| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul] 838843: ansible zuul_json callback : handle string results https://review.opendev.org/c/zuul/zuul/+/838843 | 08:44 | |
| @iwienand:matrix.org | corvus: Clark ^ I could replicate this independently with the zuul_json plugin. maybe someone has even more context on the logging issues there and we can make the patch even better, but that is my KISS proposal | 08:46 |
|---|---|---|
| @jpew:matrix.org | fungi: Well, the periodic pipeline using the wrong variables happened again, even after a full-reconfigure on the scheduler. Unfortunately, I can't easily reproduce it on my scheduler because theres no way to enqueue a Timer build manually from the command line or UI (that I can figure out anyway, it always fails). I did notice this however: for this particular job we have a lot of job definitions for the "master" branch; some of them are on non-master branches and have the variable the way it's actually being run. Is perhaps Zuul (or myself) confused about which of these job variants(?) should be selected? I had assumed that a Timer triggered job for the master branch would choose the variant from the master branch (and we have `branches: ["master"]` in this job config) but maybe it's choosing another one? | 13:39 |
| -@gerrit:opendev.org- Dr. Jens Harbott proposed: [zuul/nodepool] 838909: Add binutils to nodepool builder docker image https://review.opendev.org/c/zuul/nodepool/+/838909 | 14:50 | |
| @q:fricklercloud.de | ianw: ^ should fix Jammy builds. not nice from debootstrap to silently fail when no ar is there | 14:59 |
| -@gerrit:opendev.org- Oleksandr Kozachenko proposed: [zuul/zuul-jobs] 838919: Update promote-docker-image https://review.opendev.org/c/zuul/zuul-jobs/+/838919 | 15:44 | |
| -@gerrit:opendev.org- Oleksandr Kozachenko proposed: [zuul/zuul-jobs] 838919: Update promote-docker-image https://review.opendev.org/c/zuul/zuul-jobs/+/838919 | 15:52 | |
| @jim:acmegating.com | zuul-maint: how does this look for a zuul release? commit 8582dac3ade1959c39510449b2872590a4d5b654 (HEAD -> master, tag: 5.2.3, origin/master, gerrit/master) | 15:53 |
| @clarkb:matrix.org | corvus: looks like mostly bug fixes and dependency updates and backend database changes for performance | 16:01 |
| @clarkb:matrix.org | corvus: the one change that might want to be 5.3.0 is the one I wrote to remove collision errors from regex matching. But I think that is safe too since anyone with a collision would've had errors which means the change should be a noop for any existing users | 16:02 |
| @jim:acmegating.com | Clark: i agree it's borderline, but it sounds like we're both leaning toward 5.2.3...? | 16:05 |
| @clarkb:matrix.org | yes 5.2.3 is my inclination since existing users should't notice my change | 16:05 |
| @clarkb:matrix.org | its a behavior change but existing users shouldn't be impacted since we're relaxing what was an error before | 16:06 |
| @jpew:matrix.org | I have my repository set to use `merge-mode: cherry-pick` (because this is what Gerrit is doing), but it means that Zuul can't handle merge changes between branches | 16:07 |
| @jim:acmegating.com | pushed 5.2.3 | 16:08 |
| @mnaser:matrix.org | I've got a fun little thing I realize we'll run into: | 16:09 |
| https://review.opendev.org/c/zuul/zuul-jobs/+/838919 | ||
| This is a change to modify the `promote-docker-image` to instead use `skoepo` to allow promoting to things *other* than Docker Hub. However, there is an assumption that `promote-docker-image` runs without a nodeset, so these changes won't work -- but once we land the changes to allow unrestricted Ansible, they *will* work, and since zuul-jobs is not branched/released, we'll probably have to either.. break users .. or create another role? | ||
| Something to keep some thoughts about, if not already discussed.. | ||
| @jim:acmegating.com | mnaser: that might be better as a new role then? | 16:09 |
| @clarkb:matrix.org | mnaser: I thought the executor side of all that already uses skopeo | 16:09 |
| @mnaser:matrix.org | > <@clarkb:matrix.org> mnaser: I thought the executor side of all that already uses skopeo | 16:10 |
| nope, it's doing some `uri` fancy magic that is quite dockerhub-hard-coded | ||
