[00:20:34] bd808: i am trying to get the unit tests to run in phpstorm and am having a hell of a time. i am now stymied by this: https://dpaste.de/5kxF/raw [00:20:37] Fatal error: Uncaught exception 'InvalidArgumentException' with message 'Provided specification lacks both factory and class parameters.' in /Users/ori/www/w/includes/libs/ObjectFactory.php:86 [00:20:51] how can i make it go away? [00:21:56] ori: what is $wgMWLoggerDefaultSpi set to? [00:22:19] it looks like it's set to null based on "ObjectFactory::getObjectFromSpec(NULL)" which is wrong... [00:22:56] Looks like null [00:22:57] Not sure what would cause that [00:24:06] this is the bootstrap file: https://dpaste.de/fQSC/raw [00:26:04] /** [00:26:05] * Activate the profiler (assuming $wgProfiler is set) [00:26:05] */ [00:26:06] protected function activateProfiler() { [00:26:09] the comment is not accurate [00:26:27] the trxprofiler gets enabled regardless of whether $wgProfiler is set or not [00:28:17] https://gerrit.wikimedia.org/r/#/c/202640/ seems to have introduced that [01:18:06] legoktm: if you can run the Audit script tonight, please do. The post-rename numbers are being requested for tomorrow's Monthly Metrics :) [01:18:55] If you don't have time, no problem, we can just go with the 2.8m ballpark rounding. [05:40:09] bd808: have you seen https://github.com/composer/packagist/issues/163 ? [05:49:05] Keegan: stats are running, I assume they'll be done by the morning [05:51:03] legoktm: thanks :) [08:57:37] so... what's the most appropriate phabricator tag to associate with mediawiki/core.git ? [08:58:05] I'm totally unable to deduce where the core/infrastructure team ended up [08:58:05] one of the mediawiki-* ones, why? [08:58:46] well because phabricator has project -> git repo mapping which enables some extra functionality if it's filled in [08:58:49] historically I'm not sure there was always a WMF team assigned specifically to MW (core) [08:59:29] mediawiki-general-or-unknown [08:59:40] ok [08:59:55] that's a really bad name for a project ;) [09:00:19] But it's certainly not a specific team [09:00:30] what is the extra functionality? [09:00:43] that's what I was trying to deduce ;) [09:01:18] mainly the ability to watch for differential changes and suggest reviewers I think [09:01:23] oh, yeah [09:01:50] assigning that to WMF team projects sounds like a bad idea [09:02:02] mediawiki-general-or-unknown will do, for now [15:30:11] Keegan|Away: so..I just re-opened https://phabricator.wikimedia.org/T95927 :( [15:38:35] ^d: you should totally reply to https://phabricator.wikimedia.org/T65847#1269168 :> [15:39:11] <^d> Why would I want to do that? [15:39:47] holy crap, it just started raining [15:40:22] <^d> https://www.youtube.com/watch?v=l5aZJBLAu1E [15:40:26] <^d> cc marktraceur [15:41:39] #old [15:41:44] It's been raining here for days [15:41:51] Fuck California so hard [15:42:21] On the other hand, in California, it gets cooler when it rains. Not true here. [15:42:29] ^d: well, aren't you the person? :D [15:42:51] ^d: Just reply "no". :-) [15:43:02] <^d> I'm going to pretend nobody's talking about this. [15:43:06] * ^d sticks head in sand [15:43:07] * James_F laughs. [15:45:07] <^d> Or maybe jfdi and see what happens. [15:45:29] <^d> Meh, we'd have to rebuild our plugins [15:45:40] totally down with jfdi [15:48:24] <^d> I actually just cd'd into my old clone of gerrit [15:48:33] <^d> And then remembered I can't remember how to f'ing compile it anymore even [15:52:23] jfdiiiiiii [15:59:05] <^d> Somebody convince qchris to redo the plugins and I'll do it :p [15:59:07] +1 for jfdi [15:59:54] <^d> by jdfi of course I mean scheduled downtime, carefully planned, with backups. [16:01:08] <^d> shit now i'm quoted. [16:01:35] <^d> shit now i'm quoted. [16:01:37] you said it in a publicly logged channel! [16:02:47] hmm https://phabricator.wikimedia.org/T70271#724244 [16:06:05] <^d> (1) - it's been a year [16:06:10] <^d> But worth checking [16:06:21] <^d> The other 3 are given, but that's