Wednesday, 2022-04-13

ianwthe fedora mirror should be released from now00:03
opendevreviewIan Wienand proposed opendev/system-config master: mirror-update: clean up Fedora ignores
ianwhrw/fungi: ^ yeah the other thing about offical cloning is that we aren't really a full mirror, we just choose what we want00:08
ianwtestinfra is broken by this to :/00:58
ianwthis seems less straight forward.  i'm unsure why this would be running as a different user01:14
ianwpython -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '"'"'/home/zuul/src/
ianwi guess tox is "installing" system-config?01:16
ianwohhh, hrm, testinfra runs on bridge01:18
ianwi bet we copy the repos over into bridge as root...01:21
*** rlandy|bbl is now known as rlandy|out01:23
fungisounds probable01:32
ianwhrm, maybe not is just plain prepare-workspace-git02:00
ianw- hosts: bridge.openstack.org02:07
ianw  become: true02:07
ianwi think that would do it.  we're running it all as root now02:07
opendevreviewIan Wienand proposed opendev/system-config master: run-base: allow git into system-config
opendevreviewIan Wienand proposed opendev/system-config master: mirror-update: clean up Fedora ignores
ianw<tibbs> ianw:  Sadly pubmirror1 is out of date for alt but up to date for everything else, because the repository got way too big.  New disks are on the way.07:14
ianw<tibbs> pubmirror2 is up to date as far as I can tell.07:14
ianwa follow up on the mirror situation07:14
ianwfedora mirror situation07:14
*** pojadhav is now known as pojadhav|lunch08:00
*** pojadhav|lunch is now known as pojadhav08:40
*** rlandy_ is now known as rlandy10:32
*** dviroel|out is now known as dviroel11:11
funginoted, thanks11:35
*** soniya29 is now known as soniya29|out13:51
*** artom__ is now known as artom14:00
*** pojadhav is now known as pojadhav|afk14:06
opendevreviewyatin proposed zuul/zuul-jobs master: ensure-sphix: upgrade setuptools in venv
ykarelfungi, ianw clarkb when u get chance please see ^15:20
ykarelneeded for translation job failures in neutron15:21
*** dviroel is now known as dviroel|lunch15:28
clarkbykarel is gone apparnetly15:45
clarkbI'm not quite sure how the pip constraints conflict thing maps to needing file: support but having file: support doesn't seem problematic15:46
clarkbbasically the change is fine, but may not address the issue unless I'm missing some key piece of info15:47
fungiclarkb: see my comment15:47
clarkbah thank you. we need toget better at including that info :)15:47
funginewer pip started using the new dep solver, causing it to not install the requested version of pyroute2 because it needed a newer setuptools15:48
clarkbya 0.6.1 of pyroute2 or newer needs newer setuptools and we constrain to 0.6.915:48
fungiand yes, i would have liked to see it explained in the commit message15:48
fungibut rather than -1 i just added a comment with the results of my research there15:49
clarkbyup that works. I was just digging into it after my review and confused as to why we thought this would fix it. The change itself was fine and added functionality I had no problem with :)15:49
fungiright, the change meant we used a version of pip which performed more correct dependency solving (or any at all really). not a bug, but a feature15:50
opendevreviewMerged zuul/zuul-jobs master: ensure-sphix: upgrade setuptools in venv
*** marios is now known as marios|out16:03
*** dviroel|lunch is now known as dviroel16:35
clarkbfungi: if you have a moment should unstick opendev's CI17:06
clarkbfungi: note I think that only affects testinfra because we have to do installations of packages. I don't think it affects our production runs because ansible isn't really installing system-config just referring to it. And that playbook is zuul only I think17:08
fungii'm confused by that change17:08
clarkbif we see that the production stuff is impacted too we can move it up a level and have it affect testing and prod but I suspect that isn't necessary.17:08
clarkbfungi: aiui its only the testinfra tox target that has the problem because we run that as root (it needs access to all the ansible stuff)17:09
fungiyet we're not using become on the git config call17:10
clarkbfungi: because it is already running as root17:10
fungiplaybooks/zuul/run-base.yaml is run as root?17:10
fungiokay, that explains it17:10
clarkbyes, in system-config we connect as root to our systems. This is part of the "test like production" (but not quite heh) and so ansible is ssh'd back into itself as root there aiui17:11
fungii wasn't sure what "This section" in the commit message referred to. i guess it should have said "this playbook"17:11
clarkbthere is the top level zuul ansible that ssh's in to the test node called bridge as zuul and drives the test job. That drives a nested ansible which runs as root and sshs into bridge as root to perform those actions to mimic production17:11
clarkboh ya there is a become: true at the top of that play17:12
clarkbso that makes it explicit17:12
fungiyep, i just got confused because i was trying to find where the section of the playbook that ran as root was17:12
fungibut it was all of the playbook17:12
fungijust me taking commit messages too literally again, i guess17:12
fungidid we rule out group ownership as a solution?17:13
clarkbwe did17:13
clarkbthe git patch onyl checks user ownership (and there are accusations on hacker news that tehy did it insufficiently too for extra fun)17:13
fungigit is equally displeased by the user being a member of the group owning the files?17:13
clarkband i tested that on my held node by adding root to the stack group and then retrying17:13
fungitoo bad17:14
clarkbif you want to test that you can check on my held node. root is already part of the stack group which should own the keystone repo which you can sudo pip install to see teh failure17:14
clarkbianw's analysis ended up checking that groups are ignored on the source side too and both sides agreed. Group membership isn't evaluated and is ignored for this17:14
opendevreviewMerged opendev/system-config master: run-base: allow git into system-config
fungigot it, thanks17:27
opendevreviewClark Boylan proposed opendev/jeepyb master: DNM want to run jeepyb testing against gerrit
opendevreviewClark Boylan proposed opendev/system-config master: Update our gitea test to handle new git behavior
clarkbfungi: ^ I think that will make gitea testing happy20:10
*** dviroel is now known as dviroel|afk20:22
*** arxcruz is now known as arxcruz|out20:28
*** timburke__ is now known as timburke20:55
clarkb seems to fix the gitea testing issue if infra-root have time to review it21:12
clarkbianw: we've also found a couple issues in opendev land including ^ and the one your found21:20
*** dhill is now known as Guest186422:00
opendevreviewJames E. Blair proposed zuul/zuul-jobs master: Add per-build WinRM cert generation
opendevreviewMerged opendev/system-config master: Update our gitea test to handle new git behavior
clarkb has been rechecked to hopefully flip its -1 to a +1 now that ^ has merged22:27
opendevreviewMichael Johnson proposed openstack/diskimage-builder master: Mark our source trees as safe for git to use as other users
opendevreviewLuciano Lo Giudice proposed openstack/project-config master: Add the cinder-three-par to Openstack charms

Generated by 2.17.3 by Marius Gedminas - find it at!