| *** cdearborn has quit IRC | 00:39 | |
| *** Goneri has quit IRC | 00:47 | |
| *** rlandy has quit IRC | 01:17 | |
| *** sanjayu__ has joined #zuul | 01:18 | |
| *** sanjayu_ has quit IRC | 01:20 | |
| *** swest has quit IRC | 01:47 | |
| *** swest has joined #zuul | 02:00 | |
| *** evrardjp has quit IRC | 04:36 | |
| *** evrardjp has joined #zuul | 04:36 | |
| *** sgw has quit IRC | 04:59 | |
| *** sgw has joined #zuul | 05:16 | |
| openstackgerrit | Andreas Jaeger proposed zuul/zuul-jobs master: Add new non-npm specific javascript jobs https://review.opendev.org/726547 | 05:38 |
|---|---|---|
| openstackgerrit | Andreas Jaeger proposed zuul/zuul-jobs master: Extract ensure-javascript-build-tool role https://review.opendev.org/726900 | 05:38 |
| openstackgerrit | Andreas Jaeger proposed zuul/zuul-jobs master: Add new non-npm specific javascript jobs https://review.opendev.org/726547 | 05:53 |
| openstackgerrit | Andreas Jaeger proposed zuul/zuul-jobs master: Extract ensure-javascript-build-tool role https://review.opendev.org/726900 | 05:53 |
| *** dpawlik has joined #zuul | 06:05 | |
| AJaeger | mordred, amotoki, could you check https://review.opendev.org/726940, please? Do we need to fix that one failure or can we switch to new jobs? I think the changes 726900 and 726547 are ready to merge. | 07:01 |
| *** avass has joined #zuul | 07:07 | |
| *** jcapitao has joined #zuul | 07:12 | |
| *** sanjayu__ has quit IRC | 07:16 | |
| *** saneax has joined #zuul | 07:20 | |
| *** tosky has joined #zuul | 07:35 | |
| *** guillaumec has joined #zuul | 07:37 | |
| *** fbo|off is now known as fbo|afk | 07:37 | |
| openstackgerrit | Simon Westphahl proposed zuul/zuul master: Replace preexec_fn argument in executor popen call https://review.opendev.org/727089 | 07:40 |
| openstackgerrit | Simon Westphahl proposed zuul/zuul master: Replace preexec_fn argument in executor popen call https://review.opendev.org/727089 | 07:41 |
| openstackgerrit | Simon Westphahl proposed zuul/zuul master: Replace preexec_fn argument in executor popen call https://review.opendev.org/727089 | 07:42 |
| tobiash | zuul-maint: we see occasional deadlocks of subprocess.Popen in production (maybe 1/100.000 jobs). This is supposed to fix that ^ | 07:46 |
| *** jpena|off is now known as jpena | 07:56 | |
| *** rpittau|afk is now known as rpittau | 07:58 | |
| *** nils has joined #zuul | 08:13 | |
| openstackgerrit | Felix Edel proposed zuul/zuul master: Dequeue changes via github checks API https://review.opendev.org/709135 | 08:19 |
| *** sshnaidm|afk is now known as sshnaidm | 08:52 | |
| *** guillaumec has quit IRC | 08:58 | |
| *** guillaumec has joined #zuul | 09:02 | |
| tobiash | mordred: after py38 upgrade of our executors the upload-logs-swift role doesn't work anymore | 09:15 |
| tobiash | not sure about the reason yet because of no_log | 09:16 |
| openstackgerrit | Albin Vass proposed zuul/zuul master: Fix unused argument triggering flake8 https://review.opendev.org/727127 | 09:21 |
| avass | tobiash: I noticed the replace preexec_fn change was failing linting, can you take a look at that ^ ? | 09:21 |
| avass | tobiash: Not sure why that started erroring now | 09:22 |
| tobiash | oh weird | 09:23 |
| avass | yeah | 09:23 |
| tobiash | but that change might break the format call as the format specifier hasn't msg | 09:25 |
| avass | tobiash: but that log never used the msg anyway | 09:27 |
| tobiash | avass: nevermind, I read that the wrong way | 09:28 |
| tobiash | need more coffee ;) | 09:28 |
| avass | :) | 09:28 |
| tobiash | totally right, we supply msg but that's not in the format string and thus must be removed | 09:29 |
| tobiash | there was a flake8 release 4 hours ago | 09:30 |
| tobiash | I guess that catches this issue | 09:30 |
| *** jhesketh has joined #zuul | 09:31 | |
| tobiash | actually yesterday as well | 09:31 |
| tobiash | confirmed, flake8 3.7.9 did not find this issue, starting with 3.8.0 (released yesterday) it reports this issue | 09:32 |
| avass | ah yep, that's it | 09:34 |
| tobiash | mordred: this fixes the current zuul gate due to flake8 release: https://review.opendev.org/727127 | 09:34 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Ensure output dirs are empty https://review.opendev.org/727135 | 09:54 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Ensure output dirs are empty https://review.opendev.org/727135 | 10:03 |
| avass | zuul-jobs-maint: We noticed that zuul-output was never emptied, do we want to do that ^ in ensure-output-dirs or in a separate role? | 10:04 |
| *** bhavikdbavishi has joined #zuul | 10:08 | |
| tobiash | mordred: looks like loop is broken with py38 and ansible <2.8 in some cases. It gets a list and complains that it didin't get a list. | 10:15 |
| tobiash | sounds like this https://github.com/ansible/ansible/issues/65451 | 10:15 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Ensure output dirs are empty https://review.opendev.org/727135 | 10:15 |
| avass | tobiash: wow, that's not good | 10:18 |
| *** rpittau is now known as rpittau|bbl | 10:19 | |
| tobiash | zuul-maint: I guess we'd need to either deprecate/remove ansible 2.6 and 2.7 or revert the executor to <3.8 again | 10:20 |
| *** dpawlik has quit IRC | 10:21 | |
| *** dpawlik has joined #zuul | 10:21 | |
| *** bhavikdbavishi1 has joined #zuul | 10:34 | |
| *** bhavikdbavishi has quit IRC | 10:36 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 10:36 | |
| tobiash | and at least under macos it seems that even ansible 2.8 has problems with python 3.8 | 10:38 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: DNM: Test ansible loop with python 3.8 https://review.opendev.org/727155 | 10:47 |
| tobiash | zuul-maint: this is what fails locally with python 3.8 and ansible<2.8 ^ | 10:48 |
| *** jcapitao is now known as jcapitao_lunch | 10:51 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Remove support for ansible 2.6 https://review.opendev.org/727157 | 11:09 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Drop support for ansible 2.6 https://review.opendev.org/727158 | 11:13 |
| *** rfolco has joined #zuul | 11:16 | |
| *** smyers has quit IRC | 11:17 | |
| *** smyers has joined #zuul | 11:18 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: DNM: Test ansible loop with python 3.8 https://review.opendev.org/727155 | 11:25 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: DNM: Test ansible loop with python 3.8 https://review.opendev.org/727155 | 11:26 |
| *** ysandeep has joined #zuul | 11:29 | |
| *** jpena is now known as jpena|lunch | 11:33 | |
| *** guillaumec has quit IRC | 11:38 | |
| *** tumble has joined #zuul | 11:43 | |
| *** fbo|afk has quit IRC | 11:47 | |
| *** hashar has joined #zuul | 11:52 | |
| *** bhavikdbavishi has quit IRC | 12:16 | |
| *** rlandy has joined #zuul | 12:18 | |
| *** bhavikdbavishi has joined #zuul | 12:18 | |
| *** rpittau|bbl is now known as rpittau | 12:21 | |
| *** jcapitao_lunch is now known as jcapitao | 12:23 | |
| tristanC | tobiash: what is the purpose of not using ansible-2.9 ? Is there known issues or specific cases that prevents using 2.9? | 12:25 |
| *** rfolco is now known as rfolco|rover | 12:27 | |
| tobiash | tristanC: we're still on 2.7 for most projects. What we found out already is that 2.9 changes some checksum stuff | 12:28 |
| tobiash | and a fetch with winrm seems to fail on 2.9 (pending analysis) | 12:29 |
| *** Goneri has joined #zuul | 12:29 | |
| tristanC | tobiash: what is the checksum change? is it only the removal of the default md5 scheme? | 12:31 |
| tobiash | tristanC: get_uri doesn't seem to accept md5 anymore | 12:31 |
| tristanC | tobiash: oh ok, thank you for the details. | 12:33 |
| *** zbr_ has joined #zuul | 12:35 | |
| tristanC | tobiash: what about ansible-2.8, does it has a compatible checksum and no winrm bug? | 12:35 |
| tristanC | zuul-maint: i'd rather keep python-3.8 support and drop incompatible ansible version than the otherway around | 12:35 |
