Friday, 2023-11-03

opendevreviewTobias Rydberg proposed opendev/irc-meetings master: Public Cloud SIG holds its meeting at 0800 UTC during winter time.
frickler900k \o/09:29
opendevreviewMartin Kopec proposed opendev/irc-meetings master: [iwg] Update office hour date and time
opendevreviewMerged opendev/irc-meetings master: Public Cloud SIG holds its meeting at 0800 UTC during winter time.
amorinhey all, the mistral team is struggling figuring out why devstack keeps failing in CI while it seems working correctly when executed manually. The last change we did is the following:
amorinAs you could see, per Jan comment, apache wsgi seems not able to find mistral wsgi script, even if it was sucessfully installed. I dont get how we endup in this situation, my manual check were not doing the same result12:31
fungiamorin: is it only happening with the latest version of pip?12:31
fungiamorin: did you see this discussion?
amorinlet me check the pip version12:33
fungiamorin: the tl;dr is that with pip 23.1 or later, if the project is relying on pbr to set up wsgi entrypoints, "editable mode" (pip install -e) a.k.a. use_develop=true in tox.ini doesn't install them unless the project has a pyproject.toml file specifying pbr as a
fungiusing older pip or not using editable mode are a couple of possible workarounds, depending on the situation12:37
amorinwhy then would it work manually but not in CI?12:37
fungiwithout knowing how your local environment differs from ci (particularly what version of pip you're using and how you're calling it), it's hard to say12:38
amorinmy "system" pip is the following: pip 22.0.2 from /usr/lib/python3/dist-packages/pip (python 3.10)12:38
amorinubuntu 22.04 12:38
fungii'm checking that mistral-devstack log to see if it's using pip 23.112:40
amorin2023-10-25 14:12:17.987653 | controller |     Found existing installation: pip 22.0.212:41
amorin2023-10-25 14:12:18.219253 | controller |     Uninstalling pip-22.0.2:12:41
amorin2023-10-25 14:12:18.231134 | controller |       Successfully uninstalled pip-22.0.212:41
amorin2023-10-25 14:12:19.407380 | controller | Successfully installed pip-23.3.112:41
amorinseen in logs12:41
amorinperfect! you are amazing fungi!12:41
fungiif i can be amazing just by reading the openstack-discuss mailing list, i've probably been trying too hard ;)12:42
amorinis a decision taken regarding the solution? pyproject.yaml file is the future?12:43
fungiamorin: in the initial message in that thread sean also proposes a temporary alternative to cap the pip version in devstack, but it seems like there's a rough consensus that we'll start adding minimal pyproject.toml files with for keystone as an example12:46
fungilooks like that change is currently blocked on some test bitrot in keystone12:47
amorinok, perfect then I'll add the same for mistral and link it to the same topic, thanks12:47
fungiamorin: oh, yep looks like has a bunch now as of yesterday12:47
fungisean's not in here, but i'll mention it in #openstack-infra12:48
fungiamorin: sean points out there's a script at if you want to propose multiples12:51
fungiand also some cross-project tracking notes in there12:52
amorinack thanks, will do mistral manually for now, maybe other mistral stuff with the script later12:52
amorinhere is the patch12:57
amorinlast thing I want to ask, we are fixing multiple things in mistral (see the relation chain). I expect only the last one will get the v+1, who should I ask for a manual v+1 on previous commits ?12:59
amorinor maybe it's handled automatically by zuul?12:59
fungiamorin: if you need to fix multiple things at once, you can either squash them into a single change (we did that yesterday to get pbr's tests passing, for example), or you can temporarily disable the failing jobs and enable them again once the necessary changes to make them pass are merged13:01
amorinI was thinking that it would be better to keep the changes in separate commits.13:01
amorinI am ok with disabling the failing jobs, is it something I should do in .zuul.yaml ?13:02
amorinadding voting: false is the trick, right?13:02
fungiyeah, or just commenting out/deleting them in the check and gate pipelines13:05
kevkofungi: don't know how but i fixed it :D 14:53
kevkohello \o/14:54
fungiwhat did you change?14:55
kevkoguys, can I have two depends-on in a review ? will it work ? (btw, one is merged ...but i don't want to replace it because i want to track it )14:55
kevko << probably this fungi14:55
kevkofungi: and this
fungiyou can specify multiple depends-on yes. whether it will work depends more on what your expectations are, if your jobs are configured to consume git repository source for the projects where the depended-on changes reside, and other factors14:56
kevkobut it was really long night it's possible  that in some point i've was glad that it works so i reverted everything and i was happy that it is working :D 14:57
fungiin particular, avoid making dependency loops, since zuul won't be able to untangle them into a linear series14:57
kevkoyeah ..i know 14:57
fungialso jobs need to set required-projects to include the projects you want to depends-on to14:58
fungiotherwise zuul won't prepare speculative checkouts for them14:58
*** diablo_rojo is now known as Guest567015:18
kevkohmm, fungi, okay .. why this patch is not running letsencrypt tests ? can u check please ? 15:41
clarkbkevko: did you put the new test in a pipeline?15:44
fungilooks like it's setting a feature flag var for the job15:44
clarkbI think the jobs aren't in any pipeline so they don't run. But that is after a quick skim of the change stack15:45
fungiyeah, the parent change defines kolla-ansible-lets-encrypt-base but doesn't instantiate it into any pipeline15:46
fungisame for kolla-ansible-ubuntu-lets-encrypt and kolla-ansible-rocky9-lets-encrypt15:46
fungii guess kolla-ansible-lets-encrypt-base isn't meant to be added to any pipeline, but kolla-ansible-ubuntu-lets-encrypt and kolla-ansible-rocky9-lets-encrypt presumably should be15:47
fungikevko: i think you want to insert them into check and gate sections in
fungior whatever pipelines make sense15:49
kevkohmm thanks 15:58
kevkoi am going to try 15:58
kevkofungi: thank you very much looks like it helps :) 16:13
*** dhill is now known as Guest568318:28
opendevreviewJulia Kreger proposed openstack/diskimage-builder master: Skip glean-install for ironic ramdisks
clarkbthe gitea 1.21.0 release is imminent. Apparently the last todo on their side is writing the release documentation/blog21:09
clarkbI guess that includes a changelog because that doesn't exist yet either21:09
fungiwriting changelogs after the fact seems so 199521:26

Generated by 2.17.3 by Marius Gedminas - find it at!