[00:57:18] TimStarling: would you like me to sync https://gerrit.wikimedia.org/r/#/c/195196/ ("More secure permissions on conf cache")? Is that blocked on anything? [00:58:34] sure, you can sync it [00:58:42] permissions need to be updated manually afterwards [01:00:02] mmm, puppetize? [01:01:21] nah, a one-time salt/dsh thing, i think [01:09:05] TimStarling: sanity-check: find /tmp -maxdepth 1 -type d -name 'mw-cache-*' -exec chmod 0755 {} \; [01:09:12] running this on all the app servers should do the trick, yes? [01:10:36] that would do the directories, but what about the files? [01:11:46] files should be 0644, right? [01:12:01] how about find /tmp -path '*mw-cache-*/conf-*' -exec chmod og-w {} \; [01:12:36] symbolic permissions are really nice, I don't know why everyone uses octal ;) [01:12:46] yeah, that should work [01:19:00] 6MediaWiki-Core-Team, 7HHVM, 7Upstream: HHVM with FastCGI does not support streaming output - https://phabricator.wikimedia.org/T91468#1102524 (10tstarling) I have had a chat with sgolemon, paulbiss and afrind, and it's back into the "maybe possible" pile. It seems to me that emulating blocking I/O is prope... [01:24:32] OK, permissions updated everywhere [04:36:27] 6MediaWiki-Core-Team, 6CA-team, 10SUL-Finalization, 6operations: db1068 (s4/commonswiki slave) is missing data about at least 6 users - https://phabricator.wikimedia.org/T91920#1103151 (10Springle) >>! In T91920#1101449, @Legoktm wrote: > Thanks. How (un)likely is it that other wiki's slaves might have sim... [06:51:29] bd808: https://gerrit.wikimedia.org/r/#/c/192207/ will mean vagrant will need some updating [06:52:04] something like https://gerrit.wikimedia.org/r/#/c/195337/ [07:52:51] 6MediaWiki-Core-Team: Fix job runner service availability during redis outages - https://phabricator.wikimedia.org/T91835#1103602 (10aaron) Two places assume that a server that can be connected to works...which was wrong in the outages. Patch incoming. [13:03:23] 6MediaWiki-Core-Team, 6Multimedia, 10Multimedia-Sprint-2015-03-04, 5Patch-For-Review: SHA-1 file name support - https://phabricator.wikimedia.org/T1210#1104231 (10Gilles) [15:47:56] 6MediaWiki-Core-Team, 10Analytics, 10VisualEditor, 10VisualEditor-Performance, and 3 others: Apply Schema:Edit instrumentation to WikiEditor - https://phabricator.wikimedia.org/T88027#1104728 (10Jdforrester-WMF) 5Open>3Resolved [16:07:58] * bd808 wonders why Nik isn't in the meeting in Yongle right now :( [16:08:13] I should have gotten him and invite [16:12:48] bd808: Isn't he at a conference? [16:12:56] yeah [16:13:08] * bd808 tries to take good notes [16:17:33] * anomie emails ^demon|away about profiling, since he's |away [17:35:39] robla, bd808: I'm taking today off since Noam isn't feeling well.. just fyi. [17:35:58] ori: ok. Take care of the boy [17:40:09] csteipp: thanks for the security review of the statsd extension [17:45:37] No problem! [18:02:42] 6MediaWiki-Core-Team, 6CA-team, 10SUL-Finalization, 6operations: db1068 (s4/commonswiki slave) is missing data about at least 6 users - https://phabricator.wikimedia.org/T91920#1105596 (10Krenair) Are the other two db1062 and es1002? [18:50:36] aspiecat: Do you remember why we pinned jobrunner to a particular commit in mediawiki-vagrant? [18:54:11] probably simplicity, not sure [18:54:42] hmm... possibly [18:54:52] I'll see if I can figure that out [18:55:39] on a related note, do you have any object to running jobrunner and the new chron service using php5 instead of hhvm? [18:55:51] I think php5 would have lower resource needs for the vm [18:57:22] anomie, legoktm: Do you have a minute to review https://gerrit.wikimedia.org/r/#/c/195472/2 