| tobiash | tristanC: looks like ansible 2.8 doesn't break the winrm task but also shows an md5 issue | 12:35 |
| AJaeger | tobiash: it's time to get rid of md5 - or do you have external APIs that need it? | 12:36 |
| tobiash | so I guess we'll switch to 2.8 (the checksum issue seems easy to fix) and then further analyze the winrm problem | 12:36 |
| tobiash | AJaeger: it's just a download task in one of our base playbooks | 12:36 |
| *** zbr_ has quit IRC | 12:36 | |
| *** jpena|lunch is now known as jpena | 12:40 | |
| *** Goneri has quit IRC | 12:41 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Remove support for ansible 2.6 https://review.opendev.org/727157 | 12:46 |
| *** Goneri has joined #zuul | 12:47 | |
| tobiash | tristanC, AJaeger: our checksum issue only hits us with win_get_url | 12:51 |
| tobiash | we used already sha256, but the module added a new checksum_algorithm argument in 2.8 that defaults to sha1 :/ | 12:52 |
| tobiash | https://docs.ansible.com/ansible/latest/modules/win_get_url_module.html#parameter-checksum_algorithm | 12:52 |
| *** mhu has joined #zuul | 12:56 | |
| *** fbo has joined #zuul | 13:00 | |
| openstackgerrit | Merged zuul/zuul master: Fix unused argument triggering flake8 https://review.opendev.org/727127 | 13:08 |
| *** sanjayu_ has joined #zuul | 13:10 | |
| mordred | AJaeger: there's nothing wrong with md5 as a checksum | 13:10 |
| mordred | md5 as a security-related hash isn't good | 13:11 |
| *** saneax has quit IRC | 13:12 | |
| mordred | AJaeger, amotoki: I thnik let's switch horizon to the new jobs | 13:16 |
| AJaeger | mordred: want to review your check and confirm that my changes are fine? | 13:16 |
| *** guillaumec has joined #zuul | 13:16 | |
| mordred | tobiash: that is "thrilling" wrt ansible <2.8 and py38 :( | 13:16 |
| tobiash | mordred: since our official images are now py38 I guess we need to do something | 13:17 |
| mordred | tobiash: yeah. I'm beginning to wonder if our overall policy of supporting ansible versions in zuul for much longer than upstream does is sustainable long-term | 13:20 |
| mordred | even though I think our reasons for making that choice are good ones | 13:20 |
| mordred | AJaeger: your updates to https://review.opendev.org/#/c/726547 lgtm - I only left a +1 but I think it's ok for you to leave a +2 on that one | 13:22 |
| AJaeger | mordred: so, you leave 1.5, I leave 1.5 ;) | 13:22 |
| AJaeger | avass: want to review 726547 again, please? | 13:23 |
| avass | zuul-jobs-maint: we should merge https://review.opendev.org/#/c/726448/ today | 13:23 |
| avass | AJaeger: sure! | 13:23 |
| AJaeger | thanks, avass | 13:24 |
| AJaeger | avass: thanks for rememind us of 726448 - gave my | 13:24 |
| AJaeger | avass: thanks for rememind us of 726448 - gave my +2 | 13:24 |
| avass | AJaeger, mordred: lgtm | 13:25 |
| mordred | AJaeger: are we comfortable that 726448 isn't going to cause total havoc for openstack? | 13:26 |
| AJaeger | mordred: avass pushed changes for all repos. | 13:27 |
| avass | mordred: we should have fixed all occurances of install-* by now | 13:27 |
| AJaeger | those ar eall merged | 13:27 |
| mordred | cool | 13:27 |
| avass | at least the ones we found :) | 13:27 |
| AJaeger | mordred: check depends-on on https://review.opendev.org/#/c/719322/6 | 13:27 |
| mordred | avass: while we're landing things, what do you think of https://review.opendev.org/#/c/726547 | 13:28 |
| avass | mordred: already gave my +3 :) | 13:28 |
| mordred | \o/ | 13:28 |
| mordred | yoctozepto: new javascript jobs landing - makes me think (re your js-openstack-lib functional job) - maybe we should make a job like that in openstack-zuul-jobs | 13:29 |
| AJaeger | avass: want to review https://review.opendev.org/726900 as well, please? | 13:30 |
| avass | AJaeger: sure | 13:31 |
| AJaeger | thx | 13:31 |
| avass | AJaeger: it looks good, but tests would have been nice :) | 13:32 |
| AJaeger | mordred: what do you think of adding tests for the javascript work? I agree with avass on that one ^ | 13:33 |
| openstackgerrit | Merged zuul/zuul-jobs master: Add new non-npm specific javascript jobs https://review.opendev.org/726547 | 13:37 |
| mordred | AJaeger: I looked for existing tests for the old stuff that I could copy and update and didn't find any | 13:38 |
| mordred | but yes - we should almost certainly add some new tests | 13:38 |
| AJaeger | mordred: want to +2A this change to use the new jobs? https://review.opendev.org/#/c/726747/3/zuul.d/project-templates.yaml | 13:39 |
| openstackgerrit | Merged zuul/zuul-jobs master: Fail and direct user to use ensure-* version of roles https://review.opendev.org/726448 | 13:42 |
| *** openstackstatus has quit IRC | 13:53 | |
| *** openstackstatus has joined #zuul | 13:55 | |
| *** ChanServ sets mode: +v openstackstatus | 13:55 | |
| mordred | AJaeger: done | 13:57 |
| AJaeger | thanks, mordred | 13:59 |
| *** bhavikdbavishi has quit IRC | 13:59 | |
| mordred | AJaeger: I'm glad that we happened to work on some new javascript jobs over the weekend - it was good timing for this horizon work :) | 14:00 |
| AJaeger | indeed.. | 14:03 |
| avass | now we just need to somehow fix our dashboard bump :) | 14:05 |
| openstackgerrit | Felix Edel proposed zuul/zuul master: Dequeue changes via github checks API https://review.opendev.org/709135 | 14:06 |
| avass | I still haven't figured out a way to force it to not optimize those files away | 14:06 |
| AJaeger | mordred: https://review.opendev.org/727202 uses the new jobs in horizon, could you double check, please? | 14:10 |
| mordred | AJaeger: ++ | 14:13 |
| AJaeger | thanks | 14:13 |
| corvus | do we know why the linter job is raising new errors: | 14:19 |
| corvus | ./zuul/ansible/2.9/callback/zuul_stream.py:665:23: F522 '...'.format(...) has unused named argument(s): msg | 14:19 |
| AJaeger | corvus: new flake8 release | 14:20 |
| AJaeger | corvus: version 3.8.0 and 3.8.1 came out the last 12 hours or so | 14:21 |
| corvus | ah yep, i have 3.7.9 locally | 14:21 |
| corvus | that doesn't look like a terrible error | 14:21 |
| corvus | i'll just fix it | 14:21 |
| AJaeger | corvus: 727127 merged, didn't that fix it? | 14:21 |
| AJaeger | corvus: please pull first! | 14:21 |
| mordred | corvus: yeah - the new format checker actually caught a real bug in openstacksdk | 14:22 |
| avass | AJaeger, corvus: yep, that should already be fixed :) | 14:23 |
| corvus | mordred: https://review.opendev.org/726554 is parented to outdated https://review.opendev.org/726553 | 14:24 |
| openstackgerrit | Monty Taylor proposed zuul/zuul master: Update to new javascript jobs https://review.opendev.org/726554 | 14:25 |
| mordred | corvus: fixed! | 14:25 |
| *** bhavikdbavishi has joined #zuul | 14:26 | |
| *** ysandeep is now known as ysandeep|afk | 14:26 | |
| AJaeger | corvus: could you review https://review.opendev.org/#/c/726900 as well for javascript jobs, please? | 14:27 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Remove support for ansible 2.6 https://review.opendev.org/727157 | 14:28 |
| *** y2kenny has joined #zuul | 14:31 | |
| y2kenny | I am getting a lot of "This change depends on a change that failed to merge" on branches that has no jobs. Is there a way to silence them? | 14:34 |
| *** hashar has quit IRC | 14:40 | |
| openstackgerrit | Felix Edel proposed zuul/zuul master: Report canceled changes via Github checks API https://review.opendev.org/711023 | 14:43 |
| mordred | AJaeger: I guess we should do similar to your horizon patch for the stable banches, yes? | 14:45 |
| corvus | y2kenny: but the project is in the pipeline on other branches? | 14:48 |
| y2kenny | corvus: yes | 14:48 |
| AJaeger | mordred: yeah, I want to backport and adjust as needed | 14:48 |
| mordred | AJaeger: ++ | 14:48 |
| openstackgerrit | Felix Edel proposed zuul/zuul master: Allow check runs to be configured as required status in pipeline config https://review.opendev.org/711241 | 14:52 |
