[05:48:21] (03Restored) 10Hashar: Jenkins job validation (DO NOT SUBMIT) [integration/quibble] - 10https://gerrit.wikimedia.org/r/663557 (owner: 10Hashar) [05:48:32] (03CR) 10Hashar: "check experimental" [integration/quibble] - 10https://gerrit.wikimedia.org/r/663557 (owner: 10Hashar) [11:01:18] 10Quibble, 10Continuous-Integration-Config, 10Patch-For-Review: Switch all Quibble Selenium jobs to use apache - https://phabricator.wikimedia.org/T285649 (10hashar) [11:03:34] 10Quibble, 10Continuous-Integration-Config, 10Patch-For-Review: Switch all Quibble Selenium jobs to use apache - https://phabricator.wikimedia.org/T285649 (10hashar) 05Open→03Resolved There are a few left over tasks for tests that fail due to the concurrent requests, they got worked around or marked to b... [11:38:48] 10Quibble, 10Platform Engineering, 10Continuous-Integration-Config, 10MW-1.38-notes (1.38.0-wmf.17; 2022-01-10), 10Platform Team Initiatives (API Integration Tests): "should mark the revert as a bot edit" test flaky with Apache + php-fpm - https://phabricator.wikimedia.org/T298514 (10kostajh) Tagging #pl... [14:18:07] (03Abandoned) 10Hashar: Jenkins job validation (DO NOT SUBMIT) [integration/quibble] - 10https://gerrit.wikimedia.org/r/663557 (owner: 10Hashar) [14:29:15] (03CR) 10Hashar: [C: 03+2] "Sounds good :]" [integration/quibble] - 10https://gerrit.wikimedia.org/r/738061 (https://phabricator.wikimedia.org/T226869) (owner: 10Kosta Harlan) [14:31:53] 10Quibble, 10Platform Engineering, 10Continuous-Integration-Config, 10MW-1.38-notes (1.38.0-wmf.17; 2022-01-10), and 2 others: "should mark the revert as a bot edit" test flaky with Apache + php-fpm - https://phabricator.wikimedia.org/T298514 (10hashar) I have send https://gerrit.wikimedia.org/r/c/mediawik... [14:44:16] (03Merged) 10jenkins-bot: BrowserTests: Option to run npm install in parallel [integration/quibble] - 10https://gerrit.wikimedia.org/r/738061 (https://phabricator.wikimedia.org/T226869) (owner: 10Kosta Harlan) [14:54:56] (03CR) 10Hashar: [C: 03+2] "I originally thought we could find something easier that does not have to involve the low level os.dup or os.dup2, but that one seems to w" [integration/quibble] - 10https://gerrit.wikimedia.org/r/693456 (owner: 10Awight) [14:56:51] * awight leans forward [14:56:59] You got my attention, the recent activity is exciting! [15:02:26] (03CR) 10Hashar: "quibble.backend._stream_relay seems to achieve the same behavior?" [integration/quibble] - 10https://gerrit.wikimedia.org/r/693457 (owner: 10Awight) [15:06:54] (03CR) 10Awight: Expand redirection to capture logging (031 comment) [integration/quibble] - 10https://gerrit.wikimedia.org/r/693457 (owner: 10Awight) [15:11:34] (03Merged) 10jenkins-bot: Introduce utility to redirect stdout and stderr [integration/quibble] - 10https://gerrit.wikimedia.org/r/693456 (owner: 10Awight) [15:11:47] the progress monitor I am a bit confused about the coupling between the two :D [15:11:57] I guess that will be for tomorrow [15:14:05] Which coupling? [15:14:18] (03CR) 10Hashar: [C: 03+1] "Sorry I have misread the intent. Looks like the patch needs to be rebased :/" [integration/quibble] - 10https://gerrit.wikimedia.org/r/693457 (owner: 10Awight) [15:14:26] awight: that one needs a rebase ^ [15:14:40] Mostly, the progress monitor exists in order to output each job's results atomically [15:15:05] yeah I gotta try it out [15:15:38] (03PS8) 10Awight: Expand redirection to capture logging [integration/quibble] - 10https://gerrit.wikimedia.org/r/693457 [15:15:44] Rebased :-) [15:16:21] (03CR) 10Hashar: [C: 03+2] Expand redirection to capture logging [integration/quibble] - 10https://gerrit.wikimedia.org/r/693457 (owner: 10Awight) [15:16:28] progress [15:17:08] and there is an atexit which I always find suspicious :D [15:17:55] bwhahaha [15:32:24] (03PS2) 10Kosta Harlan: BrowserTests: Support splitting projects into two groups [integration/quibble] - 10https://gerrit.wikimedia.org/r/747904 [15:32:28] (03Merged) 10jenkins-bot: Expand redirection to capture logging [integration/quibble] - 10https://gerrit.wikimedia.org/r/693457 (owner: 10Awight) [15:34:36] (03CR) 10Kosta Harlan: "This would also be nice to include with Quibble 1.3.0; it's safe in that it's behind a command flag, so if it doesn't work out, it's a jjb" [integration/quibble] - 10https://gerrit.wikimedia.org/r/747904 (owner: 10Kosta Harlan) [15:54:39] kostajh: now that we have Apache, I am really tempted to have webdriver.io to run tests in parallel ;) [15:55:49] hashar: +1 [15:57:05] This is enabled using a configuration setting in wdio.conf.js, so we can go one repo at a time... [15:57:35] I think it'll be quite disruptive in some repos, for example when tests reuse a fixture page title. [15:57:50] hashar: sure, but I suspect it won’t be so straightforward [15:59:33] `maxInstances` [16:02:00] yeah crafting a patch for that [16:02:04] woot! [16:02:26] I am looking at wdio --help as well [16:02:38] fwiw https://webdriver.io/docs/configurationfile/ [16:03:07] ah I see what you mean, maybe we can sneak a flag in... but I think per-repo is nice because it lets us make exceptions... [16:09:19] a fun alternative would be to have a fleet of firefox on k8s or on some baremetal [16:09:27] and delegate the execution to that fleet [16:09:32] s/firefox/chromium/ [16:09:42] https://gerrit.wikimedia.org/r/c/mediawiki/core/+/751761 selenium: run 4 tests in parallel [16:11:11] 10Quibble, 10MediaWiki-Core-Tests, 10Browser-Tests, 10Patch-For-Review: Run browser tests in parallel - https://phabricator.wikimedia.org/T226869 (10hashar) [16:11:17] 10Quibble, 10Continuous-Integration-Config, 10Patch-For-Review: Switch all Quibble Selenium jobs to use apache - https://phabricator.wikimedia.org/T285649 (10hashar) [16:11:26] I don't know how to write bah [16:16:38] O_O [16:17:57] strange that the API tests failed [16:18:20] hmm https://integration.wikimedia.org/ci/job/wmf-quibble-selenium-php72-docker/129055/console#console-section-10 [16:18:43] I cant tell whether stuff runs in // [16:19:26] off-topic, I realized that the A+B selenium patch is still useful even with maxInstances, because wdio is only parallelizing within each suite. [16:20:31] hashar: Take a look at the output for "Execution of 5 spec files started at 2022-01-05T16:16:00.746Z" [16:20:39] That looks very // to me [16:20:44] looks [16:20:50] yeah based on the timing [16:20:56] it started 4 in parallel [16:22:05] I don't know what are the [0-0] [0-1] etc [16:22:20] I am guessing it is [ - ] [16:25:45] if we run npm install in parallel [16:26:01] and get at least wikibase and minvervaneue to run tests in parallel, we should have a nice boost [16:31:26] +1 and I'm pretty sure we can tie up the ParallelCommand thing which could put the browser tests parallel to API tests, etc. [16:32:07] I also have a patch somewhere to extract some of the wikibase tests to a standalone job [16:32:23] so they would only be triggered by patches to Wikibase or core [16:32:30] but well they fail and I haven't looked