[18:32:55] !log awjrichards synchronized wmf-config/InitialiseSettings.php 'Make testwiki use mobile domain for URLs' [18:32:58] Logged the message, Master [18:36:00] https://test.wikipedia.org/wiki/Special:Version is worse than usual [20:08:17] https://en.wikipedia.org/wiki/Wikipedia:Village_pump_%28technical%29#Why_did_Wikipedia_change_its_export_function_so_it_no_longer_works_with_any_other_wikis.3F ?? [20:40:04] !log awjrichards synchronized wmf-config/CommonSettings.php 'Fixing mailto links on mobilefrontend feedback form to properly populate subject lines' [20:40:07] Logged the message, Master [21:10:50] p858snake|l: does anyone in particular need to create the account? What would the email address be? I can create it and then pass the details onto whoever to plug it into the bot [21:12:12] probably better someone at the wmf does it [21:12:28] I vote Reedy, just because [21:13:45] yeah no problem, it's just one of those trivial bugs that anyone can fix :) [21:19:56] so, we were talking with ottomata about analytics & SSL [21:20:23] Ryan was telling me yesterday that we don't actually log log lines coming from the nginx (SSL) [21:20:37] well, hopefully [21:20:43] if we are then our stats are fucked anyway [21:20:50] because of missing/uncorrelated sequence numbers [21:21:06] which I just fixed (will commit in a bit, after testing it) [21:24:32] hi paravoid [21:25:09] that's better :) [21:25:30] so, if we are depending on the stats from squid/varnish rather than from https we're collecting improperly [21:25:54] i don't know exactly what sources are doing what rigih tnow, [21:26:08] we should be ignoring logs with IPs that originate from our networks [21:26:20] but, it would be weird if we were collecting from both the nginx proxy and the sources that the proxy goes to [21:26:40] do you know if nginx is actually logging to udp2log daemons right now? [21:26:44] yes. it is [21:26:50] it is but with wrong seq numbers [21:27:28] i don't think ther eis any front end filtering going on from udp2log [21:27:38] either way the logs will be bad [21:27:40] but erik or whoever might be filtering out bad IPs [21:27:42] yeah [21:27:45] they'd have to do it on that side [21:27:48] wouldn't it be bset if [21:27:54] 1. nginx added x-forwarded-for header [21:27:57] it does [21:28:02] 2. nginx logs went to a different udp2log instance [21:28:03] ? [21:28:28] squid/varnish strip XFF for anything that doesn't come from the ssl cluster, too [21:28:35] k [21:28:37] that's cool [21:28:59] ma rk had some good reasoning behind why we should be logging directly from the ssl servers [21:29:00] so right now, if all these log sources are going to the same udp2log instance [21:29:06] that will be duplicate log lines for the same request, right? [21:29:07] hell if I can remember it off the top of my head right now [21:29:20] we can still do that, but why not log to a different udp2log instance [21:29:28] rather than the ones that the main frontends use? [21:29:35] that way if we want to see SSL access logs [21:29:49] we can, but they'd be separate from the usual frontend access logs [21:29:51] that sounds good to me [21:30:08] it would be best to discuss this with ma rk [21:30:11] ayte [21:30:30] his reasoning was good enough for me to actually spend time writing a logging module for this [21:30:54] I know for sure this'll be a problem with ipv6 [21:30:57] hm [21:31:00] would it? [21:31:01] maybe not. similar with XFF [21:31:03] it's exactly the same [21:31:06] right [21:31:51] so, we better hope that people using the logs take XFF into account ;) [21:32:33] if we separate these logs from the usual ones [21:32:47] they won't have to think about it unless they are trying specifically to look at SSL logs [21:33:08] * Ryan_Lane grumbles [21:33:08] wifi [21:33:23] anyway, we better hope that people using the logs take XFF into account [21:33:27] otherwise their stats are wrong [21:33:36] it's normal to ignore XFF [21:33:39] oh, even for the proxied ones [21:33:43] right [21:33:52] i betcha they aren't [21:34:02] aren't ignoring it? [21:34:11] we want them to use it, not ignore it [21:34:14] uh, no aren't accounting fo rit [21:34:16] yeah [21:34:20] * Ryan_Lane nods [21:35:38] its too bad it is near the end of the line though [21:36:09] betcha erikz would see it more if it was right next to the main ip [21:36:15] ideally people would just ignore XFF and strip any requests from internal IPs [21:36:32] any request from an internal IP is trash data [21:37:17] why? [21:37:17] then duplicate logs messages in this context don't matter [21:37:21] because nginx is logging? [21:37:23] hmmmm [21:37:26] i guess so [21:37:32] health checks are trash data too [21:37:36] same with monitoring checks [21:37:40] aye true [21:37:51] erik z might be doing that [21:37:52] i dunno [21:38:24] he'll account for that and we'll see that we've lost 10% of our traffic :D [21:56:44] gn8 folks [22:12:39] * robla starts looking into namespace configuration on test2 wiki [22:13:18] I'd like to get test2wiki set up for ProofreadPage testing, and it looks like the Index: namespace needs configuration [22:14:42] Might be missing subpage [22:14:50] no.. apparently not [22:17:01] !log reedy synchronized wmf-config/InitialiseSettings.php 'Enable doublepage on test2wiki' [22:17:03] Logged the message, Master [22:18:16] fwiw, here's the djvu file I uploaded: http://test2.wikipedia.org/wiki/File:The_book_of_try_and_learn.djvu [22:18:36] yeah, already found it ;) [22:23:03] !log asher synchronized wmf-config/db.php 'pulling db45, last coredb on prior fb mysql build' [22:23:05] Logged the message, Master [22:25:40] https://test2.wikipedia.org/w/index.php?title=Index:The_book_of_try_and_learn.djvu [22:26:03] oh, there we go [22:26:28] one I actually created the title page, it all sorta clicked [22:26:40] nope [22:26:56] Reedy: did you just fix something? [22:26:58] I just created https://test2.wikipedia.org/w/index.php?title=MediaWiki:Proofreadpage_header_template&action=edit ;) [22:28:19] And just imported https://test2.wikipedia.org/wiki/Template:Header [22:28:47] !log asher synchronized wmf-config/db.php 'returning db45' [22:28:47] It's behaving more like enwikisource now [22:28:49] Logged the message, Master [22:29:00] thanks! [22:30:02] chrismcmahon: http://en.wikisource.org/wiki/Page:The_book_of_try_and_learn.djvu/1 [22:30:05] Most likely missing some CSS and stuff [22:30:11] but that shouldn't bean issue [22:30:13] oops [22:32:36] chrismcmahon: here we go: http://test2.wikipedia.org/wiki/Page:The_book_of_try_and_learn.djvu/1 [22:34:06] robla: test2 doesn't seem to be djvu-ing [22:34:48] chrismcmahon: how so? [22:34:57] https://en.wikisource.org/w/index.php?title=Page:The_book_of_try_and_learn.djvu/1&action=edit [22:35:37] djvu 2 :p [22:36:02] robla: better now, before I was seeing "There is currently no text in this page. You can search for this page title in other pages, search the related logs, oredit this page." [22:36:22] probably caching [22:42:45] http://en.wikisource.org/wiki/Index:The_book_of_try_and_learn.djvu and http://test2.wikipedia.org/wiki/Index:The_book_of_try_and_learn.djvu look quite different, though [23:02:40] ToAruShiroiNeko: yes I do