Thursday, 2025-10-09

opendevreviewOpenStack Proposal Bot proposed openstack/project-config master: Normalize projects.yaml  https://review.opendev.org/c/openstack/project-config/+/96255702:24
opendevreviewTony Breeds proposed opendev/system-config master: Update ansible-devel job to run on a newer bridge  https://review.opendev.org/c/opendev/system-config/+/93053802:49
opendevreviewTony Breeds proposed opendev/system-config master: Update pip3 role to work on Ubuntu Noble  https://review.opendev.org/c/opendev/system-config/+/93493702:49
opendevreviewTony Breeds proposed opendev/system-config master: Add ara and tzdata as installed requirements for the ansible-devel job  https://review.opendev.org/c/opendev/system-config/+/93491702:49
opendevreviewTony Breeds proposed opendev/system-config master: Install ARA master in the ansible-devel job  https://review.opendev.org/c/opendev/system-config/+/92401202:49
opendevreviewTony Breeds proposed opendev/system-config master: Add some debugging commands to the post job  https://review.opendev.org/c/opendev/system-config/+/92566702:49
tonybapart from wiki.openstack.org, do we have any servers that we manage that are not listed in system-config:inventory/base/hosts.yaml ?06:36
opendevreviewTony Breeds proposed opendev/system-config master: Update ansible-testing to run on the next version of bridge  https://review.opendev.org/c/opendev/system-config/+/93053807:29
opendevreviewTony Breeds proposed opendev/system-config master: Update pip3 role to work on Ubuntu Noble  https://review.opendev.org/c/opendev/system-config/+/93493707:29
opendevreviewTony Breeds proposed opendev/system-config master: Add ara and tzdata as installed requirements for the ansible-devel job  https://review.opendev.org/c/opendev/system-config/+/93491707:29
opendevreviewTony Breeds proposed opendev/system-config master: Install ARA master in the ansible-devel job  https://review.opendev.org/c/opendev/system-config/+/92401207:29
opendevreviewTony Breeds proposed opendev/system-config master: Add some debugging commands to the post job  https://review.opendev.org/c/opendev/system-config/+/92566707:29
fungitonyb: not to my knowledge, no. the only ones i could think might be missing are listed there after all, so just wiki i think11:13
clarkbI guess its worth noting we got through the PrePTG agenda yesterday and are not using the 1500-1700 block of time today15:10
clarkbin case anyone ends up looking for us on meetpad15:10
clarkbgithub is having issues this morning if anyone runs into problems that may be why15:49
fungiout to lunch, bbiab16:33
clarkbcorvus: I've just cleaned up the old nodepool images in vexxhost ca-ymq-1. Looking at sjc1 it doesn't appear we ever configured it to boot nodes16:51
clarkb(and as a consequence has no images either)16:51
corvussounds about right -- or at least, was so far in the past nodepool cleaned up our early attempts16:52
clarkbdo we recall if that was inetntional? THere are nodepool iamges that can be cleaned up but I didn't want to do that until I have a better understanding of the current sitaution there16:52
clarkbI've dug up the old nodepool config and max servers as 0 in sjc1 so that explains why zuul launcher isn't booting there16:53
clarkbcorvus: do you think I should go ahead and clean up the nodepool images anyway? I guess there isn't any reason to keep them16:53
corvusoh i thought you were saying there were no images in sjc1 because we never used it17:02
corvusbut you're saying there are images despite us never using it17:03
corvusanyway, i think you can delete the old nodepool images from sjc117:03
corvusi don't recall why we stopped using it with nodepool a long time ago17:03
corvusclarkb: https://review.opendev.org/73150817:06
clarkbcorvus: ya sorry there are no nodes but there are images17:26
clarkbI will proceed with cleaning up those images now17:26
clarkbvexxhost images are cleaned up for the most part. There is one image in sjc1 that I cannot delete and i think there were four in ca-ymq-117:41
clarkbthat means rax classic is the last region needing cleanup. I suspect this one will be the most problematic (due to its continuous use for over a decade) and as a result will defer that for now17:45
fungiokay back18:00
fungiyeah, i think we were asked to use ca-ymq-1 for nodepool nodes and not sjc118:01
fungithough it was okay to put control plane hosting in sjc118:01
clarkbyup that is/was the situation18:02
