| *** tosky has quit IRC | 00:00 | |
| openstackgerrit | Mohammed Naser proposed zuul/zuul-jobs master: chart-testing-lint-single: new job https://review.opendev.org/715665 | 00:03 |
|---|---|---|
| *** rf0lc0 has quit IRC | 01:28 | |
| *** swest has quit IRC | 01:39 | |
| *** swest has joined #zuul | 01:54 | |
| *** bhavikdbavishi has joined #zuul | 02:13 | |
| *** bhavikdbavishi has quit IRC | 02:18 | |
| *** evrardjp has quit IRC | 04:36 | |
| *** bhavikdbavishi has joined #zuul | 04:36 | |
| *** evrardjp has joined #zuul | 04:36 | |
| *** bhavikdbavishi1 has joined #zuul | 04:39 | |
| *** bhavikdbavishi has quit IRC | 04:41 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 04:41 | |
| *** bhavikdbavishi has quit IRC | 06:31 | |
| *** sgw has quit IRC | 07:59 | |
| *** bhavikdbavishi has joined #zuul | 08:43 | |
| *** bhavikdbavishi has quit IRC | 10:26 | |
| *** bhavikdbavishi has joined #zuul | 10:42 | |
| *** klindgren_ has joined #zuul | 10:47 | |
| *** klindgren has quit IRC | 10:47 | |
| *** lennyb has quit IRC | 10:48 | |
| *** lennyb has joined #zuul | 10:48 | |
| *** tosky has joined #zuul | 10:52 | |
| *** bhavikdbavishi has quit IRC | 10:53 | |
| *** sshnaidm|off has quit IRC | 10:58 | |
| *** sshnaidm has joined #zuul | 11:03 | |
| *** sshnaidm has quit IRC | 12:17 | |
| *** sshnaidm has joined #zuul | 12:17 | |
| *** Goneri has joined #zuul | 13:59 | |
| *** sshnaidm is now known as sshnaidm|afk | 14:07 | |
| *** sshnaidm|afk is now known as sshnaidm | 14:25 | |
| *** rfolco has joined #zuul | 15:40 | |
| openstackgerrit | Mohammed Naser proposed zuul/zuul-jobs master: helm: collect kubernetes logs in post https://review.opendev.org/715709 | 15:40 |
| *** evrardjp has quit IRC | 16:36 | |
| *** evrardjp has joined #zuul | 16:36 | |
| *** sgw has joined #zuul | 17:42 | |
| *** adam_g has quit IRC | 17:54 | |
| *** adam_g has joined #zuul | 18:02 | |
| mnaser | hmm -- i've got an interesting and tricky problem. i've got a repo where i build a docker image only if the Dockerfile changes | 18:22 |
| mnaser | if i make use of the use-buildset-registry role within the job, and the job to build the docker image did not run, i end up with an error | 18:23 |
| mnaser | if i don't make use of the buildset registry role in the job, then that means the image will not be pulled in from thee buildset registry _even if_ the dockerfile changed and the image was built | 18:24 |
| mnaser | you can kinda see this issue here: https://review.opendev.org/#/c/715662/2 and https://review.opendev.org/#/c/715705/5 -- the first change adds the docker image build job (only if the dockerfile changes), the second one runs the functional tests which need the image | 18:25 |
| mnaser | i could just technically build the image all the time but that seems like a waste of time, should i be instead starting a buildset registry _within_ my functional job? but that means i can end up with two buildset registries? | 18:26 |
| corvus | mnaser: i think the current 'best' solution to that is to always run the buildset registry in its own job. the image build job can list it as a dependency, so it will only start building the image once the registry is up and running (and it will automatically not run the buildset registry in the image build job if it's already running in a previous job). then the deployment job can list the buildset | 18:33 |
| corvus | registry as a hard dependency, and the image build job as a soft dependency (so iff the build job runs, the deploy job will wait for it, but it will always wait for the buildset registry to be running) | 18:33 |
| mnaser | corvus: ok, so anytime we have "consumers" for images, it's best to start up a buildset registry | 18:34 |
| corvus | mnaser: yeah, at least, i think if the producers aren't guaranteed (we always build the zuul image for instance, so we always know in zuul-quick-start that zuul-build-image will have a registry running) | 18:36 |
| corvus | mnaser: also, you should probably inherit from opendev-buildset-registry-consumer in the consumer job | 18:39 |
| corvus | mnaser: (that will invoke use-buildset-registry, and also pull the images into it from the intermediate registry) | 18:39 |
| mnaser | corvus: would that be the same as having use-buildset-registry in the role? | 18:40 |