| corvus | y2kenny: is there any chance that project-branch has a project-pipeline entry, possibly in a config project, maybe in a project stanza with a regex that applies to a group of projects or all projects? | 14:53 |
| *** felixedel has joined #zuul | 14:55 | |
| felixedel | Hi all, I've updated/rebased a few patches regarding the Github checks API https://review.opendev.org/#/q/status:open+project:zuul/zuul+branch:master+topic:github-checks-api that are open for a few weeks already. Since I was out of office for two months, I did not follow up on them in the meantime. But now it would be great to get some reviews on those :) | 14:58 |
| openstackgerrit | Felix Edel proposed zuul/zuul master: Provide some documentation for the checks API implementation https://review.opendev.org/711493 | 14:59 |
| corvus | felixedel: ++ thanks and welcome back :) | 14:59 |
| felixedel | corvus: Thanks :) | 15:00 |
| *** felixedel has quit IRC | 15:01 | |
| y2kenny | corvus: the wildcards were removed before. Looking at the projects tab in the web-ui, there's only one pipeline and the two jobs that were defined for the specific branches and nothing else. | 15:02 |
| corvus | y2kenny: hrm. well, it's not *supposed* to report that in that case. it should only report on projects with branches in the pipeline. it's probably worth digging further to try to find out why it's doing that. | 15:05 |
| *** jcapitao is now known as jcapitao_afk | 15:06 | |
| y2kenny | corvus: ok, I will do a second check and double make sure there's no wildcard in any project. | 15:07 |
| fungi | the scheduler's debug logging will likely shed some light on what decisions led it to leave those reports | 15:07 |
| y2kenny | corvus: another issue may be my merger... I am not sure but I don't think I've got working merger outside of the executors at this point | 15:07 |
| tobiash | y2kenny: executors are fully functional mergers as well so standalone mergers are optional | 15:08 |
| y2kenny | tobiash: right... but I ran into bottleneck before that needs extra merger | 15:09 |
| tobiash | y2kenny: k, the mergers are exactly for this use case :) | 15:09 |
| y2kenny | but unfortunately, my merger is getting error like this: http://paste.openstack.org/show/793453/ | 15:10 |
| tobiash | y2kenny: hrm, looks like your merger vm/container lacks the ps command | 15:11 |
| *** ysandeep|afk is now known as ysandeep | 15:11 | |
| clarkb | I think the zuul published containers may suffer that. I ran into it when trying to signal things and common tools like that were missing | 15:12 |
| clarkb | I didn't think to check if that was giving zuul itself problems though | 15:12 |
| openstackgerrit | Merged zuul/zuul master: Replace preexec_fn argument in executor popen call https://review.opendev.org/727089 | 15:12 |
| tobiash | oh, right, I had to add procps as well when I switched our containers to the python base image | 15:12 |
| clarkb | (I was running docker exec) | 15:12 |
| y2kenny | tobiash: I am using docker.io/zuul/zuul-merger:latest | 15:12 |
| tobiash | y2kenny: that explains it | 15:13 |
| y2kenny | is that not ready for production use? | 15:13 |
| clarkb | y2kenny: we (opendev) are using them in production | 15:13 |
| clarkb | y2kenny: but I think we are still refining and fixing bugs as we find them | 15:13 |
| y2kenny | ok | 15:13 |
| clarkb | we need du, ssh-agent, ssh-add, and kubectl as well based on a quick skim of things | 15:15 |
| clarkb | (I think many/most/all of those may already be there but we can check easily enough) | 15:16 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Require procps also for dpkg platforms https://review.opendev.org/727223 | 15:16 |
| tobiash | clarkb, y2kenny: this should fix that ^ | 15:16 |
| clarkb | tobiash: ^ fyi I tried to pull a quick list of other fork+exec requirements | 15:16 |
| AJaeger | avass: did you see my comments on https://review.opendev.org/#/c/725678/ | 15:16 |
| y2kenny | tobiash: thanks | 15:16 |
| clarkb | tobiash: du is in coreutils which I expect is already in our images | 15:18 |
| clarkb | but maybe we should list it explicitly? | 15:18 |
| tobiash | makes sense, we never know how stripped down a newer python image will be | 15:19 |
| clarkb | tobiash: and https://packages.ubuntu.com/eoan/amd64/openssh-client/filelist is where we get the ssh-* utilities | 15:19 |
| clarkb | I think we should maybe list that too? | 15:19 |
| tobiash | openssh-client is already in bindep | 15:20 |
| clarkb | ah there it is thanks | 15:20 |
| avass | AJaeger: looking | 15:22 |
| AJaeger | avass: consider my comment a soft -0.5 ;) Mainly nits | 15:22 |
| avass | AJaeger: yeah, looked like it's mainly documentation, the undocumented varaibles are so we can flip the delegate_to and set a different working directory since the role is supposed to be used in a trusted context | 15:24 |
| *** jcapitao_afk is now known as jcapitao | 15:24 | |
| nils | fbo, hey, sorry for bothering. I'm still hitting dead-ends with running Zuul jobs in fedora-infra/ansible. Do you have some time? | 15:24 |
| avass | AJaeger: oh, and I need to remove the changes to fake-ansible | 15:24 |
| AJaeger | avass: oh.. | 15:25 |
| AJaeger | zuul-jobs maintainers, please review https://review.opendev.org/726900 to update the javascript jobs | 15:25 |
| avass | AJaeger: me and corvus came if with four different solutions to test it, and this seemed like a good balance between making it easy to use and test | 15:26 |
| AJaeger | 4? Wow | 15:28 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Add upload-artifactory role https://review.opendev.org/725678 | 15:31 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Split browser and non-browser targetting js jobs https://review.opendev.org/727231 | 15:33 |
| avass | AJaeger: done :) | 15:33 |
| mordred | AJaeger: since the js jobs are new - what do you think about doing that real quick ^^ | 15:33 |
| mordred | AJaeger: I think it would mean changing the template in ozj and the horizon patch to add -browser - but maybe it's soon enough that we can do that? | 15:33 |
| AJaeger | avass: Now I need to find another problem so that you can reach at least 50 patch sets ;) | 15:34 |
| mordred | AJaeger: (inspired by discussion about https://review.opendev.org/#/c/702132/) | 15:34 |
| fbo | nils: no problem, I haven't had the time to investigate more today the ansible plugins issue. But yes let me know what the others problems your are facing/ | 15:34 |
| AJaeger | mordred: reviewing | 15:34 |
| avass | AJaeger: oh but it's already at 50 patchsets :) | 15:35 |
| nils | fbo, mostly this one ;). I poked at it with a stick for a while and wanted to share my findings. | 15:36 |
| yoctozepto | mordred: re js functional - that would make sense; I don't think it's that standardized though; most projects keep themselves to unit tests | 15:36 |
| nils | fbo, just for testing I removed the *_plugins directories, then the Zuul jobs ran as I expected. | 15:36 |
| AJaeger | avass: you're iterating too fast for me ;) | 15:37 |
| openstackgerrit | Albin Vass proposed zuul/zuul master: Differentiate between failed and skipped builds https://review.opendev.org/727233 | 15:37 |
| nils | of course, that won't fly in real life :) | 15:37 |
| fbo | nils: ok I'm preparing a zuul instance and debug this live with a flork of fedora-infra/ansible | 15:37 |
| fbo | fork | 15:37 |
| avass | anyone who likes to work with js, wanna help me with that ^ :) | 15:37 |
| nils | fbo, unfortunately I don't even seem to find logs of the failed run, am I missing something | 15:38 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Split browser and non-browser targetting js jobs https://review.opendev.org/727231 | 15:38 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add firefox to javascript browser testing depends https://review.opendev.org/727234 | 15:38 |
| mordred | AJaeger: ^^ found one more place where it needs to have different logic | 15:38 |