| @jim:acmegating.com | the current version of that role is designed to use minimal api calls so that the promotion is near instantaneous | 16:10 |
| @jim:acmegating.com | it is intentionally called promote-*docker*-image | 16:11 |
| @clarkb:matrix.org | ah | 16:11 |
| @jim:acmegating.com | i suggest the name promote-container-image for the new role | 16:12 |
| @jim:acmegating.com | (reducing the delay between merging and promotion reduces the window during which zuul might use a wrong image in a build; so that's important for some users) | 16:14 |
| @mnaser:matrix.org | Yeah, I think the intention is that this can also run inside of the executor once the new Ansible changes land but as of today, it’s not | 16:15 |
| @jim:acmegating.com | yes, though it will still be slower than the docker api call (we could have implemented the original with skopeo on the executor as well but chose not to because of the delay pulling and pushing incurs) | 16:16 |
| @mnaser:matrix.org | I thought skopeo did the smart thing and didn’t pull and push but I guess not | 16:17 |
| @mnaser:matrix.org | So if we use Google artifact registry maybe it’s worth looking at how we can craft something similar with API calls only to avoid the push/pull | 16:17 |
| @jim:acmegating.com | even if it doesn't have to move the body for any layers, it should still be slower than just the re-tagging operation. it might be fast enough. i'm just saying let's leave the docker role the way it is and explore the alternative in a new role. solves all problems. :) | 16:19 |
| @jim:acmegating.com | * even if it doesn't have to move the body for any layers, it should still be slower than just the re-tagging operation. it might be fast enough for some/most purposes. i'm just saying let's leave the docker role the way it is and explore the alternative in a new role. solves all problems. :) | 16:19 |
| -@gerrit:opendev.org- Joshua Watt proposed: [zuul/zuul] 838924: merger: Handle merges with cherry-pick merge-mode https://review.opendev.org/c/zuul/zuul/+/838924 | 16:22 | |
| @avass:vassast.org | Apparently zuul-web doesn't correctly handle job names with characters that uses special characters not allowed in a url ("@", "/", ":", and so on). Is anyone opposed to url encoding/decoding job names for job pages (https://zuul.opendev.org/t/zuul/job/apply-helm-charts)? | 16:23 |
| We set up rules to make sure job names are prefixed with the project name, so they look something like: <org>/<project>@<test-command> | ||
| This way job definitions are unique and it's easy to copy/paste the test command and run it locally. | ||
| @avass:vassast.org | Also I think this is also true for tenants with special characters :) | 16:24 |
| @jim:acmegating.com | Albin Vass: that seems like something that only applies to the js app, right? | 16:25 |
| @avass:vassast.org | corvus: I think so yep, we didn't notice until we used it for a couple of weeks | 16:26 |
| @jim:acmegating.com | ie, the fix would be to have the js app just needs to encode the urls it generates internally; anyway, that sounds reasonable | 16:26 |
| @jim:acmegating.com | * ie, the fix would be to have the js app encode the urls it generates internally; anyway, that sounds reasonable | 16:27 |
| @avass:vassast.org | Yeah, i took a quick look at it yesterday and it looks like a simple fix | 16:27 |
| @jim:acmegating.com | Albin Vass: we have limited unit testing of react components -- but this might be a good one for a test since we're unlikely to encounter it otherwise. | 16:29 |
| @avass:vassast.org | Actually i think the python app also needs to decode the name. From some testing it looked like cherrypy doesn't do that automatically since it's in the path and not a query | 16:29 |
| @avass:vassast.org | I'll try find some time to push a change tomorrow with a fix :) | 16:30 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 838929: Don't report NO_JOBS to the database https://review.opendev.org/c/zuul/zuul/+/838929 | 16:48 | |
| @jim:acmegating.com | zuul-maint: sadly it was the actual release of 5.2.3 which brought that problem to my attention ^ | 16:48 |
| @jim:acmegating.com | i think we should merge that asap, roll opendev, and issue 5.2.4 | 16:48 |
| @jim:acmegating.com | (if that sounds familiar, we fixed it before, but didn't add a test, once again verifying our mantra "if it's not tested it's broken") | 16:50 |