| corvus | mnaser: there's another important role, 'pull-from-intermediate-registry' which is used to make the cross-change dependency stuff work (so if change A builds the image, and change B runs the image, change B gets the image from change A). that requires a secret, so you have to inherit from an opendev base job that already has it. | 18:41 |
| mnaser | corvus: hmm, okay, so at the time i added something the the install-docker role to automatically run use-buildset-registry but it didn't include pull-from-intermediate-registry | 18:42 |
| corvus | mnaser: depending on the complexity of your consumer job, you may end up wanting to use multiple inheritance to get that. we did that in system-config. so if you take a look at your situation, and can't draw a straight line through opendev-buildset-registry-consumer, then you may want to take a look at how we did that for the system-config jobs here: | 18:43 |
| corvus | https://opendev.org/opendev/system-config/src/branch/master/.zuul.yaml#L898-L917 | 18:43 |
| mnaser | (i wanted to make it easier to reuse the existing helm jobs and simply adding a var that says "use buildset registry") | 18:43 |
| mnaser | does that mean https://opendev.org/zuul/zuul-jobs/commit/ae05322a857730aa0351f586c625fbd618b3ad15 is not really effective for much? | 18:44 |
| corvus | mnaser: well, you might be able to get almost the same thing by just having a new job with the extra inheritance path. so "apply-helm" and "vexxhost-apply-helm-buildset-registry" or whatever | 18:44 |
| mnaser | ok, interesting | 18:45 |
| corvus | mnaser: re the install-docker role -- halfway -- it is useful within a single buildset, but still something needs to run pull-from-intermediate-registry to get an image build in a different buildset | 18:45 |
| mnaser | yeah so it's effective in terms of pulling images _within_ the same buildset but will probably not be effective in pulling an image change from another one, ok | 18:46 |
| corvus | yup | 18:46 |
| mnaser | ok, let me try this multi-parent thing with the functional job and see what happens after this run finishes | 18:49 |
| *** Goneri has quit IRC | 18:49 | |
| *** Goneri has joined #zuul | 19:13 | |
| mnaser | does anyone know why this failed at the time? https://review.opendev.org/#/c/611549/ | 20:21 |
| mnaser | it seems like a really neat cool feature, but the log server is gone and the reference to why it failed doesn't help | 20:21 |
| AJaeger | mnaser: maybe we discussed it on IRC? Best check eavesdrop for #openstack-infra and #zuul for that time | 20:27 |
| mnaser | AJaeger: yeah that's likely, okay, i'll try to dig for it | 20:28 |
| AJaeger | mnaser: I only remember that the job failed but don't have details why. Yeah, a lovely feature if it works... | 20:29 |
| corvus | mnaser: iirc, i think there was an error which broke zuul and caused a whole bunch of pep8 jobs to fail when they shouldn't have, but zuul is now hardened against it now; it should be okay to revert and test out, but probably with a bunch of depends-on, because i don't know if we found the problem with pep8 that triggered it in the first place. | 21:01 |
| mnaser | corvus: i see, i found a change inside openstacksdk which depended on it and it was working there, maybe that might be a good example to try out | 21:03 |
| corvus | mnaser: yeah, but we should also probably check nova and other things -- it was tested somewhat the first time | 21:04 |
| corvus | (so a single "ok" probably isn't enough to have confidence) | 21:04 |
| mnaser | corvus: fair enough, i'll workout a few changes with some "failures" | 21:04 |
| corvus | mnaser: and probably some successes too :) | 21:09 |
| openstackgerrit | Mohammed Naser proposed zuul/zuul-jobs master: Revert "Revert "Extract pep8 messages for inline comments"" https://review.opendev.org/715727 | 21:09 |
| mnaser | corvus: yep :) | 21:12 |
| tobiash | Next time I'll put traces into the commit message :) | 21:15 |
| corvus | i don't think we expected it to stay reverted out so long | 21:16 |
| tobiash | But as far as I think I remember there was a parse error for some pep8 outputs | 21:16 |
| tobiash | But don't know anymore. Could be mypy tests which also run in pep8 jobs | 21:17 |
| mnaser | tobiash: perhaps you can run a depends-on that at your side too? | 21:23 |
| mnaser | the linters failure on that right now is just the revert acting up, we probably don't need to add zuul into the jobs | 21:24 |