| mordred | yoctozepto: ^^ | 15:39 |
| nils | fbo, and before I forget, Zuul seems to use the configuration from the next-to-last push into the PR, not the very latest one, see e.g. my comment here: https://pagure.io/fedora-infra/ansible/pull-request/54#comment-119348 | 15:39 |
| nils | fbo, is there anything I can help you with? | 15:40 |
| fbo | nils: this last error, zuul won't give more info. | 15:40 |
| mordred | AJaeger: oh - I'm dumb - wait a sec, this is much easier | 15:40 |
| AJaeger | mordred: why firefox? We've never used that before... | 15:40 |
| nils | fbo, ah so it's something where you have to access to Zuul, its logs, etc. | 15:40 |
| yoctozepto | mordred: thanks, commented | 15:41 |
| fbo | nils: yes that's why I need a bit of time to dig into it. | 15:41 |
| fbo | but I'm on it | 15:41 |
| yoctozepto | AJaeger: jslib used firefox | 15:41 |
| nils | fbo, I'll try spinning a Zuul instance in a VM (maybe my earlier problem was with podman-compose vs. docker-compose) | 15:41 |
| yoctozepto | AJaeger: most projects test only v8 in chromium but firefox's javascript engine can behave differently | 15:42 |
| yoctozepto | also due to browser sandboxing thingies | 15:42 |
| yoctozepto | I guess that was why it was tested | 15:42 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Split browser and non-browser targetting js test jobs https://review.opendev.org/727231 | 15:43 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add firefox to javascript browser testing depends https://review.opendev.org/727234 | 15:43 |
| fbo | nils: for zuul not using what you have in the PR, that's the first time I see it. Is it the only time you see that across all you PR updates ? | 15:43 |
| mordred | AJaeger: the js-openstack-lib installs both chromium and firefox for hysterical raisins - but it made me think that maybe that's a thing people might want? I split it into a second patch because maybe it's a bad idea | 15:44 |
| nils | fbo, I don't have much experience with Zuul in PRs yet, can't say | 15:44 |
| nils | and usually it would be from the perspective of someone getting a PR in that's not related to Zuul config in the project :) | 15:44 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Split browser and non-browser targetting js test jobs https://review.opendev.org/727231 | 15:45 |
| nils | fbo, I'm also not quite sure if all of this belongs here or rather in #fedora-ci -- i.e. what is more related to how we have deployed Zuul in Fedora rather than its code | 15:46 |
| mordred | yoctozepto: yes - I believe the new update to the job does the thing you suggest- and makes the whole thing much more simple | 15:47 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-output: introduce zuul_use_fetch_output https://review.opendev.org/681905 | 15:47 |
| mordred | yoctozepto: turns out we already had a playbook that only got called in test for installing that stuff :) | 15:47 |
| *** guillaumec has left #zuul | 15:47 | |
| AJaeger | mordred: https://review.opendev.org/727202 fails ;( | 15:58 |
| AJaeger | mordred: since it's only js-openstack-lib, let's keep it there | 15:58 |
| fbo | nils: we can discuss on fedora-ci and raise here issues if that make sense. | 15:59 |
| nils | fbo, +1 | 15:59 |
| mordred | AJaeger: kk | 16:00 |
| mordred | AJaeger: zuul_work_dir undefined - hrm | 16:01 |
| mordred | AJaeger: oh - you know - I think that gets fixed in https://review.opendev.org/#/c/726900 | 16:01 |
| AJaeger | mordred: I think as well. avass, clarkb, could you review 726900, please? | 16:01 |
| AJaeger | mordred: I'll add depends-on the browser change now | 16:02 |
| mordred | AJaeger: ++ - I thnik that's the best path forward | 16:02 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 16:03 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Support dynamic badges https://review.opendev.org/702128 | 16:04 |
| tristanC | corvus: i've updated https://review.opendev.org/#/c/681870/ and https://review.opendev.org/#/c/681905 . Please let me know if I missed something | 16:05 |
| clarkb | mordred: does the ansible stat in https://review.opendev.org/#/c/726900/7/roles/ensure-javascript-build-tool/tasks/main.yaml need a failed when false? I never know when ansible will treat a file not existing as an error or not | 16:05 |
| tristanC | (without those, sphinx or reno jobs are failing on kubectl host | 16:06 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Add upload-artifactory role https://review.opendev.org/725678 | 16:09 |
| mordred | clarkb: AJaeger has a horizon check patch up with a depends-on - and horizon doesn't have a yarn.lock | 16:10 |
| AJaeger | clarkb: stat will not fail | 16:10 |
| mordred | clarkb: https://review.opendev.org/#/c/727202/ | 16:10 |
| *** rpittau is now known as rpittau|afk | 16:10 | |
| AJaeger | clarkb: https://review.opendev.org/726940 was earlier | 16:11 |
| AJaeger | (it tested both old and new jobs) | 16:12 |
| clarkb | yup just ran a test locally | 16:12 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-output: introduce zuul_use_fetch_output https://review.opendev.org/681905 | 16:15 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 16:15 |
| openstackgerrit | Merged zuul/zuul master: Require procps also for dpkg platforms https://review.opendev.org/727223 | 16:21 |
| corvus | mordred: can you look at tristanC's change https://review.opendev.org/681870 ? i'm not sure how fetch-output is supposed to be used in this case. | 16:22 |
| mordred | corvus: why would there be a flag to incidate a site doesn't use fetch-output? | 16:24 |
| mordred | oh - this is to switch sphinx tarball to using fetch-output because it wasnt' before | 16:24 |
| corvus | mordred: my question is how are we supposed to use the docs/ dir specified by fetch-output | 16:25 |
| tristanC | corvus: `~/zuul-output/logs/docs` seems identical to `~/zuul-output/docs` | 16:25 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Remove support for ansible 2.6 https://review.opendev.org/727157 | 16:25 |
| *** sugaar has quit IRC | 16:26 | |
| mordred | well - the interface for docs is supposed to be ~/zuul-output/docs - so that documentation publishing jobs can publish output. but in check jobs we grab docs/ and put them into logs/docs/ for preview purposes | 16:27 |
| openstackgerrit | Merged zuul/zuul-jobs master: Extract ensure-javascript-build-tool role https://review.opendev.org/726900 | 16:28 |
| mordred | by having it in a different interface dir, doc publication jobs can say "I found content in docs/, I'm going to publish it to {{ zuul.project }}/docs" and that logic doesn't have to get copied around and can be in a trusted base job | 16:29 |
| mordred | BUT - we should loop in AJaeger here | 16:30 |
| AJaeger | mordred: the javascript changes pass on https://zuul.opendev.org/t/openstack/status/change/727202,3 | 16:30 |
| mordred | \o/ | 16:30 |
| mordred | AJaeger: piddle - I have a linter failure on the browser patch. fixing | 16:31 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Split browser and non-browser targetting js test jobs https://review.opendev.org/727231 | 16:31 |
| AJaeger | mordred, corvus , commented on 681870 - I think we're inconsistent in variable usage | 16:34 |
| *** evrardjp has quit IRC | 16:36 | |
| *** evrardjp has joined #zuul | 16:36 | |
| mordred | AJaeger, corvus: is it possible that the stuff I said above about a docs/ was an initial plan/design that we never did anything with? | 16:37 |
| openstackgerrit | Albin Vass proposed zuul/zuul master: Differentiate between failed and skipped builds https://review.opendev.org/727233 | 16:38 |
| tristanC | corvus: mordred: actually using `~/zuul-output/docs` is not identical to what fetch-sphinx-output does, docs/ result in `work/docs` while fetch-sphinx-output reulsts in `work/logs` | 16:38 |
