Wednesday, 2024-01-10

opendevreviewTony Breeds proposed opendev/system-config master: Switch meetpad servers to jammy
opendevreviewTony Breeds proposed opendev/system-config master: DNM: force jitsi_meet node failure for testing
frickleriiuc Feb 10 is the start of the Chinese New Year this year, with the week after that being a major holiday05:41
tonybahh that's good to know.  we'll double check and work with you and the rest of the TC to accommodate that.05:49
opendevreviewTony Breeds proposed opendev/system-config master: Switch meetpad servers to jammy
opendevreviewTony Breeds proposed opendev/system-config master: DNM: force jitsi_meet node failure for testing
opendevreviewTony Breeds proposed opendev/system-config master: Switch meetpad servers to jammy
opendevreviewTony Breeds proposed opendev/system-config master: DNM: force jitsi_meet node failure for testing
*** blarnath is now known as d34dh0r5315:01
opendevreviewIan Y. Choi proposed opendev/irc-meetings master: Revitalize I18n meeting slot
fungii'm headed out to grab lunch now that the flood waters seem to be receding, but should be back within an hour15:54
fungithere's a new ticket from rackspace about an outage for the trove instance zuul is using for its mysql database. i can give it a quick health checkup when i get back if nobody beats me to that15:55
clarkbreading the gerrit mailing list I've noticed that there is a Gerrit NPE that occurs on 3.8 when pushing code. This was apparently fixed on the 3.8 stable branch but not release has been made. We may want to rebuild our 3.8 images and deploy unreleased gerrit to fix that if it becomes a problem for us16:26
clarkbfungi: I am able to get a build listing from zuul which implies the db is up17:10
jrosseris there a correct way to refer to a playbook outside the repo that a job is defined in for `job.pre-run`?17:18
clarkbjrosser: no, playbooks must live in the same repo17:18
jrosser(the ordering of parent pre jobs running before child pre jobs is opposite of what i need)17:18
jrosserclarkb: is that enforced, or could i have a playbook in the same repo that `import_playbook` from another repo?17:20
clarkbthe lookups for the values defined in pre-run/run/post-run only look in the current repo. You might be able to hack around it doing imports but I'm not sure. Zuul may also manipulate search paths in ansible17:21
jrosserhmm ok i will have a try17:21
jrosseri'm attempting to drop some very small bits of config into place from the child job before the parent does anything17:22
fungiokay, back17:28
fungijrosser: the clean, zuulish way is to use roles instead, since you can tell it to mix in roles from other repos17:29
jrosserunfortunately i have to make this work outside of zuul as well17:30
jrosserfungi: i see that i can specify roles in `job.roles`, when are these run?17:38
fungijrosser: per it's a list of projects that your playbooks need roles from. you still call the roles from job playbooks17:44
fungithe idea is that if you have ansible which is needed by jobs defined in different repositories, you put it into a role in one of those repositories and then playbooks in the other repository can reuse those roles17:44
jrosserhmm i am still then stuck with the ordering of parent first, child second17:46
jrosserand why i originally asked about referring to playbooks from other repos in order to work around that17:46
fungirevisiting that part, you have tasks which a child needs run before pre tasks in a parent?17:47
jrosseryes, thats right17:47
jrosserpre tasks in the parent does "all the setup", child wants to drop some config first to influence that setup17:48
fungithe way i've seen that tackled elsewhere is to create a divergent ancestry at the point before the pre tasks you need to inject17:48
fungior use variables17:48
jrosseryeah, so thats what i'm doing17:48
jrosserand in the child i then wanted to call what would have been "all the setup" from the parent repo17:48
fungivariable-based logic branching can work, you put the logic in the parent but specify (or override) the variable's value in the child job definition17:49
fungiparticularly if it's something as simple as a toggle for a boolean decision17:50
fungilike is done for fips job setup, for example17:50
jrosserit's writing some config files, actually copying them out of a zuul repo into /etc/....17:50
fungican the parent job do that, gated behind a switch?17:52
jrosserdo you have a documentation link for that kind of thing?17:55
fungi and subsequent entries cover it, has an example job definition17:56
bqianzuul job failed merging in starlingx/update repo, I ran the failed tox tests on my build machine, no error. the zuul failure seems to be server specific bandit: 754 W install_deps> pip install -U -c -r /home/zuul/src/ [tox/tox_env/] ERROR: 404 Client Error17:57
