Thursday, 2023-06-01

opendevreviewTony Breeds proposed zuul/zuul-jobs master: [configure-mirrors] Allow per distribution disabling of mirrors
tonybSo I see that the mirror_info spec? is a nice solution to the problem.  I'm happy to work towards implementing that but I don't really know how/enough about how things work to get it done quickly.00:13
tonybI think it's a good opportunity to learn but I really would like feedback/advice on priority.00:13
tonybIt seems like there is stuff that'd benefit from being done soon (server migrations, podman tooling, gerrit are a few) so thoughts?00:14
tonybfungi, clarkb, ianw: Can y'all have a look over: for obvious errors or omissions befoer I send it to openstack-discuss tomorrow?00:54
ianwtonyb: one thing i'd say is that checking on hound there are other references
ianwthings like windmill i doubt work anyway00:57
tonybYeah lots of windmill hits, I'm worried that I didn't find the bindep ones myself00:58
tonybWhat is windmill?01:00
ianwfedora-36 also brings up x/tobiko that seems to use that node directly01:01
ianwwindmill was set of ansible roles setup a long time by pabelanger to deploy various things01:01
tonybAhh okay01:02
tonybI'll do more digging01:02
ianwi think it was in the era where we were trying to make system-config a very generic thing, so it was a lot of different components you could plug into01:03
ianwi don't know exactly when it happened, but during the puppet->ansible transition it just became impossible workload to keep totally generic roles working *and* do the work of keeping opendev production working01:05
ianwi.e. when changing anything in production, you'd have to think about a whole bunch of other platforms that production doesn't use, which held up the actual production :)01:06
ianwone of the only remnants of that is in system-config where we have a top-level roles/* which are supposed to be more generic; and the things under playbooks/roles are basically targeted at opendev deployment only01:07
tonybOh I did wonder about playbooks/roles01:07
tonyband now I know01:08
ianwwell, that and zuul-jobs of course, that became a place for generic work01:08
tonybYeah. of course01:08
opendevreviewMerged zuul/zuul-jobs master: Force cgroupfs cgroup manager with podman on ubuntu
*** dmellado17 is now known as dmellado104:55
*** amoralej|off is now known as amoralej06:12
frickleransible 8 was out two days ago according to pypi
opendevreviewMerged openstack/project-config master: Stop testing github hosted projects in openstack tenant
opendevreviewMerged openstack/project-config master: Stop caching infrequently-used CirrOS images
opendevreviewDr. Jens Harbott proposed openstack/project-config master: Cache new cirros images
*** amoralej is now known as amoralej|lunch12:09
fungitonyb: it may also make sense to include an observation that some of the jobs you linked haven't run at all for months and years, and the rest are all non-voting/experimental12:26
fungiexcept for the last one, for tripleo, which is another kind of situation all its own anyway12:26
fungiand also they're running on fedora 36 which is officially eol12:27
fungiwe normally don't continue building images for eol distro versions without a very good reason, so it's to the point where we probably need to decide between updating the fedora images or dropping them12:30
*** amoralej|lunch is now known as amoralej12:59
tonybfungi thanks all very good points 14:42
clarkbtonyb: ianw fungi windmill was pabelangers ansible to deploy zuul when working at ansible. We in opendev were'nt really ready for ansible at thatpoint so he wrote new stuff. At this point it is effectively dead and unmaintained and some (maybe all?) of the repos have been removed from zuul's tenant configs14:46
clarkbthe idea there is that we can easily add the projects back to zuul if someone starts maintaining it again but until then the simplest thing for resolving config issues is to remove them from zuul which we can do here. I'll take a look at that etherpad once I'm a bit more settled in this morning14:47
clarkbfwiw I did not attempt to join the gerrit community meeting this morning because i never got an answer on whether or not it would be held...15:17
tonybclarkb: do we need to warn Paul?15:57
clarkbtonyb: when I pulled things out of zuul previousl I messaged him and he basically said he didn't have time to do anything himself so whatever we did was fine15:57
clarkband it is easy to add it back when someone does show up to clean things up15:57
tonybokay cool.15:58
tonybI'll try to do that today.  I guess it isn't urgent and it'll be another good learning opportunity 16:05
clarkbsoudsn good.16:06
fricklerwhen doing cherry-picks via the gerrit UI, it amends the topic from "original" to "original-targetbranchname", does someone know if it is possible to override this? I'd like to keep all patches under the same topic16:08
fricklerin addition, updating the topic requires first deleting and then adding it again16:09
clarkbfor the second thing that is how it works. YOu clikc the 'x' and then add a new topic. For the second thing I assume that if it isn't an option on the cherry pick widget you can't change that. One thing that can mitigate the pain is that you can use the bulk actions on change lists to set the topic though16:12
fricklerlastly, there is some issue in the "set topic" autocompletion. if I type "fix-", "fix-zuul-errors" is listed among the completions. if I add to "fix-z", only some weird "fix_z*" strings are shown16:12
clarkbbasicall after you cherry pick everything go to your changes list and then bulk set the topic to what you want it to be16:12
clarkbthe autocompletion is likely based on lucene index content and that may be an artifact of how lucene does tokenization16:13
frickleroh, I never saw these bulk actions before, that is likely helpful, thx16:14
clarkbit is new with 3.716:21
clarkba feature that gertty has had forever16:21
*** amoralej is now known as amoralej|off16:23
fricklerI never got a working gertty setup set up. it was always thrashing around syncing git things for ages, like literally days16:23
clarkbI stopped using it around when I had kids because using a browser was easier when dealing with small children16:24
clarkbre thrashing I think the more repos you subscribe to the more it has to do, but not sure I ran into that16:24
opendevreviewDr. Jens Harbott proposed openstack/project-config master: Cache new cirros images
fungivery large repos like nova or neutron with tons of open changes are problematic to subscribe gertty to, and the rest api round-trips probably get much worse the higher your rtt/latency to the gerrit server17:04
fungis/problematic/very slow/17:04
opendevreviewClark Boylan proposed opendev/system-config master: Add zp02 zuul-preview server to replace zp01
opendevreviewClark Boylan proposed opendev/ master: Add zp02 to DNS
opendevreviewClark Boylan proposed opendev/system-config master: Remove zp01
opendevreviewClark Boylan proposed opendev/ master: Remove zp01 from DNS
clarkbok I think we can iterate through those changes making sure things are happy at each step of the way in order to replace an older server17:11
clarkbI picked it because it is easy >_> no volumes17:11
opendevreviewClark Boylan proposed openstack/diskimage-builder master: fedora: don't use CI mirrors
opendevreviewMerged openstack/diskimage-builder master: fedora: don't use CI mirrors
clarkbI'm going to approve now to get things moving on the zuul-preview server21:09
*** JayF is now known as Guest196221:37
*** JasonF is now known as JayF21:37
opendevreviewMerged opendev/system-config master: Add zp02 zuul-preview server to replace zp01
clarkbonce we're happy with the deployment that results in we should be safe to update DNS22:02
clarkbok I think that is looking good and we can switch DNS over now. But I'm about to eat dinner so may be best to do that tomorrow morning when I can watch it23:37

Generated by 2.17.3 by Marius Gedminas - find it at!