| tobiash | mnaser: I can do tomorrow, but I think it was more an emergency revert for openstack projects outside us working hours | 21:25 |
| tobiash | http://eavesdrop.openstack.org/irclogs/%23zuul/%23zuul.2018-10-18.log.html#t2018-10-18T08:59:37 | 21:27 |
| corvus | tobiash: do you think this was the error? that maybe the problem was just on repos running ansible lint with zuul_return? http://eavesdrop.openstack.org/irclogs/%23zuul/%23zuul.2018-10-18.log.html#t2018-10-18T09:13:30 | 21:32 |
| *** hashar has joined #zuul | 21:32 | |
| tobiash | corvus: I think the missing lines translation was the error: http://eavesdrop.openstack.org/irclogs/%23openstack-infra/%23openstack-infra.2018-10-18.log.html#t2018-10-18T11:12:38 | 21:33 |
| tobiash | If that's the case zuul should handle it correct now | 21:33 |
| corvus | tobiash: ah, good find. that sounds familiar. so we may be all set. :) | 21:34 |
| corvus | (should still test of course, but i feel better about it :) | 21:34 |
| tobiash | And there was also hardening in zuul for invalid results from the job in the meantime | 21:34 |
| corvus | the "file 2018-10-18 10" part looks like a parse error | 21:35 |
| corvus | i wonder if the part that caused that got fixed | 21:35 |
| tobiash | I think zuul now validates what it receives from the job and ignores invalid stuff like this | 21:36 |
| corvus | yeah, but i wonder what generated that in the first palec | 21:37 |
| corvus | place | 21:37 |
| tobiash | I guess it wrongly parsed non pep8 output | 21:37 |
| mnaser | https://review.opendev.org/#/c/611552/ was the change it failed to report on | 21:38 |
| mnaser | in this case, perhaps openstack-zuul-jobs-linters running there caused the issue | 21:39 |
| *** hashar has quit IRC | 21:39 | |
| tobiash | Is that job running yamllint or ansible-lint? | 21:40 |
| mnaser | it just runs 'linters' env, let me check | 21:40 |
| mnaser | (might be good to make a patch against there then i guess) | 21:41 |
| mnaser | tobiash: the answer is many things -- https://opendev.org/openstack/project-config/src/branch/master/tox.ini#L35-L50 | 21:41 |
| tobiash | Commented on the re-revert | 21:46 |
| tobiash | I thing that was the bug that caused the revert in the first place | 21:46 |
| corvus | i also left a comment, which i think will address the linters error | 21:52 |
| tobiash | Oh now I understand, it really misparsed something as a correct line comment and zuul reported this to gerrit without further validation (which is now performed as part of the line translation) | 21:52 |
| tobiash | My initial comment was bogus, too late here ;) | 21:56 |
| openstackgerrit | Merged zuul/nodepool master: DiskImage : alphabetise entries https://review.opendev.org/713381 | 22:01 |
| openstackgerrit | Mohammed Naser proposed zuul/zuul-jobs master: Revert "Revert "Extract pep8 messages for inline comments"" https://review.opendev.org/715727 | 22:03 |
| openstackgerrit | Merged zuul/nodepool master: diskimage: make name primary key https://review.opendev.org/713382 | 22:08 |
| openstackgerrit | Merged zuul/nodepool master: Add parent and abstract flags for diskimages https://review.opendev.org/713157 | 22:08 |
| openstackgerrit | Merged zuul/nodepool master: Move config merge into DiskImage object https://review.opendev.org/713550 | 22:08 |
| mnaser | interesting, i wonder if "Depends-On: http://..." doesn't get picked up and "Depends-On: http://" does | 22:36 |
| *** tflink has quit IRC | 22:37 | |
| mnaser | i think i just saw a case of that happening | 22:37 |
| *** tflink has joined #zuul | 22:37 | |
| mnaser | yep, the regex implies a "single" whitespace | 22:38 |
| mnaser | that seems a little user unfriendly, thoughts? | 22:39 |
| *** SpamapS has quit IRC | 22:44 | |
| *** SpamapS has joined #zuul | 22:45 | |
| *** tosky has quit IRC | 22:54 | |
| openstackgerrit | Mohammed Naser proposed zuul/zuul-jobs master: Revert "Revert "Extract pep8 messages for inline comments"" https://review.opendev.org/715727 | 23:08 |
| mnaser | cool, figured out the root cause, used some regex to make it a bit more strict | 23:09 |
| openstackgerrit | Mohammed Naser proposed zuul/zuul-jobs master: Revert "Revert "Extract pep8 messages for inline comments"" https://review.opendev.org/715727 | 23:22 |
| mnaser | wee, okay, that works much better | 23:56 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!