why we'll do a database dump first. [16:08:23] <^d> I guess we can start talking about it and poking it. I don't see it happening pre-Lyon though [16:09:22] <^d> Which means not until June, I go on vacation after lyon until the 31st. [16:10:49] Keegan|Away: I got access to dbstore1001 which is only 24h behind (intentionally) [16:11:53] ummm, search backend? can't possibly be any worse than it is now [16:20:43] <^d> They rewrote it to use an embedded Solr backend [16:20:56] <^d> Early iterations were rough, but that was 2+ major releases ao [16:20:58] <^d> *ago [16:22:11] <^d> It's probably at least on-par with what we have now which I think is acceptable [16:22:20] <^d> Long as the syntax didn't get hosed or something [17:04:16] legoktm: errr, was that ping for me? [17:04:35] Keegan: yes. stats are re-running now [17:05:25] Ahhhh [17:05:38] Okay, makes sense now [17:05:53] I just saw it in the course of other conversation, out of context :) [17:34:03] Keegan: I don't think the stats are going to be done in 30 min...it's only at enwikibooks [17:34:33] legoktm: No problem. We're just gonna go with the rounding [17:34:51] I explained (again) that we're finalizing finalization, it's not finally final until it's final. [17:56:41] Keegan: Finally finally final? [17:57:09] James_F: And the the OED changes the definition of final. [17:57:14] *then the [18:51:55] AaronSchulz: thanks for the reviews :D [19:36:29] in recentchanges API, what is the difference between revid and rcid? [19:37:35] revids are for revisions [19:37:37] SMalyshev: revid is the rev_id from the revision table (also rc_this_oldid, IIRC). rcid is the rc_id from the recentchanges table. [19:37:52] rcids are for any recentchanges, they could have a log id instead of revid [19:49:39] MatmaRex: Awesome! (re 209570 and jsduck) [20:32:22] :D [20:48:13] Keegan|Away: https://meta.wikimedia.org/wiki/Special:Log/gblrename [20:51:50] legoktm: heh. "Invalid username 2667~enwiktionary" is a great username ;) [20:59:54] :P [20:59:58] at least they can login now! [21:00:09] if they somehow know they were renamed [21:07:01] <_joe_> manybubbles: you guys are going to do the full monty of scrum I see :) [21:07:31] _joe_: probably not [21:07:56] <_joe_> manybubbles: I'm just trolling you :P [21:07:57] just creating more boards so that the search team board can be Deskana's place. [21:08:09] I feal like I'm trolling me, from time to time [21:08:11] <_joe_> I was referring to kevin's email about meetings [21:08:47] <_joe_> I quickly calculated that participating in that number of meetings will quadruple the time I spend in meetings in a week :P [21:09:29] <_joe_> err s/will/would/ [21:12:37] err- I still haven't read that one.... and I said I would [21:31:36] does someone want to review https://gerrit.wikimedia.org/r/#/c/208988/ ? I already finished running it in production [22:55:38] yuvipanda: I was thinking about the logstash for tools thing last night. [22:55:45] How will we do auth for it? [22:56:14] bd808: have a proxy in the middle that adds a field based on identd? [22:56:14] If we run the instances inside labs we can't take ldap creds [22:56:16] like, just uid [22:56:28] bd808: ah, yeah. I might have to make someone pony up some hardware at some level. [22:56:45] bd808: *or* convvince people that LDAP creds are ok if it’s in a project that’s got NDA’d roots [22:57:00] slippery [22:57:32] bd808: yeah [22:57:40] bd808: the ponying-up-hardware is probably doable [22:57:40] *or* build an auth system that takes creds on wikitech or elsewhere safe and only gives a token to the app [22:57:58] bd808: we can re-enable OAuth on wikitech, yes [22:58:01] that’s an option too [22:58:09] bd808: it was enabled, just got lost in the migration to silver [22:58:30] bd808: that might be the simpest one, maybe [22:58:49] is there a mod_authn that works with our OAuth setup? [22:58:58] probably not [22:59:06] but: won’t it need to be more fine grained anyway? [22:59:20] because we’ve to deal with it / restrict it in kibana / ES too [23:00:13] that's an additional problem, yes [23:00:42] I looked into what the ES security plugin does (not how it does it). Even if we had it it would be a huge pain [23:00:48] bd808: yeah. [23:01:07] bd808: I was wondering if we should use something that’s *not* ES for the backend, but then that becomes a much bigger project, maybe [23:01:42] actually ES as the backend wouldn't be bad if we put something smarter than kibana in front [23:01:48] but that's a pain too [23:02:24] yeah [23:02:34] bd808: how about we put a layer between kibana and ES? [23:02:46] maybe... [23:02:53] that might not be too hard acutally [23:02:54] which basically mirrors the ES API (is simple pass through) but with authn [23:02:55] yeah [23:03:03] so you just pass cookie or whatever from kibana ‘onward' [23:03:23] there's this option too -- https://github.com/elastic/kibana/blob/kibana3/sample/filtered-alias-example/nginx.conf [23:03:37] crazy nginx proxy config magic [23:05:29] bd808: haha wow [23:05:37] bd808: but yeah, since it’s all http we can put proxies at different levels [23:05:49] one thing for sure is that it’s all going to be a fucking PITA :P [23:05:54] bd808: what does heroku do? [23:05:59] * yuvipanda hasn’t used [23:06:06] * yuvipanda checks [23:06:09] cli that you can use to tail logs [23:06:29] and then there are cloud ES solutions you can buy [23:12:27] bd808: maybe we can just do that [23:12:35] because they do the same thing now [23:14:25] let's figure out how to put all the logs in a pile for you to see first I guess [23:14:39] which is what led me to wonder about auth last night [23:15:16] I guess we could make it something that you had to ssh into the project and open a reverse tunnel to see to start with [23:21:23] TimStarling: have you been able to use /tmp/perf-PID.map to get PHP symbols in `perf report`? [23:21:29] bd808: yeah. [23:22:25] I think so, I remember using that feature for something, most likely perf report [23:22:52] yeah, pretty sure I did [23:23:34] bd808: step one is to just pto have a cluster that takes in something [23:24:04] TimStarling: I can't get it to work. A nonempty file with 'PID' matching HHVM's exists, but perf ignores it. I tried chowning it to root but it didn't make a difference. [23:25:50] weird, now it is reading it but not resolving any symbols [23:26:11] it's just a text file isn't it? you wouldn't think there would be a breaking format change [23:27:31] yuvipanda: which of tools logstash or beta nfs logging is a higher priority? [23:27:39] bd808: right now? beta nFS [23:27:45] because it’s waking me up at night :P [23:28:00] yuck! I'll work on that first then [23:28:01] and is an easier win [23:28:11] bd808: yeah, 3am catchpoint yesterday night... [23:28:26] well that’s not from beta itself, but still. [23:28:45] it happened day before, and if it were deployment-bastion again I’d have just shut the VM down and gone back to sleep but that sounds mean now that I say it [23:29:17] I think the first thing to do is just make it all like prod; fluorine clone and change apache configs to log via syslog [23:29:53] what i'm seeing : http://i.imgur.com/lmibpcj.png [23:31:11] i can't get perf to use HHVM's perf-PID.map to resolve symbols. here's what I'm seeing: http://i.imgur.com/lmibpcj.png . perf shows that it is using perf-31743.map , but i don't get any symbols. [23:31:20] grr, that was meant for #hhvm, sorry. [23:36:28] bd808: do you think you can start doing that? [23:36:47] I’m somewhat burnt out on NFS atm, and would like to finish up some webservice consolidation work... [23:37:40] yuvipanda: I think I can, yes. Might not see any results until the weekend though [23:38:04] bd808: <3 let me know if there’s anything I can do to help [23:56:25] legoktm: Lovely :D [23:57:13] legoktm: When can you close the task? [23:58:18] Keegan: a script is running right now to check there are no more invalid usernames left and whenever that finishes (it's really really slow) I'll close it [23:58:31] Yay [23:58:34] Thanks :)