[16:17:52] I’m wondering - when WMF JobRunner is called - is it called by shell or by web/curl request, and if with web - does it get the `X-Request-ID` header? [16:21:55] Sure pmiazga! Krinkle, I want to +2 https://gerrit.wikimedia.org/r/c/mediawiki/core/+/947374, any feedback or objection from your end? [16:25:32] xSavitar: LGTM. We broke the override method so if we update EventBus, as long as the deprecated method exists that's still a regression. [16:25:49] I believe the affected patch hasn't gone to prod yet, so no backport needed for this one. [16:26:11] Thanks! [16:26:57] * xSavitar waits for CI to pass before I plus two [16:28:32] Krinkle, yeah, it hasn't hit production yet. ReleaseTaggerBot says it'll hit production next week (1.41.0-wmf.22; 2023-08-15) [16:29:05] pmiazga: At WMF we run jobs in one process per job, so it won't enter that method a second time unlike for most setups. The MW default is to run one or more jobs at the end of a pageview web request, by calling Special:RunJobs. The most common non-default set up is to use runJobs.php as CLI, that's what MediaWiki-Docker uses for example. [16:30:00] pmiazga: In prod we use wmf-config/rpc/RunJobs which is similar to Special:RunJobs (we should switch at some point probably, there's an unfinished task from the core platform team). [16:31:02] pmiazga: I don't recall off hand whether jobrunners, which don't go through Varnish etc, actually have req IDs generated externally. Possibly not, so maybe those autogen on first call. But only once. [16:37:10] Not sure who has the rights to do this but looks like pmiazga should be voiced? Like the rest of the team members? [16:39:33] xSavitar: try it! https://wikitech.wikimedia.org/wiki/MediaWiki_Engineering/Runbook/Access_control [16:41:43] Krinkle, nice! Thanks! Book marking that right away :D [16:42:40] Thanks for voice [16:58:41] * xSavitar grins [16:58:49] You're welcome pmiazga [17:10:36] Amir1: I think our proposed API for external DB handles doesn't cover the use case of an external cluster with a non-external dbname. i.e. access domain=false (current wiki) from x1.growthexperiemnts for example. I see (new) adoption of ILoadBalancer as a way to make this easier, curious what you'd recommend in this case? Stick with LB, or remember to pass cluster in each call when we have the new method, or local wrapper? [17:10:36] https://gerrit.wikimedia.org/r/c/mediawiki/extensions/GrowthExperiments/+/939737 [18:06:47] Krinkle: I think we should support the pseudo databases, something that was brought up in T330590. I can go over the details in depth in call or something. It's a bit weird [18:06:47] T330590: External LBs should not be exposed to developers - https://phabricator.wikimedia.org/T330590