| *** thuvh1 is now known as thuvh | 03:47 | |
| dtantsur | Hi folks, could you clarify this failure? https://zuul.opendev.org/t/openstack/buildset/a41346f64bd04bc2b48f7c837d37a737 | 14:40 |
|---|---|---|
| dtantsur | (I hope it does not mean that dependencies no longer work across projects) | 14:42 |
| fungi | dtantsur: it means that zuul isn't aware of a relationship between those projects and so won't enqueue the later change until the dependency has merged | 14:43 |
| dtantsur | fungi: what sort of a relationship is it looking for? and why is it a failure, not just a delay? | 14:44 |
| fungi | normally it just wouldn't enqueue the later change, but we've been doing a bunch of gate queue reordering and dequeuing/enqueuing to prioritize the ossa-2024-001 fixes so they merge sooner | 14:44 |
| dtantsur | ahhh, so it's a one-off situation? | 14:45 |
| fungi | possible we enqueued it by mistake, i'll have to look at the timeline | 14:45 |
| dtantsur | fungi: don't worry, it's not worth your time then. Thank you for explanation. | 14:45 |
| fungi | ah, i think zuul does try to enqueue it but immediately rejects it because the dependency hasn't merged yet, hence the -2 | 14:46 |
| fungi | if you declared a shared queue for openstack/ironic-inspector and openstack/python-ironic-inspector-client then they could be enqueued together | 14:46 |
| dtantsur | fungi: haven't we recently removed such a queue because it's no longer supported? Oo | 14:46 |
| dtantsur | in any case, a better error message would be nice to have | 14:47 |
| dtantsur | I mean, https://opendev.org/openstack/ironic-lib/commit/179772a05a08b2e826509c8de22679e467698f07 etc | 14:47 |
| frickler | dtantsur: iiuc you should have moved the queue definition to the project level instead of dropping it completely | 14:49 |
| dtantsur | oh | 14:49 |
| dtantsur | JayF: we have misunderstood it ^^^^ | 14:49 |
| dtantsur | also cc rpittau | 14:49 |
| fungi | dtantsur: depends on how you define "recently" but after zuul 7.0.0 it was necessary to move queue declarations from pipelines to projects: https://zuul-ci.org/docs/zuul/latest/releasenotes.html#relnotes-7-0-0-upgrade-notes | 14:49 |
| frickler | "you" as in the ironic team | 14:49 |
| fungi | "The deprecated syntax of specifying project change queues on pipeline configurations has been removed. Specify queues using the project stanza now. See queue for more information." | 14:49 |
| dtantsur | Thank you frickler and fungi! Would it be possible to improve the error message for the time being? I imagine newcomers very confused by the current message. | 14:50 |
| fungi | i think the reason behind that failure depends on a fair amount of context outside zuul itself, so zuul is limiting the error message to explaining why zuul rejected gating the change | 14:51 |
| fungi | it could be that you intended them to gate together but didn't declare a shared queue for them (or messed it up when you did), or it could be that they're not meant to be gated together and the second change was approved prematurely | 14:52 |
| frickler | yes, I'm not sure how the wording of that message could be improved | 14:52 |
| dtantsur | "The change cannot be put in the gate pipeline because one of its dependencies has not merged yet, and the projects don't share the same queue". followed by the warning you already output. | 14:53 |
| fungi | i'm still not convinced that trying to enqueue the change and then immediately creating a gate failure is entirely user-friendly, but neither was ignoring the approval (one of the past behaviors for that situation) | 14:53 |
| dtantsur | yeah | 14:54 |
| dtantsur | it's also confusing that the only explanation of the error is prefixed as a warning. so I was wondering it's the cause of the failure or just a part of the problem or completely unrelated. | 14:54 |
| dtantsur | at the very least, something should start with "Fatal: cannot enqueue the change", even if we cannot provide a better explanation because of the missing context for Zuul. | 14:55 |
| fungi | dtantsur: https://opendev.org/zuul/zuul/src/branch/master/zuul/manager/dependent.py#L213-L223 is where you'd propose alternative wording, i think | 14:58 |
| dtantsur | thanks, will make a mental note | 14:58 |
| fungi | i think the reason it's a "warning" is because zuul has added a framework for reporting additional "warning" text in review comments | 15:00 |
| fungi | but yes calling them all "warnings" could be misleading in some situations | 15:00 |
| fungi | they were originally intended to be able to warn reviewers about situations alongside successful results | 15:01 |
| rpittau | bye everyone, have a great weekend! o/ | 15:12 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!