[00:26:37] hi lwelling1 [00:26:51] welcome again :-) (now that your welcome email has gone out) [03:58:46] hi pastakhov [14:05:46] New patchset: Hashar; "enable operations-puppet validation job" [integration/jenkins] (master) - https://gerrit.wikimedia.org/r/29966 [14:07:16] New patchset: Hashar; "take fingerprint before PHPUnit tests" [integration/jenkins] (master) - https://gerrit.wikimedia.org/r/29967 [14:07:22] New review: Hashar; "synced from production" [integration/jenkins] (master); V: 1 C: 2; - https://gerrit.wikimedia.org/r/29966 [14:07:22] Change merged: Hashar; [integration/jenkins] (master) - https://gerrit.wikimedia.org/r/29966 [14:07:32] New review: Hashar; "synced from production" [integration/jenkins] (master); V: 1 C: 2; - https://gerrit.wikimedia.org/r/29967 [14:07:33] Change merged: Hashar; [integration/jenkins] (master) - https://gerrit.wikimedia.org/r/29967 [17:47:57] Hey MaxSem: were you the one working on CologneBlue? [17:50:02] csteipp: MatmaRex, in #mediawiki [17:50:33] csteipp, nope [17:50:49] Amgine: Thanks! [17:50:59] Sorry about that Max! [18:02:06] csteipp, I don't know if you saw but labs recently decided to split the admin role on labsconsole so they had an extra group without interface-editing rights [18:02:22] So that it couldn't be used to take over cloudadmin accounts [18:02:59] Yeah, I saw that was happening, but I hadn't actually dug into any details. Thanks! [18:03:10] I don't know if WMF has ever considered restricting it in production for similar reasons with steward/OS/CU accounts [18:04:14] <^demon> Krenair: [[w:WP:BEANS]] [18:27:49] Hi mlitn : Checking in on the status of today's AFT5 deployment. Are there any issues on your end? Spage says that he +1'd your revision for the metrics study, and I understand that you and kaldari are discussing feedback. Please let us know if there is anything we can do to help. It would be great if you could give us an estimate of when you expect these revisions to be on testing, so DarTar can check that we are collecting the data [18:27:49] properly. Thanks! [18:29:43] fabriceflorin, mlitn: I'm back in the office and ready to help with the AFT5 deployment when you are [18:30:17] fabriceflorin: I've requested some insight from ori-l on the clicktracking change as well, just waiting as he reviews it :) [18:30:35] hoi mlitn [18:30:51] provided he thinks it's ok, we're good to go [18:30:53] hi DarTar [18:31:12] double checking: we're now using consistently the anon token set by mw.user.id() for all events and dropped the CT session token, correct? [18:31:25] mlitn: Thanks for the update! Keep us posted. Fingers crossed … [18:34:43] DarTar: CT session token will be set as "token", for every event [18:34:52] mw.user.id() will be set as first param for "additional", for every event [18:35:31] which is the same behaviour as acux, judging from the log excerpts you sent last time :) [18:35:32] I see, excellent - we'll have to adjust the specs and the parser [18:35:33] well not quite [18:35:48] acux uses mw.user.id() as the token for all client side events [18:36:26] DarTar / mlitn: reviewing now [18:36:45] whether in AFT we store this in the actual token field or as additional data, I don't care - provided we have this [18:54:53] hashar: thanks for resurrecting https://gerrit.wikimedia.org/r/#/c/5174/! [18:55:42] gwicke: I was looking at the list of open changes for mediawiki/core and found that very old change [18:55:52] since it is very simple.. I have simply merged it :-] [18:55:57] (after test completions though) [18:56:11] thanks! [18:56:39] had kind of given up on that one.. [18:58:16] !b 28810 [18:58:17] https://bugzilla.wikimedia.org/show_bug.cgi?id=28810 [18:58:51] andre__ that one is specific to FF and not to Chromium ^^ [19:01:04] andre__ and the problem exists on test2 as well. [19:03:37] mdale: does TMH have any influence on how UploadWizard displays thumbs during the upload process? Or is that solely handled by UW? [19:06:07] marktraceur: you might be interested in BZ 28810 as well [19:09:46] I see, didn't know [19:10:22] andre__: and I just confirmed thumbs are black for an ogv on real Commons also. [19:10:26] in FF [19:10:53] I should probably check if there's an upstream ticket than if we know that it works well in several other browsers? [19:11:03] s/than/then [19:12:12] andre__: it's odd that 28810 is so old. UW had a lot of changes not long ago, the FF thumbs problem probably slipped through. [19:15:17] andre__: I think maybe we make a new bug report for the black thumbs and close 28810, does that make sense to you? [19:20:36] chrismcmahon, feel free to, sure [19:21:09] andre__: OK. the original few entries on 28810 are really obsolete [19:25:05] andre__: heh. fwiw, thumbs in UW are broken in Safari also [19:44:27] mlitn: Could we call you on Skype now to discuss the status of this AFT5 metrics deployment? I'm here