minimal not ubuntu specifically bceause the published uimages have not always been great starting points. And we point our ubuntu builds at our own package mirrors which only update if reprepro can validate them which helps prevent package repo problems | 18:58 |
| clarkb | but the changes in scrollback do ubuntu buidls (of noble too I think) so you could check them. But also the logs on https://nb01.opendev.org and https://nb02.opendev.org | 18:58 |
| JayF | I can reproduce this in both ubunutu | ubuntu-minimal | 18:58 |
| JayF | #ubuntu on libera is suggesting it might be an upstream issue w/their repos | 18:58 |
| JayF | yeah, they're confirming basically the linux-modules package wasn't uploaded for the linux pacakge they uploaded | 18:59 |
| clarkb | https://nb02.opendev.org/ubuntu-noble-dc76d3cba27749308d3d693c957998d9.log is our most recent build | 19:00 |
| clarkb | it was successful possibly because we use reprepro to avoid super broken repos | 19:00 |
| JayF | I suspect so, yes | 19:02 |
| JayF | in the meantime, if anyone hits this issue as well: DIB_UBUNTU_MIRROR_DISTS: updates,security,backports,proposed is a workaround | 19:08 |
| JayF | (enabling proposed) | 19:08 |
| JayF | this only works on ubuntu-minimal, there's no workaround to enable proposed on the cloud-image based build | 19:09 |
| JayF | Looking through DIB review backlog: | 22:46 |
| JayF | https://review.opendev.org/c/openstack/diskimage-builder/+/936209 fairly simple change (updating default ubuntu version to noble) | 22:46 |
| JayF | https://review.opendev.org/c/openstack/diskimage-builder/+/936206 (2x+2 just needs approval, another minor fix) | 22:46 |
| JayF | I have another one for DIB with review feedback I'll be updating shortly. | 22:47 |
| clarkb | my focus is going to be on getting the ci jobs working before anything else | 22:48 |
| clarkb | any idea if noble package repos are happy again? | 22:48 |
| JayF | I can run a quick local test :) | 22:49 |
| JayF | it fails in all of a minute when it's in the bad state | 22:49 |
| clarkb | oh thats conveniently quick | 22:50 |
| JayF | yeah, looks like it's OK, it's gotten past the point it was failing before | 22:51 |
| JayF | honestly the build is pretty fast in general | 22:51 |
| clarkb | ya dib caches a lot of stuff too | 22:51 |
| clarkb | I'll reapprove fungi's test fix | 22:52 |
| fungi | much obliged | 22:52 |
| JayF | now I'm going to go poke my third dib PR, which had feedback I had to respond to :) | 22:53 |
| opendevreview | Jay Faulkner proposed openstack/diskimage-builder master: Stop using deprecated pkg_resources API https://review.opendev.org/c/openstack/diskimage-builder/+/907691 | 23:09 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!