clarkbbut nodepool was still uploading images (zuul-launcher is not)18:02
fungiright, so i agree it should be safe to blow all those away18:10
clarkbI've got a school thing in about half an hour and I'm going to use that as motiviation to get out on the bike. I'll be back in a bit19:56
fungihave fun!19:56
claygzigo: yooo!! I'm confused!  (as usual) Missing sub-modules when building w/ setuptools · Issue #1071 · eventlet/eventlet https://github.com/eventlet/eventlet/issues/1071 would love it if you could set me straight brother!  🫶20:49
clarkbclayg: I don't know any of the answers, but `python setup.py $command` is deprecated and going away. You should probably be using the python build command: https://pypi.org/project/build/ instead22:23
clarkbregardless of the pyproject.toml contents i would stop using any setup.py commands as step 0 in debugging if they don't produce the expected results any longer22:24
fungithough it's possible that running `pyproject-build` instead of `setup.py sdist` is still going to have the observed problem with the packaged modules not getting included due to overriding setuptools module search routine22:27
clarkbright I don't know if this will fix the problem, more of ensuring the foundation is in place and correct before doing additional debugging22:28
fungihttps://setuptools.pypa.io/en/latest/userguide/pyproject_config.html indicates that packages is for explicitly listing all the ones that should be included, and doesn't indicate it's recursive so probably what clayg has proposed is a viable solution22:31
fungihowever, it looks like [tool.setuptools.packages.find] can be used to fine-tune the discovery mechanism rather than overriding it completely22:32
fungiit also mentions py-modules but doesn't really say how it's used, making me suspect it's deprecated/discouraged22:40
fungieven though it's not called out as being discouraged or obsolete like some other settins listed there22:40
clarkbfungi: note that it looks like setuptools is no longer even used anymore?22:41
clarkbI guess for me what I would so is stop using python setup.py sdist and/or setuptools entirely and let build make a hatch built tool22:42
fungiah, yeah it does look like hatch/hatchling is what eventlet is using now22:42
clarkbfi you're backporting fixes you don't need to backport the entire build system just the fixed code22:43
fungiso running pyproject-build probably will entirely bypass setuptools22:43
clarkband in that case I would expect setuptools to keep working as is22:43
fungiso yes, i do think your suggestion (not invoking setuptools, using a pep 517 compliant "build frontend") is the actual solution22:46
fungioh, though clayg points that out in his issue, so obviously not a viable solution in his case since he wants to continue using setuptools instead22:49
fungithe equivalent version handling would be to add the setuptools-scm plugin22:49
fungibut really it needs a different pyproject.toml to support setuptools instead of hatch22:50
clarkbright I think my confusion is why you need to use new pyproject.toml with new security patches22:51
clarkbif you just want to update setuptools to latest then I would use the new build system. If you just want security patches I would expect those to backport onto the old setuptools based system22:51
fungiit looks to me like the current debian packages aren't relying on setuptools either? https://setuptools.pypa.io/en/latest/userguide/miscellaneous.html22:52
fungier, wrong paste22:52
fungihttps://salsa.debian.org/openstack-team/third-party/python-eventlet/-/blob/debian/flamingo/debian/rules?ref_type=heads#L1322:52
fungi--buildsystem=pybuild22:52
clarkbso its possible the work zigo did was a dead end and never worked as expected23:02
clarkb?23:02
fungior was later superseded23:02
fungianyway, i commented on the issue with some links23:02
clarkbI just responded to the openstack discuss list with some quick details on how we stuff many keys into one nova key object and that makes thinsg magically work23:05
clarkbpretty sure that when ianw first generated an ed25519 key nova wouldn't accept it as the first key but was happy with it later in the list23:05
fungiyeah, it's sort of an abuse of the feature, but seems to be a still-viable workaround23:05
clarkbwell I would argue at this point if the api is not capable of recognizing valid keys otherwise then its definitely grown into a proper feature itself23:06
fungiaccidental feature23:06

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