| @clarkb:matrix.org | corvus: did that need a git add for the yaml config file? | 16:50 |
| @jim:acmegating.com | yep | 16:51 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 838929: Don't report NO_JOBS to the database https://review.opendev.org/c/zuul/zuul/+/838929 | 16:51 | |
| @clarkb:matrix.org | couple of questions inline | 16:54 |
| @clarkb:matrix.org | if you have a second to look | 16:54 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 838929: Don't report NO_JOBS to the database https://review.opendev.org/c/zuul/zuul/+/838929 | 16:55 | |
| @jim:acmegating.com | Clark: fixed | 16:55 |
| @jim:acmegating.com | copypasta | 16:55 |
| @clarkb:matrix.org | +2 thanks | 16:57 |
| @jim:acmegating.com | tristanC: ^ would you mind re-reviewing? thanks :) | 16:58 |
| @jim:acmegating.com | i'd really like to get that moving, so i've approved it under the assumption tristanC is still okay with it | 17:05 |
| @michiel:piscaer.com | I'm looking at https://docs.opendev.org/vexxhost/ansible-collection-atmosphere/latest/user/quickstart.html to see if I can use it with Zuul. It uses tox to deploy Openstack Helm. Does this mean that is also integrate easily with the tox in Zuul? | 17:37 |
| @mnaser:matrix.org | Michiel Piscaer: I think at the moment, we depend pretty heavily on collections and newer versions of Ansible that are not (yet) supported inside Zuul. Eventually, they will and we can run this stuff natively, however in the meantime, I think it's best to use a combination of tox (or directly calling `ansible-playbook`) in a Zuul job | 17:38 |
| @mnaser:matrix.org | but in due time, hopefully we can have it run natively via Zuul once collections can be a bit more native (which should be soon IMHO with all the unrestricted ansible changes coming in) | 17:39 |
| @michiel:piscaer.com | ok, thanks. | 17:41 |
| -@gerrit:opendev.org- Joshua Watt proposed: [zuul/zuul] 838924: merger: Handle merges with cherry-pick merge-mode https://review.opendev.org/c/zuul/zuul/+/838924 | 17:46 | |
| @clarkb:matrix.org | corvus: the image builds for 838929 will use the assemble update. I don't expect problems based on my local checking but thought I'd call that out | 17:55 |
| @jim:acmegating.com | Clark: thanks, gtk :) | 18:01 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 838933: Add a gearman-style backoff for executor job starts https://review.opendev.org/c/zuul/zuul/+/838933 | 18:19 | |
| -@gerrit:opendev.org- Joshua Watt proposed: [zuul/zuul] 838924: merger: Handle merges with cherry-pick merge-mode https://review.opendev.org/c/zuul/zuul/+/838924 | 18:26 | |
| -@gerrit:opendev.org- Zuul merged on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul] 838929: Don't report NO_JOBS to the database https://review.opendev.org/c/zuul/zuul/+/838929 | 19:27 | |
| @jim:acmegating.com | that's ^ running on opendev now | 20:02 |
| @clarkb:matrix.org | and a hard refresh on the zuul status page returns a working website. assemble updates look good | 20:18 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 838933: Add a gearman-style backoff for executor job starts https://review.opendev.org/c/zuul/zuul/+/838933 | 21:13 | |
| @spamaps:spamaps.ems.host | https://news.ycombinator.com/item?id=31108828 <-- A zuul-able conversation. | 21:19 |
| @clarkb:matrix.org | Someone has mentioned zuul at least | 21:32 |
| @clarkb:matrix.org | ianw I left a thought on the zuul json change. I suspect its fine but theorized a way it might not be if we can test that better | 22:19 |
| @clarkb:matrix.org | zuulians https://review.opendev.org/c/zuul/zuul/+/837852 got a new patchset. I think corvus wanted some broad agreement on that one so would be good to rereview | 23:04 |
| @jim:acmegating.com | zuul-maint: how does this look for a zuul release? commit dd0135baa51cdf21a18831926c04227caa060878 (HEAD -> master, tag: 5.2.4, origin/master, gerrit/master, refs/changes/29/838929/3) | 23:42 |
| @iwienand:matrix.org | Clark: yeah the type is AnsibleUnsafeText at that point, I think | 23:53 |
| @iwienand:matrix.org | i'm trying to think of something that sets results, but doesn't actually install packages for testing | 23:54 |
| @iwienand:matrix.org | (sets results as a list of strings) | 23:55 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!