[00:15:33] legoktm: What's the timeline for UrlShortener? This quarter? Next quarter? [00:47:00] * legoktm will respond on phab [00:56:17] tgr: haven't forgotten about sentry, still working on a change [01:25:22] tgr: yt? [01:50:32] ori: just returned [01:50:53] the auth script looks great, thanks! [03:12:59] tgr: do you remember who told you not to have the database created by puppet? [03:25:44] ori: Alex, here: https://phabricator.wikimedia.org/T96054 [04:10:42] well, ok [04:10:51] i think he's wrong in this case, but i don't blame you for not wanting to argue [15:25:32] hashar, aude: i'll now move the jobs that run core with composer from experimental to gate, that should prevent such errors: https://gerrit.wikimedia.org/r/#/c/240110/ [15:26:41] jzerebecki: \O/ [15:27:23] jzerebecki: {be_bold} and feel free to announce it on wikitech-l / qa [15:27:50] jzerebecki: a potential side effect is that it will introduce a few more copy of mw/core on the Jenkins slaves [15:29:51] hashar: from what we last discussed for nodepool it seems we have enough hardware... [15:30:01] yeah [15:30:20] gotta run the jobs on Nodepool instances though [15:30:33] haven't looked yet at installing all dependencies needed for mw/core tests :-( [15:30:39] yea so many things to improve with ci so few people :( [15:31:09] jzerebecki: sounds good [15:31:42] jzerebecki: I wanted to migrate a bunch of tox jobs today but got delayed :D [15:32:02] !log reloading zuul for 82153d8..5ddf198 [16:44:20] tgr|away: sentry landed [16:46:55] poof [17:01:04] can't beat an Airbus Beluga landing ... https://www.youtube.com/watch?v=0KVrkJmDH9w [17:04:34] anyone know when bd808 is coming back? [17:11:38] phabricator says Mon, Sep 28, 12:59 AM [17:12:07] So Sun, Sep 27, 23:59 [17:15:02] thanks [17:17:45] ori: thank you so much! [17:18:41] tgr: thanks for the awesome work. i think alexandros is wrong on the manage_db thing in this case -- the module ought to manage the database in all cases, for the sake of sanity/simplicity. but i didn't want to hijack your patch to make that argument, so i'll submit a follow-up patch and we can debate it there [17:18:49] either way it doesn't need to block deployment or anything [17:21:39] ori: I generally feel like the new "every DB change on prod must be done by a DBA" policy is a but unrealistic, but then I know exactly nothing about a DBA's job so I'll do whatever people will tell me to do wrt database handling [17:22:02] so what's next for sentry? [17:22:56] it needs a nice way to manage users, LDAP or mw.org OAuth [17:24:10] after that I'll start advertising it on wikitech-l, see if labs and other non-prod projects are interested [17:24:43] enable it on translatewiki [17:25:15] replace the current ad hoc instance in the beta cluster and see about pushing PHP error data from logstash to it [17:25:53] https://github.com/getsentry/sentry/issues/1228 is useful, sentry uses django-social-auth [17:26:01] which has a pluggable backend system [17:26:02] try to find somone from the mobile app teams who is interested in replacing the current proprietary crash reporting code [17:26:44] there's a django_auth_ldap plugin [17:27:03] social auth in Sentry is deprecated, they want to roll their own SSO [17:27:16] actually did so but did not opensource it yet [17:27:44] that said it's django app so no reason why any django plugin wouldn't work [17:27:48] right [17:28:08] https://phabricator.wikimedia.org/T97133 is the ticket for that [17:28:40] django-auth-ldap seems like the way to go, IMO [17:29:44] the other next step is https://phabricator.wikimedia.org/T91649 if the multimedia team is still interested [17:39:38] anomie: (re: using UserGroupsChanged in Echo) we probably dont' want to send notifications for autopromotions...will our current check of $wgUser == $user prevent that? Or is there a better way? [17:41:42] legoktm: User::addAutopromoteOnceGroups() looks to only get called from page editing, and most of the time the user on the edit being saved is going to match $wgUser. [17:42:12] most? [17:42:53] if something manually calls WikiPage::doEditContent() I suppose [17:43:03] Does anything call WikiPage::doEditContent() and pass some other user for the $user parameter? [17:44:07] probably [17:44:15] Although I'd say that "You made enough edits to automatically become a patroller" wouldn't be a bad notification anyway. [17:44:52] yeah, but we need to detect whether it was autopromote or not, because the message text should be different [17:46:45] can we add another parameter to the hook indicating whether it was autopromotion? [17:49:06] I don't see why not. [17:52:47] ooh actually [17:53:09] I'll pass in the user who performed the user rights change, and false if it was autopromotion [17:53:21] that'll let us kill the $wgUser usage entirely [17:54:13] Nice