fungijrosser: as long as it's not a trusted playbook (like in a config repo), the variable values set in the child job definition will propagate to the parent job playbooks17:57
clarkbbqian: I can reproduce the 404 error17:57
clarkbbqian: I believe that is a valid error probably due to branch cleanup in openstack17:58
fungibqian: yes, like clarkb said, no longer has a stable/stein branch, it's eol17:59
clarkbbqian: is the chagne that had that happen17:59
fungibqian: you could use instead17:59
bqianoh, thanks for reply clarkb and fungi18:00
clarkbhow does this look for wheel build dropping announcement:
clarkbinfra-root ^ I'll probably wait long enough for tonyb and frickler to weigh in on that before sending it18:18
fungithe error condition details seem odd to jump straight into. might cover them nearer the end of the message or in a footnote?18:20
clarkbfungi: like that maybe?18:20
fungithe important part of the announcement is the intent to stop building wheels for those platforms, so yeah lgtm now18:21
fungii was just worried that getting too far into the weeds at the beginning of the message would cause lots of people's eyes to glaze over before they got to what we're trying to tell them18:22
clarkbya it is good feedback. hit them with the important part first18:22
fungi"we're turning this off, and here's why..."18:22
fungiit also tells them a lot sooner whether they need to read further or can skip because they don't care/aren't affected18:24
clarkbtonyb: I slept on it and realize that if we really want to do a completely new meetpad setup and then update DNS once the way to do it is with hostvars to override the groupvars values. This will end up being a bit verbose but will enable us to have a new jvb talk to a new meetpad and not have them talk to anything else since hostvars win the var precedence battle against group vars18:27
clarkbI think that is doable and would be happy to take the appraoch if you prefer it18:28
clarkb this shows a list of projects that may not be python3.12 compatible because they rely on the removed imp module. I'm poking at zuul currently and I think PBRs is a fallback case for using importlib and should be fine19:22
clarkbotherwise I don't see anything that will affect us after a quick skim19:23
fungii saw some recent changes in... dib? for that as well19:23
clarkboh yup sorry dib is listed and ya I rememebred ianw pushed fixes iirc19:24
clarkbor maybe it was stevebaker?19:24
fungibut yes, it's just a normal part of python 3.12 readiness. some projects already addressed it because imp has been deprecated for years19:24
clarkbthe new importlib apis feel less approachable. Which I guess may be intentional on Python's part as they want people to stop messing with import internals19:25
fungiin addition to the zuul db ticket, the trove instance used by cacti had a similar incident today19:31
clarkbchecking the ubuntu-ports I don't see any 32bit arm or ppc pacakages. We are just doing arm64 binary packages. However, I see that we are mirroring source packages? At least we have them listed here:
clarkbthere is also a debian-installer/ dir19:32
clarkbthe jammy content does not have source content. I suspect taht we can clear out the source content for focal and bionic19:32
clarkband then xenial is listed but hasn't updated in years so I don't think we are actually mirroring it anymore. Unsure if we have stale xenial packages that need to eb cleared19:32
fungiapart from a brief gap in graphs around 18:00-18:15 utc, i don't see any problem with cacti now19:33
fungiand that seems to correspond to the time in the ticket19:33
clarkbfungi: sounds good re cacti19:34
clarkblooking at our reprepro config we don't seem to configure jammy any differently than the other releases so I'm not sure why it doesn't have source packages19:34
clarkbinterestingly our main ubuntu mirror appears to include source packages for everything including jammy?19:36
clarkbhrm the debian wiki says that reprepro uses source pacakges to track changes in binary pacakges so maybe we can't?19:43
clarkbI think the best next step we can take for ubuntu-ports is to cleanup the bionic images and drop bionic ports mirroring19:46
clarkblooking at codesearch it doesn't seem like much is really using that (system-config checks our base roles against it which is likely holdover from when we ran nodepool builder on bionic)19:46
fungithe last openstack release to include bionic in its pti was ussuri:
fungiin theory, some projects may still have tests on that platform:
fungisince codesearch doesn't include those branches, they may not show up readily20:09
clarkbfungi: right, but also this is arm64 specifically which didn't have as much traction then iirc20:10
fungioh! just arm, yeah should be fine20:10

Generated by 2.17.3 by Marius Gedminas - find it at!