with Dario, Ori, and SPage and we're trying to figure out what to do. [19:51:43] siebrand, are you satisfied with the new version of https://gerrit.wikimedia.org/r/#/c/28863/ ? [20:13:49] * DarTar is off to grab food, will be back in 10 [20:22:22] back [20:25:49] <^demon> Krinkle: Apache docs says default for UseCanonicalName is Off anyway, so I don't know why we need to set it anywhere. [20:25:58] ok [20:26:21] <^demon> Good thought, though. [20:30:25] ^demon: I didn't expect there to be a problem, but I'm generally interested in ops so I peek in every now and then and try to get a grasp of what it all means and how it works. It was more a "this is probably right, but I don't get it" than a "this is wrong, right?" [20:30:45] <^demon> Hey, I'll take any feedback I can get :) [20:30:59] ^demon: I guess though that removing it from wiktionary is a "if it ain't broken, don't fix it" kind of thing? [20:31:22] <^demon> Yeah. Put it under "Apache config still needs a lot of cleanup" [20:33:16] whatever works for now is okay i think [20:33:20] always can be improved later [21:12:52] fabriceflorin: it's on test ;) [21:14:17] Thanks, mlitn ! I just told DarTar, who will check the data now. Is there any other consumer-facing feature that's being deployed today? [21:14:58] there is, though hardly testable: the abusefilter threshold increase for feedback [21:20:04] fabriceflorin: I didn't know we were making further changes to abusefilter [21:20:33] please keep track of the exact timeline [21:26:30] DarTar: didn't actively plan to have that change released today, just happened to be merged in today [21:26:44] ic [21:27:16] it's only a small config change to make it less common for a filter to accidentally be auto-disabled [21:27:33] ok, so not a change to an actual rule [21:28:18] no, it isn't, these should remain unchanged (and active ^^) [21:33:44] mlitn, fabriceflorin: mail with results of CTA1 tests [21:33:58] one minor glitch + one potential concern [21:34:39] but we can try and exclude the concern by measuring page save time after removing AFT5 from the same article [21:37:57] mlitn: disabling AFTv5 on /San_Francisco to see if this has any effect [21:38:10] Thanks, mlitn, and DarTar! So glad the CTA4 test is now working. Sorry to hear that the CTA1 seems to introduce 30-second waits on the test.wikipedia.org site. Let's make sure we can reproduce this and that it's not related to test.wiki being sluggish. [21:40:56] mlitn, fabriceflorin: it actually takes the same time when AFT5 is disabled, must be the crazy number of templates to be parsed on that article [21:41:14] or some weirdness with testwiki [21:41:32] mlitn, DarTar : I will test the Abuse filter change directly on production. DarTar, this change will primarily impact only 1 filter (very short posts), and I will try to increase the length of disallowed posts from 2 chars to 5 chars, which could increase the number of disallowed AFT posts from 5% to 10%. I will also create an AFT5 history page to log in events like these, though keep in mind that the abuse filter will constantly be tweaked, [21:41:32] so it's not realistic to keep track of every single tweak. [21:41:49] fabriceflorin: it's not yet on production [21:47:42] DarTar: That is excellent news! Looks like we're ready to deploy to production. However, kaldari reports that Reedy may have accidentally deployed everything that was on fenari, so we may already be on production. Can you check? [21:48:20] mlitn: Is there a way to check whether or not our latest code is in fact on production, as suggested by kaldari ? [21:48:38] config changes may not be live however [21:49:00] fabriceflorin: there is no way to check it [21:49:22] and indeed, the extension was pushed, but I just reverted it because it shouldn't go out without the config :) [21:49:32] give me 5 min :) [21:49:48] DarTar: Here is the link to our Google spreadsheet, where we track the impact of the 7 abuse filters for Article Feedback -- it would be great if we could automate the tracking of that metric: https://docs.google.com/a/wikimedia.org/spreadsheet/ccc?key=0AiGAdIp7VYlbdDdKUm9naXhxOXVweWZ5YkU3Wk5lSlE#gid=0 [21:51:09] mlitn: Never a dull moment! So many plot twists today ;o) … Let us know when you are ready for us to test on production! [21:52:39] mlitn, fabriceflorin: completed a CTA4 funnel on enwiki but still see the old code [21:53:12] i.e., missing AFT impression, missing token from the additional data field - I think it's not live yet [21:53:54] DarTar: it's not yet on production, it had been pushed accidentally [21:54:00] and was reverted [21:54:33] ok [21:55:03] we're good to go any time [21:58:19] fabriceflorin: DarTar: we're on production now [21:58:38] Both the TimeGate and TimeMaps are separate resources and hence need different entry points. The different entry points is the only reason for us to create the 3 