| corvus | mordred: well, i don't think we've used fetch-output for docs yet, so i think this is the time to decide :) | 16:38 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-output: introduce zuul_use_fetch_output https://review.opendev.org/681905 | 16:38 |
| AJaeger | mordred: I don't remember, sorry. | 16:38 |
| mordred | tristanC: yeah- that's right - but the log publishing jobs are supposed to put the content of work/docs into work/logs/docks | 16:38 |
| mordred | so that doc-producing jobs can always just put the content into docs/ - and when it's appropriate for it to get published to logs, it does, and when it's appropriate to get finally published it does | 16:39 |
| tristanC | mordred: right now, https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/fetch-sphinx-output/tasks/main.yaml#L3 which is `zuul_executor_dest: "{{ zuul.executor.log_root }}"` | 16:40 |
| *** dpawlik has quit IRC | 16:40 | |
| mordred | merge-output-to-logs is the role that does the syncing | 16:41 |
| mordred | which you'll see is in the base job post playbook | 16:43 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Provide some documentation for the checks API implementation https://review.opendev.org/711493 | 16:43 |
| mordred | tristanC: you're right - but that's because it hadn't been migrated to the fetch-output interface yet | 16:44 |
| *** smyers has quit IRC | 16:44 | |
| mordred | if you have to go to work/docs - merge-output-to-logs should put it into work/logs/docs on the executor | 16:44 |
| AJaeger | mordred: if we merge your browser change, we should merge https://review.opendev.org/727271 as well | 16:45 |
| avass | any thoughts on this? https://review.opendev.org/#/c/727233/ we have a lot of skipped builds so it would be nice if we could easily differentiate between those and any failed builds | 16:45 |
| tristanC | mordred: ok, so is it required to use `merge-output-to-logs` when using `fetch-output` ? | 16:45 |
| mordred | well - it's the idea - it's expected for artifact processing too | 16:45 |
| tristanC | in our case, we only integrated `fetch-output`, not `merge-output-to-logs` | 16:45 |
| mordred | gotcha | 16:45 |
| tristanC | why `fetch-output` doesn't `merge-output-to-logs` automatically? | 16:46 |
| mordred | yeah- I was just about to say the same thig | 16:46 |
| tobiash | avass: I'll have a look when the dashboard build is done | 16:46 |
| mordred | I think maybe that was before we were ok with include_role | 16:46 |
| *** smyers has joined #zuul | 16:46 | |
| mordred | tristanC: perhaps we should just include_role merge-output-to-logs at the end of fetch-output ? | 16:46 |
| mordred | corvus: can you think of any reason we woudn't do that? | 16:47 |
| tristanC | mordred: sure, but should we also remove it from base post playbook to avoid double execution? | 16:47 |
| avass | tobiash: sure, the colors aren't that important, I just used the classes from patternfly atm | 16:47 |
| mordred | tristanC: yeah (although it should be idempotent, so it should be safe to add to fetch-output then remove from post everywhere | 16:48 |
| mordred | tristanC: I certainly agree that having 2 roles you're always supposed to use together is a little silly | 16:49 |
| tristanC | well as you prefer, i don't mind either way | 16:49 |
| mordred | especially as you need both for the interface to make sense | 16:49 |
| tristanC | but that seems like something we could do after the zuul_use_fetch_output addition to fetch-sphinx | 16:50 |
| mordred | yeah. and in fact, it could be a patch that adds the include_role and changes the dest dir - so that it's clear to read what's happening | 16:50 |
| corvus | mordred: i can't think of a reason not to do that right now :) | 16:50 |
| mordred | or we could just do the addition in this patch :) | 16:50 |
| mordred | then we reduce the number of patches | 16:51 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 16:54 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: add missing zuul_success default https://review.opendev.org/727272 | 16:54 |
| tristanC | corvus: shouldn't the merge of `merge-output-to-logs` and `fetch-output` be announced though? | 16:55 |
| AJaeger | mordred, zuul-maint, could we merge https://review.opendev.org/#/c/727231 and https://review.opendev.org/727271 ? That's one javascript job change that we can do before the jobs are used. | 16:56 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Remove support for ansible 2.6 https://review.opendev.org/727157 | 16:56 |
| corvus | tristanC: yes i think so | 16:58 |
| *** nils has quit IRC | 17:01 | |
| openstackgerrit | Albin Vass proposed zuul/zuul master: Enables whitelisting and configuring callbacks https://review.opendev.org/717260 | 17:02 |
| openstackgerrit | Merged zuul/zuul-jobs master: Split browser and non-browser targetting js test jobs https://review.opendev.org/727231 | 17:10 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-website master: Switch website to Gatsby https://review.opendev.org/717371 | 17:10 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-website master: Switch website to Gatsby https://review.opendev.org/717371 | 17:12 |
| mordred | AJaeger, corvus: related to the above convo ^^ is that the right thing to do in the promote pipeline there? | 17:13 |
| mordred | also - it makes me think build-javascript-deployment and build-javascript-deployment-tarball should be merged and they should just return both the preview and the tarball artifacts - sort of wasteful to do a second build just to build a tarball | 17:14 |
| *** jpena is now known as jpena|off | 17:14 | |
| *** ysandeep is now known as ysandeep|away | 17:19 | |
| *** jcapitao has quit IRC | 17:20 | |
| *** fbo is now known as fbo|off | 17:24 | |
| *** iurygregory has quit IRC | 17:27 | |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: WIP: Remove requiring tox_envlist https://review.opendev.org/726829 | 17:33 |
| *** yolanda has quit IRC | 17:35 | |
| openstackgerrit | Monty Taylor proposed zuul/zuul-website master: Switch website to Gatsby https://review.opendev.org/717371 | 17:36 |
| mordred | corvus, AJaeger: ok I think _that_ ^^ is maybe more like it | 17:38 |
| mordred | but I think we should merge build-javascript-deployment and build-javascript-deployment-tarball - and should just have build-javascript-deployment always build a tarball and return both artifacts | 17:40 |
| mordred | mnaser: ^^ also, I kinda took that patch over | 17:42 |
| AJaeger | mordred: I agree, we should have a single job instead of two | 17:43 |
| AJaeger | mordred: similar how we handle docs | 17:43 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 17:43 |
| mordred | AJaeger: ++ | 17:45 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-website master: Switch website to Gatsby https://review.opendev.org/717371 | 17:48 |
| mordred | AJaeger: cood call in readme | 17:48 |
| mordred | I updated a few more things too | 17:48 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Refactor github auth handling into its own class https://review.opendev.org/710034 | 17:50 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Remove support for ansible 2.6 https://review.opendev.org/727157 | 17:54 |
| AJaeger | avass has a new role to upload to artifactory, anybody else to +2A, please? https://review.opendev.org/#/c/725678/51 | 17:55 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Deprecate ansible 2.7 https://review.opendev.org/727344 | 17:57 |
| avass | AJaeger: oops, found an error with that note in the readme | 18:00 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Add upload-artifactory role https://review.opendev.org/725678 | 18:00 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Add upload-artifactory role https://review.opendev.org/725678 | 18:02 |
| avass | AJaeger: oh, and I just noticed the artifact section got duplicated :) | 18:02 |
| *** dustinc has joined #zuul | 18:05 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Default to Ansible 2.9 https://review.opendev.org/727345 | 18:07 |
| *** hashar has joined #zuul | 18:08 | |