back ;) [16:41:11] 10Quibble, 10MediaWiki-Core-Tests, 10Browser-Tests, 10Patch-For-Review: Run browser tests in parallel - https://phabricator.wikimedia.org/T226869 (10hashar) As of this morning, CI exposes MediaWiki through Apache which allow concurrent requests (T285649) we should thus be able to get webdriver.io to run t... [16:43:16] hehehe [16:46:25] looks like mw-core tests finished instead of 10 seconds instead of the usual 26 seconds [16:46:37] https://integration.wikimedia.org/ci/job/wmf-quibble-selenium-php72-docker/129059/console#console-section-7 [16:50:16] and Minerva at 26 seconds instead of 43, though part of that may be Apache https://integration.wikimedia.org/ci/job/wmf-quibble-selenium-php72-docker/129061/console#console-section-9 [16:53:04] and yeah, awight I think the A+B (or A+B+C split) of browser tests would be helpful, because I think ParallelCommand is going to be tricky to implement [16:53:59] fwiw I'm hoping to do a similar A+B split of tests for PHPUnit, for the same reason that parallelization in the same environment is going to be hard for various reasons [16:58:41] (03PS3) 10Kosta Harlan: BrowserTests: Support splitting projects into two groups [integration/quibble] - 10https://gerrit.wikimedia.org/r/747904 (https://phabricator.wikimedia.org/T226869) [16:58:42] I am not so fan of the split [16:59:15] cause that means more jobs running in parallel [16:59:26] each doing git operations + composer/npm install etc [17:00:28] also npm ci is super slow [17:06:06] and i found https://github.com/pnpm/pnpm#readme :D [17:10:40] it install node modules in the home directory [17:11:09] then just uses a bunch of symlink in the project that link back to the home directory storage [17:11:20] that saves disks and a lot of write i/o [17:29:30] presumably we can get more resources from WCMS to have more jobs running in parallel? That is cheaper than rewriting testsuites. I mean, rewriting is better longer term but I’d rather have something that works now [17:30:05] yeah possibly [17:30:15] meanwhile Wikibase with maxInstances at 4 fails miserably https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/545650/ [17:30:48] and the wmf-quibble-apache jobs are still triggerd somehow [17:32:06] I forgot to deploy the zuul change https://gerrit.wikimedia.org/r/c/integration/config/+/751697 [17:32:07] done [17:34:23] I am off [17:59:32] 10Quibble, 10MediaWiki-Core-Tests, 10Browser-Tests, 10Patch-For-Review, 10User-zeljkofilipin: Run browser tests in parallel - https://phabricator.wikimedia.org/T226869 (10zeljkofilipin) [18:56:12] 10Quibble, 10MediaWiki-Core-Tests, 10Browser-Tests, 10Patch-For-Review, 10User-zeljkofilipin: Run browser tests in parallel - https://phabricator.wikimedia.org/T226869 (10Krinkle) @kostajh If I understand correctly, this last patch would split at the job/quibble level, and thus not share a common db or w... [19:30:21] 10Quibble, 10MediaWiki-Core-Tests, 10Browser-Tests, 10MW-1.38-notes (1.38.0-wmf.17; 2022-01-10), and 2 others: Run browser tests in parallel - https://phabricator.wikimedia.org/T226869 (10kostajh) >>! In T226869#7599620, @Krinkle wrote: > @kostajh If I understand correctly, this last patch would split at t... [21:32:12] (03CR) 10Hashar: [C: 04-1] "I have no idea what excimer is but if we need a lot of configuration I highly advise to move it to mediawiki/core includes/DevelopmentSett" [integration/quibble] - 10https://gerrit.wikimedia.org/r/748314 (https://phabricator.wikimedia.org/T225730) (owner: 10Ladsgroup)