extensions. We created two special pages for timegates and timemaps. [21:58:40] * Reedy facepalms [22:00:56] mlitn: Cool! Will check the abuse filter now. Is it safe for me to increase the minimum from 2 to 5 characters per post on the 'Extremely short post' filter? https://en.wikipedia.org/wiki/Special:AbuseFilter/458 … or should I work gradually -- or wait some more? [22:02:13] fabriceflorin: I have no clue what percentage will be hit by a certain amount of characters [22:02:29] however, I would work my way down, starting by 5, going to 2 if it keeps being blocked [22:03:01] fastest way to figure out how much it can handle [22:04:52] mlitn: Sounds good. I will work incrementally -- however the order will be reversed, so I will start by disallowing posts with 3 characters or less , then 4 characters or less, then 5 characters or less. The higher the number of characters, the higher the volume of disallowed posts and the more likely the filter is to be disabled, if you follow my math ;o) [22:06:45] indeed, that's why I'd start with the highest number, you won't have to wait 3 days to know if it holds up, you'll likely find it's disabled after only a couple of hours [22:07:02] going down, as soon as it start holding up a couple of days, you know you've hit the threshold [22:07:21] going up, if it holds up a couple of days, adding another character might still hold a couple of days :p [22:07:42] anyway, doesn't really matter how you test it, either way we'll get there :) [22:09:33] mlitn: testing on enwiki [22:11:42] mlitn: You make an excellent point, thank you! I will adopt your method now and start at 10 characters, then go down if need be. Stay tuned … [22:19:00] CTA4 passed, check my report in your mailbox [22:19:27] my mail box? [22:20:10] mlitn: OK, abuse filter 458 is now set at 10 characters or less. Fingers crossed ... [22:21:28] DarTar: Wooho! Let's hope CTA1 tests as well ... [22:27:52] mlitn: I think 10 characters may prove to be too high. 5 minutes after I made that change , we are already at 5.88% (of the last 85 actions, this filter has matched 5). But I'm willing to settle for 5 characters, if that proves to be too much for the filter. I wonder if there is a mathematical relation between the 5% to 10% limit and the number or characters, which might suggest that we could double it up to 4 characters, but no more. We'll [22:27:53] soon find out ... [22:29:13] fabriceflorin: if it hits 5.88%, the filter shouldn't disable (should only disable at 10%) [22:31:41] CTA1 passed \o/ [22:32:49] DarTar: you may also notice the fixed event names (the missing dash is now there) ;) [22:33:15] ok, phew, got it out ^^ [22:35:58] yeah I saw that, terrific [22:39:06] mlitn: Yes, it looks like it may have been a temporary peak. It just went down to 0.48% (of the last 2,516 actions, this filter has matched 12). I wonder how they pick the number of 'last actions': could it be the total number of actions in the last minute? That might explain why that number keeps changing. [22:41:26] fabriceflorin: I don't know where those numbers come from, though I doubt that 2516 entries have been submitted in the last minute [22:43:40] mlitn: Yes, you're right, can't be the last minute. I guess we'll just have to live with the mystery for now. I'm just happy that it seems to be holding up, even if I don't fully understand how it works … ;o) [22:45:58] :D I'm curious how well it'll hold up [22:46:59] mlitn: Hey Matthias, kudos for your amazing work this week! You pulled it off against all odds, and your calm in the middle of the storm was very inspiring. Thanks for making this important test possible … now get some rest if you can. Cheers! [22:49:37] :) later! [22:56:36] hey chrismcmahon got a moment? [22:56:44] hi sumanah what's up? [22:57:06] chrismcmahon: are you up to speed on the situation regarding the new batch of problems with thumbnails? [22:57:07] chrismcmahon: if not, I wanted to update you. [22:59:10] siebrand: FYI I just reported in the itwiki channel that the translation of the "Create an account" link is inaccurate [22:59:22] it currently reads "Create an access" [23:00:07] sumanah: andre__ and I hacked on some thumbs issues earlier and found a new issue. I created https://bugzilla.wikimedia.org/show_bug.cgi?id=41403 and retired an old vague one https://bugzilla.wikimedia.org/show_bug.cgi?id=28810 as a duplicate of the new, specific one. [23:01:14] DarTar, that's 'nologinlink' [23:01:23] Should be dealt with on translatewiki I think [23:01:29] Krenair: indeed [23:02:10] not sure who's active there among itwiki admins/editors, I just wanted to report it b/c it doesn't make any sense [23:02:38] Actually, itwiki has it's own version: https://it.wikipedia.org/w/index.php?title=MediaWiki:Nologinlink&action=history [23:06:34] chrismcmahon: ok, lemme look [23:08:02] Krenair, siebrand: I just registered on translatewiki