| tobiash | avass: is there a reason to upload from the executor? When working with large artifacts it can be much more efficient to directly upload from the build node. | 18:09 |
| mordred | tobiash: maybe we should just go ahead and deprecate 2.8 too - since I think maybe it has the python3.8 issue too? and once we have 2.9 as a min we can start thinking about how collections fit into things | 18:10 |
| fungi | uploading from the build node can make it challenging to protect your upload credentials if arbitrary code is run on the node | 18:10 |
| mordred | fungi: yah - but might not be as much of a concern in a more trusted scenario | 18:10 |
| clarkb | what is the python3.8 issue? | 18:11 |
| tobiash | mordred: 2.8 seems to be ok with python 3.8 so far but I can prep that as well | 18:11 |
| clarkb | as a side note ianw discovered ansibles service module is broken with focal and newer fedora | 18:11 |
| tobiash | clarkb: ansible 2.7 and below doesn't work with python 3.8 | 18:11 |
| avass | tobiash: hmm, I think the idea was to be able to easier support windows and linux environments that way. | 18:12 |
| tobiash | more specifically it fails to do loops | 18:12 |
| fungi | and yeah, sounds like the fixes for the service module are merged but not yet released | 18:12 |
| avass | tobiash: otherwise, yes you're right | 18:12 |
| y2kenny | corvus, fungi: so I have the summaried scheduler log here (specifically for change# 358904): http://paste.openstack.org/show/793467/ The three "Dequeuing change" with "because it can no longer merge" shows up in Gerrit as "This change depends on a change that failed to merge." I am not sure if that is valid and I am not sure why there are three | 18:12 |
| y2kenny | of them. Looks like each of them is a unique event though... | 18:12 |
| tobiash | avass: hrm, maybe we can enhance that later | 18:13 |
| fungi | y2kenny: note that the dequeue messages are followed with "adding change ... to queue" | 18:15 |
| avass | tobiash: there's also no src parameter in win_uri | 18:15 |
| fungi | y2kenny: so something seems to have triggered readding them, could it be that the osg pipeline is configured to react to comments and so the report comment is causing reenqueuing? | 18:15 |
| y2kenny | fungi: oh right... um... | 18:15 |
| fungi | y2kenny: also in the debug logs for the scheduler there should be entries for the triggering events themselves, indicating what matched the pipeline's conditions | 18:16 |
| avass | tobiash: maybe, instead of toggling the delegate_to: during testing, we could actually allow that? | 18:16 |
| y2kenny | fungi: yes, many of my pipelines are configured to react to comment for things like !recheck | 18:17 |
| tobiash | avass: I think that might make sense | 18:17 |
| y2kenny | fungi: so the multiple copies of the failed to merge is a form of "gate loop" I have encountered previously | 18:19 |
| fungi | y2kenny: you're filtering comment-added events for that pipeline with a comment pattern i guess, something like https://opendev.org/opendev/project-config/src/branch/master/zuul.d/pipelines.yaml#L19 | 18:19 |
| y2kenny | fungi: and it essentially retrying until the merge become successful | 18:19 |
| y2kenny | fungi: yes that's it | 18:20 |
| fungi | y2kenny: is it possible the pattern is too greedy and winds up matching the report comment from zuul's account somehow? | 18:20 |
| tobiash | clarkb: with python 3.8 ansible 2.7 fails when doing this: https://review.opendev.org/#/c/727155/3/tests/fixtures/config/ansible/git/common-config/playbooks/hello-post.yaml | 18:21 |
| tobiash | so basically it's completely broken | 18:21 |
| tobiash | there is also an upstream issue that got rejected because 2.7 is in security fix only maintenance mode | 18:22 |
| y2kenny | fungi: I basically use the same regex but switched out the "recheck" with !<pipeline name> (so !reviewed) but may be ! mean something special I am not aware of in regex. | 18:23 |
| y2kenny | fungi: looking at the full log, I don't think I ran the scheduler in debug mode so I don't see the trigger | 18:25 |
| y2kenny | fungi: I do see log from GerritConnection on updating change | 18:25 |
| fungi | y2kenny: it's using the re module in python, so you can test the patterns yourself like this: | 18:26 |
| fungi | python3 | 18:26 |
| fungi | import re | 18:26 |
| fungi | re.match(r"""(?i)^(Patch Set [0-9]+:)?( [\w\\+-]*)*(\n\n)?\s*recheck""", "Patch Set 123: recheck") | 18:26 |
| fungi | y2kenny: https://docs.python.org/3/library/re.html indicates ! can be special in certain negative lookahead expressions, so you may need to use \! in the pattern to escape it | 18:30 |
| *** iurygregory has joined #zuul | 18:31 | |
| fungi | (also used in negative lookbehinds) | 18:32 |
| y2kenny | fungi: I ran the pattern in python3 just now and it seems fine (did not match the failed to merge comment) | 18:33 |
| fungi | interesting. well, if you can manage to reproduce under debug logging on the scheduler that may provide some additional clues | 18:34 |
| y2kenny | fungi: ok... I will keep observing | 18:35 |
| mordred | fungi, clarkb: re service module - they _just_ made some point releases this morning - so maybe that's been fixed since ianw was awake | 18:37 |
| mordred | tobiash: nod. so maybe it's still too early to retire 2.8 - but I definitely agree about 2.7 at this point | 18:37 |
| tobiash | mordred: so what do we do about py38 since the official images are basically broken with ansible 2.7? | 18:38 |
| mordred | tobiash: it's a good questino. we could switch the images back to py37 (I don't think we know of any issues with 3.7) until we drop 2.7 | 18:39 |
| tobiash | py37 should work I guess | 18:40 |
| mordred | and then we could re-raise to 3.8 in the patch to drop ansible 2.7 | 18:40 |
| clarkb | and keep testing py38 in the gate to ensure we are ready for smooth transition there | 18:41 |
| clarkb | its somewhat surprising they found a way to make invalid 3.8 but va.id 3.7 | 18:42 |
| clarkb | usually its the other way around | 18:42 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Switch back to python 3.7 https://review.opendev.org/727367 | 18:43 |
| y2kenny | fungi: so I am looking at the full log again and I noticed this. If I have a patch series A->B->C. If A "did not merge because it did not have any jobs configured", would it cause B to fail and lead to a "failed to merge" message? | 18:43 |
| fungi | y2kenny: yes, if you have changes dependent on changes which zuul thinks cannot merge, then the depending changes also cannot merge from zuul's perspective | 18:45 |
| fungi | and it will report that on them, as an explanation for why it did not attempt to run any jobs for them | 18:45 |
| y2kenny | fungi: even because the no merge was due to no job? | 18:46 |
| fungi | y2kenny: for that case, zuul is likely determining that before it gets to the point of realizing there would have been no jobs to run anyway. that's likely something we can improve in the scheduler | 18:47 |
| fungi | but i'm not sure how involved that switch in logic could be | 18:48 |
| corvus | y2kenny: aha -- i think that makes sense -- let me sammarise to make sure i understand: your zuul is not reporting anything on change A, but it's reporting the failed to merge message on B and C? | 18:48 |
| y2kenny | fungi: I think this is probably what's happening in my case. Surveying these failed to merge gerrit changes, the first patchset always seems to be fine | 18:48 |
| *** avass has quit IRC | 18:48 | |
| y2kenny | it's the subsequent patchset that has the failed to merge message pops up. (and in some what random repetitions) | 18:49 |
| y2kenny | corvus: yes... and report multiple failed to merge messages on B and C | 18:49 |
| y2kenny | corvus: so far I have not detected a pattern on the number of repeats for the subsequence change | 18:50 |
