| -@gerrit:opendev.org- Zuul merged on behalf of Benjamin Schanzel: [zuul/zuul] 934310: zuul-web UI: Fix initial show all pipelines state https://review.opendev.org/c/zuul/zuul/+/934310 | 00:18 | |
| -@gerrit:opendev.org- Artem Goncharov proposed: [zuul/zuul] 859940: Implement tests of the gitea driver https://review.opendev.org/c/zuul/zuul/+/859940 | 08:57 | |
| @0weng:matrix.org | @stephenfin: test ping | 17:13 |
|---|---|---|
| @jangutter:matrix.org | HALT WHO DISTURBS THE GINGER! | 17:13 |
| @jangutter:matrix.org | apologies, popup confused me completely! | 17:14 |
| @0weng:matrix.org | Haha no worries :D | 17:14 |
| @0weng:matrix.org | Just troubleshooting some Matrix stuff | 17:14 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul-jobs] 935574: WIP: Add mirror-container-images role and job https://review.opendev.org/c/zuul/zuul-jobs/+/935574 | 18:35 | |
| @clarkb:matrix.org | where did we end up with not using the old zuul-registry for the buildset registry | 21:15 |
| @clarkb:matrix.org | * where did we end up with not using the old zuul-registry on docker hub for the buildset registry | 21:15 |
| @clarkb:matrix.org | looks like that change is failing on the docker hub request limits now too | 21:16 |
| @clarkb:matrix.org | I'm thinking opendev should maybe drop the docker hub mirror by default so that we can spread out the requests and give us a better shot of success then we should land this change as its one less thing to pull from docker hub | 21:16 |
| @mnaser:matrix.org | Yeah it’s stuck on that but one of the tests pulls registry:2 which doesn’t help Clark | 21:18 |
| @clarkb:matrix.org | ya but if we drop the config that forces everything through the single proxy cache we in theory get many more requests | 21:22 |
| @clarkb:matrix.org | not perfect but may help alleviate some of the problems | 21:22 |
| @clarkb:matrix.org | mnaser: fwiw not sure if you saw https://www.docker.com/blog/november-2024-updated-plans-announcement/ which indicates a 10 requests per hour limit. However, I checked in the tokens themselves and they report 100 requests per 6 hours | 21:22 |
| @clarkb:matrix.org | anyway I suspect that the rough order of operations here will be to stop using proxy caches, then reduce our reliance on docker hub with changes like yours as well as movement to quay or elsewhere | 21:23 |
| @clarkb:matrix.org | and as we do that the remaining use of docker should become more reliable | 21:24 |
| @clarkb:matrix.org | for the test that pulls registry:2 I wonder if we can use zuul-registry or something hosted on quay | 21:24 |
| @jim:acmegating.com | it could use zr with a little bit of rework. it was a bootstrapping problem that is largely alleviated. but there may still be some value in diversity of testing. it would be easier to just mirror it to quay a la 935574 if we want to get it out of the path. | 21:27 |
| @clarkb:matrix.org | that also seems reasonably | 21:28 |
| @clarkb:matrix.org | * that also seems reasonable | 21:28 |
| @clarkb:matrix.org | in any case I feel liek getting the proxy caches out of the euqation is at least a temporary help. Would a flag to check at either https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/ensure-docker/tasks/docker-upstream.yaml#L12 or https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/use-docker-mirror/tasks/main.yaml work? not sure if there is a preference | 21:29 |
| @jim:acmegating.com | Clark: looks like use-docker-mirror doesn't use the new style of mirror config. ideally we would solve this problem by migrating to that. that may be slightly more work than adding a throwaway flag, but it would be forward progress. may not be too hard for something like this. | 21:31 |