for andrewbogott? Looks to be a simple schema change for the OSM tables [18:57:32] bd808: I just +1'd it :P [18:57:37] perfect [18:57:40] or did he want a +2? [18:58:38] He wants to SWAT it I think, so he probably needs help knowing what to do [18:58:50] +2, backport prepared, db change made [19:01:32] ok, we're talking in -labs now [19:01:42] sweet [19:02:54] php5 is ok [19:03:57] legoktm: Is "ALTER TABLE foo SET COLUMN" valid MySQL? [19:05:22] I didn't test it... [19:05:33] * legoktm tries [19:06:03] * anomie had it fail locally with MySQL 5.5.42 [19:07:19] bd808: does it use HTTP requests or runJobs.php ? [19:07:46] aspiecat: it's using junJobs.php [19:07:48] the former could use even less ram [19:08:15] That's how you're running on prod now right? [19:08:16] * aspiecat should commit his basic web runjobs script somewhere [19:08:39] yeah, we have our own script that accounts for multiversion, but it's pretty tiny [19:09:25] bd808: in the /rpc dir of mw-config [19:09:52] perfect. I knew I'd seen it but wasn't remembering where [19:10:18] so it would be on extra hhvm server with one heap instead of a heap (plus slow startup) per temporary worker [19:12:28] So that runs a dedicated apache2 vhost -> fcgi HHVM in prod? Or is it just an endpoint on aother vhost? [19:12:49] Or better asked, could it just be an endpoint on mw-vagrant [19:12:57] like something in /w/ [19:14:05] I guess the downside of that would be dealing with folks who have turned on the zend role [19:14:22] but I think we still spin up hhvm even when that's the case [19:14:26] 6MediaWiki-Core-Team, 10Analytics, 10VisualEditor, 10VisualEditor-Performance, and 4 others: Apply Schema:Edit instrumentation to WikiEditor - https://phabricator.wikimedia.org/T88027#1105830 (10Jdforrester-WMF) [19:15:21] it not just virtual, since it's a local server on the runners which don't have apache for any other reason [19:15:33] *nod* [19:15:34] for vagrant, you might be able to not actually have another server but just a vhost [19:15:39] which of course means less ram [19:16:10] yeah. I'll make some notes on the task. Sounds like a fun thing to play with some evening soon [19:27:52] 6MediaWiki-Core-Team, 10VisualEditor, 10VisualEditor-Performance, 3VisualEditor 2014/15 Q3 blockers: Stash VisualEditor edits using edit stash - https://phabricator.wikimedia.org/T90040#1105890 (10aaron) [19:29:15] aspiecat: https://gerrit.wikimedia.org/r/195636 [19:30:47] probably needs to be backported [19:33:17] right, there was some ext [19:33:25] why would someone implement that anyway? [19:33:41] context already depends on MW core stuff a super lot [19:33:53] aspiecat: the whole extension is a huge mess, that's just one part of it [19:33:57] why not extend it? [19:34:05] well, whatever, heh [19:35:06] https://phabricator.wikimedia.org/T45975 [19:47:53] legoktm: why can't that use $this->getContext()? [possibly stupid question] [19:51:12] oh, it probably can [19:51:30] it's funny how stuff seems circular at first [19:51:33] aspiecat: updated [19:52:25] like the IContextSource special page taking in another one to proxy to...and then you wonder what if that is a special page too, and then you go "well it would have it's own context"...so it stops somewhere [20:00:14] csteipp: what's the best way to force users log back in? https://gerrit.wikimedia.org/r/#/c/195445/ current code doesn't work, but I don't know what the right way is [20:02:31] anomie: Updated RfC words to match your patch -- https://www.mediawiki.org/w/index.php?title=Requests_for_comment%2FAuthManager&diff=1439466&oldid=1432611 -- and changed PostAuthenticationFilter to PostAuthenticationAction [20:03:01] legoktm: User::doLogout? [20:03:02] bd808: Nifty [20:03:52] $this->user->logout() most likely... [20:04:46] csteipp: aha, missed that function. $this->user is actually an OpenStackNovaUser even though the documentation claims it's a string [20:46:20] ori: https://gerrit.wikimedia.org/r/#/c/195521/1 [21:13:39] <^demon|away> bd808: Big thing for Logstash 2.x? REST apis for stuff like managing your filters and log sources [21:13:42] <^demon|away> Rather than config files [21:14:07] so logstash becomes more like elasticsearch then [21:14:12] <^demon|away> Yep [21:14:16] not a horrible thing [21:15:42] <^demon|away> Logstash is going to be a cluster service in 2.x instead of a single process running on the ES instance [21:15:48] <^demon|away> (so you can hit any instance) [21:15:53] <^demon|away> afaict [21:16:31] oh? so they are moving it into elasticsearch as a plugin? Or ? [21:17:02] <^demon|away> Not sure. I don't /think/ so? But it sounds like a lot of wheel reinvention of not [21:17:08] <^demon|away> *if not [21:17:24] hmmm [21:17:36] that would have some interesting consequences [21:17:46] and probably make a better product for most needs [21:18:24] <^demon|away> Also, Elasticsearch [the company] is now just Elastic. [21:18:32] <^demon|away> They announced a rebranding today at opening keynote [21:18:55] <^demon|away> https://www.elastic.co/ [21:19:05] assembling multi-packet messages could be an interesting challenge in a cluster solution [21:19:18] .co please [21:19:43] <^demon|away> .co -> the .com I wanted was taken already :p [21:20:04] * bd808 only sees an animated gif because the site content is behind a javascript wrapper [21:20:49] <^demon|away> They're talking about folding the logstash forwarder into logstash itself [21:20:58] <^demon|away> (because logstash on jruby is too heavy to run on your edge nodes) [21:21:06] it really is [21:21:36] I need to look at what they have been doing in the kibana4 branch too [21:21:59] when I saw a ruby process come back I looked away and haven't returned [21:22:06] <^demon|away> I was going to go to the kibana4 talk but it was crazy overpacked. [21:24:00] oh goody. it's nodejs instead of ruby now [21:25:39] <^demon|away> logstash is going to pass MRI now so they can possibly move beyond jruby [21:25:50] <^demon|away> (because start up times suck) [21:26:09] <^demon|away> also: better packaging we promise!! [21:26:27] heh [21:26:46] that would have been easier if they hadn't hired the fpm guy :) [21:27:32] <^demon|away> But: Jruby is still best! [21:27:59] Only if they are using subbu's new experimental compiler [21:28:10] s/compiler/runtime/ [21:30:04] He's the #4 contributor these days -- https://github.com/jruby/jruby/commits?author=subbuss [21:50:55] for the travel request to Lyon form, what do we put for "Who will be approving your travel cost?" ? [21:53:47] <^demon|away> "Not me" [21:54:25] lol [21:55:11] legoktm: Quim [21:55:25] thanks [22:06:21] ori: https://gerrit.wikimedia.org/r/#/c/195507/ [22:16:50] legoktm: https://gerrit.wikimedia.org/r/#/c/195773/ [22:17:01] * autismcat ran into that testing [22:22:56] bd808: help me understand what is going on with the quarterly review planning: what is the difference between the meeting you had scheduled with RelEng and the meeting Quim scheduled with reps from all groups? and does the ordering matter? [22:23:42] greg-g: *shrug* Quim made the big meeting all on his own :( [22:23:57] Mayb eI should take people off of that [22:24:56] I can send an email to you, me, quim, and rob-la [22:25:09] just a "hey, what's the plan yo?" [22:26:45] taht would be swell [22:28:29] {{done}}