| fungi | it's possible the multiplicity of those messages could be disconnected from the reason for them (but they also might be related) | 18:50 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Combine javascript deployment and deployment-tarball jobs https://review.opendev.org/727370 | 18:51 |
| mordred | AJaeger: ^^ combined | 18:51 |
| corvus | y2kenny, fungi: i agree reporting on B and C is a bug in zuul; i think we can fix it. | 18:51 |
| *** bhavikdbavishi has quit IRC | 18:51 | |
| y2kenny | corvus, fungi: I can paste a fuller log for analysis if you like (I truncated it to a specific change before because there's just a lot of log) | 18:51 |
| corvus | y2kenny: i think i have enough to go on for now, thanks :) | 18:52 |
| y2kenny | corvus: ok. | 18:52 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Deprecate ansible 2.7 https://review.opendev.org/727344 | 18:54 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Drop support for ansible 2.6 https://review.opendev.org/727157 | 18:54 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Drop support for ansible 2.7 https://review.opendev.org/727373 | 18:54 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Update images to use python 3.8 https://review.opendev.org/727374 | 18:54 |
| mordred | tobiash: ++ | 18:55 |
| tobiash | now we need to agree on a timeline/release plan :) | 18:55 |
| tobiash | I think we could immediately deprecate 2.7 and drop 2.6 | 18:56 |
| tobiash | dropping 2.7 and upping python should wait one release | 18:56 |
| y2kenny | corvus: and just a bit of additional data point... I just sampled another 5 series that has similar issue and none of the first patch of those series has the message. Only the subsequent patches.) | 18:57 |
| *** guillaumec has joined #zuul | 18:57 | |
| *** sgw has quit IRC | 19:05 | |
| *** sgw has joined #zuul | 19:07 | |
| mnaser | mordred: please feel free to take it over :) | 19:10 |
| mordred | mnaser: done! | 19:11 |
| mnaser | mordred: i see you've already broken it so that's perfect :) | 19:12 |
| mnaser | error gatsby@2.21.22: The engine "node" is incompatible with this module. Expected version ">=10.13.0". Got "6.14.1" | 19:12 |
| mordred | mnaser: I also split blog into its own thing | 19:12 |
| mordred | mnaser: yay! that's now what should have happend | 19:12 |
| mnaser | well i'm glad its broken the way it's supposed to :P | 19:12 |
| AJaeger | mordred: do we need a promote job as well? See https://opendev.org/opendev/base-jobs/src/branch/master/zuul.d/jobs.yaml#L472 | 19:13 |
| mordred | mnaser: s/now/not/ | 19:15 |
| mordred | but also - why is it getting 6 instead of 14 :( | 19:15 |
| mnaser | mordred: TASK [ensure-nodejs : Output node version] ? ubuntu-bionic | v6.14.1 | 19:16 |
| mnaser | mordred: did we bump the upstream version? i dont see it here https://review.opendev.org/#/c/717371/19/.zuul.yaml | 19:16 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Combine javascript deployment and deployment-tarball jobs https://review.opendev.org/727370 | 19:16 |
| mordred | mnaser: it should be defaulting to 14 in the shiny new jobs | 19:17 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-website master: Switch website to Gatsby https://review.opendev.org/717371 | 19:18 |
| mordred | mnaser: maybe I'm misunderstanding how role defaults and include_role work - so let's just set a job variable :) | 19:18 |
| mordred | AJaeger: I think that promote job should work yes? | 19:18 |
| mordred | AJaeger: oh - that's just publishing the tarball | 19:19 |
| AJaeger | mordred: it won't work in https://review.opendev.org/#/c/726554/ since the artifact name is different | 19:19 |
| AJaeger | download_artifact_job: build-javascript-content-tarball | 19:20 |
| mordred | AJaeger: nod. also - we want to publish the contents here, not the tarball anyway | 19:25 |
| mordred | AJaeger: so yeah, I might need a promote job for this | 19:25 |
| *** avass has joined #zuul | 19:47 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 19:50 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Drop support for ansible 2.6 https://review.opendev.org/727158 | 19:51 |
| clarkb | mordred: the OSF is asking if we have any content for the newsletter. The biggest thing that stands out to me is the multi arch docker images builds. Are we ready to advertise that functionality? | 19:51 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Drop support for ansible 2.7 https://review.opendev.org/727410 | 19:51 |
| tobiash | clarkb, mordred: did we have the initial checks api support for github already in the last one? | 19:52 |
| clarkb | tobiash: I want to say that got mentioned but maybe it only made it onto the community update stuff | 19:53 |
| clarkb | tobiash: we can add that too if it wasn't previously mentioned. Thanks for the reminder | 19:53 |
| mordred | clarkb: sadly no - we've found a few new unknown unknowns | 19:54 |
| mordred | clarkb: I mean - we could advertise it - it's likely by the time someone notices for real we'll have it sorted | 19:54 |
| clarkb | mordred: or maybe even a "Initial support is there pelase help us test it" thing? | 19:54 |
| mordred | yeah | 19:55 |
| openstackgerrit | Gonéri Le Bouder proposed zuul/zuul-jobs master: bindep: ensure virtualenv is present https://review.opendev.org/727413 | 19:55 |
| clarkb | mordred: ok I'll go with that then | 19:55 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 19:55 |
| tristanC | mordred: corvus: AJaeger: please let me know if I missed anything with https://review.opendev.org/#/c/681870 and https://review.opendev.org/#/c/681905/6 (re: building docs on kubectl node) | 19:56 |
| clarkb | ianw: I didn't approve https://review.opendev.org/#/c/726032/1 because i was hoping corvus might be able to double check my assertion in the comments there first | 19:59 |
| clarkb | but I think it is ready otherwise | 19:59 |
| *** y2kenny has quit IRC | 20:01 | |
| mnaser | btw -- i assume the 3pci hasn't been too noisy and not providing a bunch of bad output? | 20:01 |
| corvus | clarkb: +3 | 20:04 |
| mordred | mnaser: it's been great! | 20:05 |
| mnaser | sounds good | 20:07 |
| AJaeger | mnaser: thanks, it's appreciated! | 20:09 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Don't require tox_envlist https://review.opendev.org/726829 | 20:13 |
| openstackgerrit | Albin Vass proposed zuul/zuul-jobs master: Don't require tox_envlist https://review.opendev.org/726829 | 20:14 |
| *** dpawlik has joined #zuul | 20:17 | |
| *** sshnaidm is now known as sshnaidm|afk | 20:18 | |
| avass | can we try to deprecate having 'venv' as default tox_envlist? I'm guessing that might be a bit of work but I'm willig to do it if there's no problem with that :) | 20:21 |
| corvus | avass: wasn't there a change about that? | 20:21 |
| avass | corvus: not sure, I pushed a change to see what breaks in the tox role if we remove it | 20:22 |
| avass | corvus: but I at least think that we should work towards tox having the same behaviour as tox :) | 20:23 |
| corvus | avass: wow, i thought i remembered a change like https://review.opendev.org/726830 but older than yesterday :) | 20:26 |
| corvus | avass: i guess yesterday just seem like a long time ago :) | 20:27 |
| avass | corvus: I think I updated it to handle an empty string, maybe that's what you're thinking about? :) | 20:28 |
| corvus | avass: ah that may be it | 20:28 |
| avass | corvus: noticed, that and empty string probably breaks tox_sibling so I'm trying to fix that atm | 20:28 |
| avass | wow that sentence | 20:29 |
| corvus | i got it :) | 20:29 |
| avass | :) | 20:29 |
| avass | corvus: anyway, I guess that it's enough to create a tox-venv job and make any job that inherits from the tox job that doesn't tox_envlist inherit from that instead | 20:32 |
| tobiash | mordred, clarkb: it would be great if I could get a second review on https://review.opendev.org/720458 which makes repo reset more robust | 20:33 |
| corvus | avass: yeah, left a comment; it seems worthwhile to me. i don't know if we need a tox-venv job, or just need to update some openstack jobs to set tox_envlist:venv | 20:33 |