| @jim:acmegating.com | https://zuul-ci.org/docs/zuul-jobs/latest/mirror.html | 21:31 |
| @jim:acmegating.com | (so in that system, the way to turn off dockerhub would be to simply remove "dockerhub" from zuul_site_mirror_info) | 21:31 |
| @clarkb:matrix.org | I think if we do that we may still need a separate flag if we want to maintain backward compat? | 21:31 |
| @clarkb:matrix.org | otherwise we'd just go aehad and configure like we do today? | 21:32 |
| @jim:acmegating.com | if zuul_site_mirror_info is defined and dockerhub is not in zuul_site_mirror_info do not configure dockrehub mirror | 21:32 |
| @clarkb:matrix.org | ah ok that works | 21:33 |
| @clarkb:matrix.org | I'll work on something along those lines now | 21:33 |
| @jim:acmegating.com | (also, worth rereading that real quick -- i think the var name is actually `mirror_info`) | 21:33 |
| @clarkb:matrix.org | the docs say either mirror_info or zuul_site_mirror_info so we'll check both | 21:34 |
| @clarkb:matrix.org | we've got quite a matrix of existing conditionals to logic out. THis may take me a minute | 21:45 |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul-jobs] 935722: Support new style mirror_info in use-docker-mirror https://review.opendev.org/c/zuul/zuul-jobs/+/935722 | 21:55 | |
| @clarkb:matrix.org | corvus: ^ that feels a bit more complicated than it needs to be but I think I got the logic right | 21:55 |
| @mnaser:matrix.org | “Unauthenticated users will be limited to 10 Docker Hub pulls/hr/IP address.” Yeah that explains it. It’s kinda pretty much squeezing it out to make it impossible | 21:59 |
| @clarkb:matrix.org | mnaser: ya the only thing is the tokens don't indicate that has happend that said the tokens may have the wrong data in them. But experimentally I was able to do more than 10 pulls locally | 21:59 |
| -@gerrit:opendev.org- Clark Boylan proposed: | 22:10 | |
| - [zuul/zuul-jobs] 935722: Support new style mirror_info in use-docker-mirror https://review.opendev.org/c/zuul/zuul-jobs/+/935722 | ||
| - [zuul/zuul-jobs] 935726: Fully qualify openvswitch_bridge to make linter happy https://review.opendev.org/c/zuul/zuul-jobs/+/935726 | ||
| @jim:acmegating.com | Clark: one comment on that | 22:11 |
| @clarkb:matrix.org | corvus: I didn't do that bceause ansible lets you include specific files out of roles so I didn't want to drop what it was already doing in there | 22:11 |
| @clarkb:matrix.org | that may be overly cautious and ya we could move everything up into main then only run mirror.yaml once we've decided we're going to do any work | 22:12 |
| @clarkb:matrix.org | but that was why I didn't refactor in this way | 22:12 |
| @jim:acmegating.com | yeah, i don't think the individual files are an api we're trying to support | 22:12 |
| @clarkb:matrix.org | ok I'll refactor then | 22:12 |
| @jim:acmegating.com | i think it's better to avoid the tasks, especially here since we're thinking that may end up being the normal case | 22:12 |
| @clarkb:matrix.org | the way that makes sense for refactoring to me only cuts the skipped tasks in half (not compeltely removes them) fwiw | 22:15 |
| @clarkb:matrix.org | but let me push that and you can see if that is good enough | 22:15 |
| @jim:acmegating.com | i was going for "don't include mirror.yaml if mirror_info exists and dockerhub not in mirror_info" | 22:16 |
| @clarkb:matrix.org | the linter error is still happening so I'm trying to figure that out and fix both in the same set of pushes | 22:19 |
| @clarkb:matrix.org | I think the issue is ansible removed openvswitch.openvswithc collection in ansible 11 | 22:20 |
| @clarkb:matrix.org | and so now ansible-lint is thinking you can't use it? | 22:20 |