| mordred | yeah - I mean - honestly just running "tox -e venv" in a zuul job isn't likeluy to be super useful | 20:33 |
| avass | corvus: yeah, one of those | 20:34 |
| mordred | corvus: yah | 20:34 |
| mordred | I think we shoudl do that | 20:34 |
| corvus | avass: openstack is releasing this week, so probably shouldn't merge it before then, but maybe we can research this week and give it a try next week | 20:35 |
| avass | corvus: sure, there's no hurry anyway. | 20:35 |
| mordred | corvus: http://codesearch.openstack.org/?q=tox_extra_args&i=nope&files=&repos= | 20:36 |
| *** cdearborn has joined #zuul | 20:36 | |
| avass | corvus: on the other hand. what do I need to do to get this through? https://review.opendev.org/#/c/724855/ :) | 20:36 |
| mordred | corvus: I have found very few - if _any_ - openstack jobs actually using the venv default value | 20:36 |
| avass | corvus: We're keeping a mirror of zuul-jobs at the moment to be able to do quickfixes but I'm working towards using zuul-jobs directly | 20:38 |
| mordred | https://opendev.org/openstack/horizon/src/branch/master/playbooks/horizon-tox-django/pre.yaml#L9 <-- I think that's the only occurance | 20:38 |
| avass | corvus: Third party ci would have been nice but I don't think we have resources to do that at the moment | 20:38 |
| mordred | although I agree, we should wait until after openstack release | 20:38 |
| corvus | avass: i think we can probably make that change next week too? | 20:38 |
| avass | corvus: sure, I'll try to get it ready | 20:39 |
| corvus | avass: it should be a noop, so we shouldn't need to announce it or anything. you can test several of those with Depends-On from the zuul repo (at least the js, sphinx, subunit ones) | 20:39 |
| corvus | avass: let's get an ack (maybe just some +1s) from some people on the policy change. tristanC, tobiash, clarkb would be good reviewers -- can you confirm the policy goal in https://review.opendev.org/724855 is good? | 20:41 |
| avass | corvus: I think I just need to work out some quirks to make sure the linter only catches any file transfers between executor and remote | 20:42 |
| corvus | avass: let's get an ack (maybe just some +1s) from some people on the policy change. tristanC, tobiash, clarkb would be good reviewers -- can you confirm the policy goal in https://review.opendev.org/724855 is good? | 20:43 |
| corvus | sorry for repeat msg :( | 20:43 |
| clarkb | tobiash: corvus re https://review.opendev.org/720458 I think I'm ok with the goal there, but it seems weird to intentionally set HEAD to an arbitrary value | 20:43 |
| *** sanjayu_ has quit IRC | 20:43 | |
| tristanC | corvus: those ansible-lint rules are looking great to me, thanks! | 20:43 |
| tristanC | avass: ^ | 20:44 |
| clarkb | but I guess if the remote HEAD isn't valid then we can't realy rely on that so any valid head is what we are doing | 20:44 |
| avass | I'm logging off for today, I'll check in tomorrow :) | 20:44 |
| clarkb | corvus: avass the policy goal there makes sense to me | 20:44 |
| openstackgerrit | Merged zuul/nodepool master: Use fqdn for builder hostname info https://review.opendev.org/726032 | 20:45 |
| tobiash | clarkb: any valid HEAD works because since we delete and re-create all refs we just need to detach HEAD from any local branch | 20:45 |
| corvus | tobiash, mordred: i +3d the ansible stack except for the last one; i think we may want to wip https://review.opendev.org/727373 for a bit? | 20:45 |
| clarkb | tobiash: but we don't recreate HEAD | 20:46 |
| tobiash | corvus: yes, that's the intention, I'll -2 it until the next release | 20:46 |
| clarkb | tobiash: I think we set HEAD later when we set it up to the zuul ref ? | 20:46 |
| clarkb | basically this is an intermediate state and we should update HEAD later for the actual zuul refs when the jobs run right? | 20:47 |
| tobiash | clarkb: repo reset is not intended to set HEAD to anything special. Zuul will take care of this later when e.g. fetching or merging | 20:47 |
| mordred | corvus, avass: I checked - that one horizon entry I found specifies an envlist of venv explicitly in a job def - so I think we're good to go | 20:47 |
| tobiash | clarkb: so yes I guess we could call it an intermediate state effectively | 20:48 |
| clarkb | tobiash: thanks for confirming that was what I thought | 20:48 |
| clarkb | I'ev approved it | 20:49 |
| *** avass has quit IRC | 20:49 | |
| tobiash | clarkb: thanks! | 20:50 |
| tobiash | maybe a tiny additional one (removes dead code): https://review.opendev.org/718670 | 20:51 |
| openstackgerrit | Merged zuul/nodepool master: Allow disabling build-log-retention https://review.opendev.org/723782 | 20:59 |
| *** sgw has quit IRC | 20:59 | |
| *** hashar has quit IRC | 21:01 | |
| *** rfolco|rover is now known as rfolco|rover|off | 21:06 | |
| openstackgerrit | Merged zuul/nodepool master: Functional container tests: update to CentOS 8 https://review.opendev.org/721509 | 21:09 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 21:12 |
| openstackgerrit | Merged zuul/zuul master: Allow check runs to be configured as required status in pipeline config https://review.opendev.org/711241 | 21:15 |
| *** sgw has joined #zuul | 21:18 | |
| openstackgerrit | melanie witt proposed zuul/zuul-jobs master: DNM Try out running sphinx-build in parallel for releasenotes https://review.opendev.org/727473 | 21:49 |
| *** panda has quit IRC | 21:51 | |
| *** panda has joined #zuul | 21:53 | |
| openstackgerrit | Guillaume Chauvel proposed zuul/zuul-jobs master: tox siblings installed packages: Add PEP 440 direct reference format https://review.opendev.org/727475 | 21:54 |
| *** zxiiro has quit IRC | 22:19 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul-jobs master: fetch-sphinx-tarball: introduce zuul_use_fetch_output https://review.opendev.org/681870 | 22:32 |
| mordred | corvus: so - I've got a weird ansible thing I don't understand | 22:36 |
| mordred | corvus: https://review.opendev.org/#/c/727370/1 (patchset 1 is important) - https://review.opendev.org/#/c/717371/19/ (patchset 19 is important) | 22:37 |
| mordred | if you look at the build log from 717371,19 : https://zuul.opendev.org/t/zuul/build/75f0b66f8a1041b48e159802e8228248 | 22:38 |
| mordred | corvus: you'll see it's installing node version 6 - even though the roles in question have '14' in their defaults | 22:38 |
| mordred | when I put node_version: 14 in the zuul job variables it works | 22:38 |
| mordred | I'm wondering if there's just something dumb I'm missing - or if maybe a role with a default value that include_role: another role for some reason doesn't pass that value along? | 22:39 |
| mordred | clarkb: ^^ you like weird things | 22:39 |
| openstackgerrit | Merged zuul/zuul master: Detach head before reset https://review.opendev.org/720458 | 22:40 |
| *** guillaumec has quit IRC | 22:44 | |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Combine javascript deployment and deployment-tarball jobs https://review.opendev.org/727370 | 22:45 |
| clarkb | mordred: include_role is weird ime | 22:46 |
| clarkb | it wouldn't surprise me if it was related | 22:46 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Combine javascript deployment and deployment-tarball jobs https://review.opendev.org/727370 | 22:46 |
| mordred | clarkb: I'm trying passing it explicitly | 22:47 |
| mordred | clarkb: and will see if that works | 22:47 |
| openstackgerrit | Merged zuul/zuul master: Remove unused function prune in merger https://review.opendev.org/718670 | 22:50 |
| *** sanjayu_ has joined #zuul | 23:02 | |
| openstackgerrit | melanie witt proposed zuul/zuul-jobs master: DNM Try out running sphinx-build in parallel for releasenotes https://review.opendev.org/727473 | 23:14 |
| *** tosky has quit IRC | 23:14 | |
| *** sanjayu_ has quit IRC | 23:19 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!