| @clarkb:matrix.org | however we cap the ansible version we lint with which you'd think would avoid this problem | 22:21 |
| @clarkb:matrix.org | https://zuul.opendev.org/t/zuul/build/6bb81e73e4b04526a5593bf88c390a4f/log/job-output.txt#883 but we install newer ansible for some reason | 22:22 |
| @clarkb:matrix.org | linters requirements are separte | 22:23 |
| @clarkb:matrix.org | for now I'm going to cap ansible there too I think? | 22:23 |
| @clarkb:matrix.org | and add a TODO in the code | 22:23 |
| -@gerrit:opendev.org- Clark Boylan proposed: | 22:28 | |
| - [zuul/zuul-jobs] 935726: Cap the ansible version used by ansible-lint https://review.opendev.org/c/zuul/zuul-jobs/+/935726 | ||
| - [zuul/zuul-jobs] 935722: Support new style mirror_info in use-docker-mirror https://review.opendev.org/c/zuul/zuul-jobs/+/935722 | ||
| @clarkb:matrix.org | corvus: ^ is that better? | 22:28 |
| @jim:acmegating.com | Clark: yeah that might be a good balance between brevity and readability :) | 22:38 |
| @clarkb:matrix.org | corvus: the other thing I'm noticing is taht we don't directly test it | 22:42 |
| @clarkb:matrix.org | should be able to depends on it somewhere and have it exercise things though | 22:43 |
| @clarkb:matrix.org | unless that role is in a trusted job | 22:44 |
| @clarkb:matrix.org | don't think it is | 22:45 |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul] 935728: DNM exercising parent change https://review.opendev.org/c/zuul/zuul/+/935728 | 22:46 | |
| @jim:acmegating.com | yeah, it should be testable. | 22:46 |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul-jobs] 935722: Support new style mirror_info in use-docker-mirror https://review.opendev.org/c/zuul/zuul-jobs/+/935722 | 22:51 | |
| @clarkb:matrix.org | that should run ensure-docker tests too which include use-docker-mirror | 22:51 |
| @clarkb:matrix.org | it also does a docker pull so I think that will help ensure we don't regress | 22:52 |
| @clarkb:matrix.org | corvus: https://zuul.opendev.org/t/zuul/build/2112ca8e28d14635b37a27a45a636179/log/job-output.txt#311-321 it skipped everything which I didn't expect. zuul_site_mirror_fqdn should be set so now I need to look at that | 22:55 |
| @clarkb:matrix.org | ha i see it. testing ftw | 22:56 |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul-jobs] 935722: Support new style mirror_info in use-docker-mirror https://review.opendev.org/c/zuul/zuul-jobs/+/935722 | 22:57 | |
| @clarkb:matrix.org | of course now the jobs will fail bceause we're using the caching proxy and will hit rate limits :) | 22:57 |
| @clarkb:matrix.org | https://zuul.opendev.org/t/zuul/build/c0037da0260a4353967efb1ef2b00c8e/log/job-output.txt#311-328 ok I think we're covered now and it is working as expected | 23:00 |
| @clarkb:matrix.org | I think my ansible dictionary element checks may not be correct. What is the canonical way to do that? foo.bar is defined or bar in foo? | 23:31 |
| @clarkb:matrix.org | oh `is in` works because `is` is a jinjaism that you follow with a test and `in` is a jinja test | 23:33 |
| @clarkb:matrix.org | https://jinja.palletsprojects.com/en/latest/templates/#jinja-tests.in | 23:34 |
| @clarkb:matrix.org | corvus: ^ do you have a preference on how we express that in ansible? | 23:34 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: | 23:35 | |
| - [zuul/zuul] 935729: Batch fake build zookeeper additions https://review.opendev.org/c/zuul/zuul/+/935729 | ||
| - [zuul/zuul] 935730: Batch fake build database additions https://review.opendev.org/c/zuul/zuul/+/935730 | ||
| @jim:acmegating.com | Clark: no pref | 23:37 |
| @clarkb:matrix.org | ok I'll avoid another roundtrip and leave it as it is | 23:37 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!