[00:00:14] halfak: yep, that makes sense [00:55:36] ori-l: Got a good run at https://meta.wikimedia.org/wiki/Research:Module_storage_performance [00:55:42] I'm calling it for the night. [00:56:18] I left little boldly stated TL;DRs to give a quick gist of the results. [01:31:04] halfak: quick question [01:31:21] "Mobile doesn't really benefit from caching." is ambiguous -- do you mean, repeated page views doesn't indicate that native browser caching is helping any [01:31:25] or module storage caching isn't helping any? [01:31:28] the latter, right? [01:32:06] It's hard to tell a significant performance increase due to caching in mobile browsers. [01:32:47] Only mobile firefox, chrome and safari get a substantial performance increase. [01:33:09] All of the desktop OS's (and their browsers) tend to get a substantial load time decrease by the second request. [01:33:25] ori-l: ^^ [01:34:28] "doesn't really" isn't fair. It's just that we see this nicely structured benefit in all desktop browsers and not in mobile browsers that aren't Firefox. [01:34:54] innnnteresting [01:35:32] Yeah. I'm amending my tldr to "Mobile doesn't benefit from caching as much or as consistently as non-mobile." [01:36:03] k! I am wondering if there's something funky in how MFE is using ResourceLoader [01:36:14] but that's for me to investigate, don't worry [01:36:16] I spent a lot of time fiddling with this on Saturday. I have a lot of caveats to add, but none that I expect you'll find would affect the interpretation. [01:36:36] just one last question -- is there a TL;DR figure? a geometric mean across all platforms / page loads? [01:36:40] does it make sense to have one? [01:36:45] or a concise range I could specify? [01:37:05] The best I've got for that is under Grouped Analysis. [01:37:15] It says "Module storage is faster" [01:38:01] Oh wait. I see what you mean. I'll generate that. [01:38:10] When do you need it. [01:38:12] ^? [01:38:15] ori-l [01:38:28] ideally soon, how hard is it to generatE? [01:38:45] my naive thought was that perhaps averaging the diff for page views would do it [01:39:02] but i guess the distribution of overall pageviews to load index is not even [01:39:06] but tilts to the left [01:39:40] is it something that i could easily compute? [01:40:22] Give me 10 minutes? [01:40:31] sure! [01:58:21] Commons takes forever! [01:58:26] https://commons.wikimedia.org/wiki/File:Enwiki.load_time.geo_mean.cached_vs_not.svg [01:58:35] ori-l: ^ Something like this. [01:59:03] halfak: is there a number to go with that? [01:59:13] Yes. Getting that. [01:59:49] aweeeeesome [02:00:31] *everything* after that has no deadline on it at all; I'm super grateful that you put in all this time [02:00:41] consider me your dev-slave! [02:02:34] Bah! The things I do for formating a 2x2 table. [02:02:34] cached test control [02:02:35] 1: not cached 4263.163 4274.376 [02:02:35] 2: cached 1690.320 1846.696 [02:02:47] Oh wait. let me throw a diff on there. [02:04:38] cached control test diff [02:04:38] 1: not cached 4.274376 4.263163 -0.01121321 [02:04:38] 2: cached 1.846696 1.690320 -0.15637645 [02:04:53] I'll go throw this up onto the wiki page too. [02:05:25] the 'not cached' doesn't seem esp. interesting, no? [02:05:35] Nope. [02:05:37] especially since localStorage persists through cache clears [02:05:52] Oh yeah. I was simplifying. [02:05:54] Let me explain. [02:06:10] so it's really only one request ever that these measruements apply to [02:06:11] Not cached == Load_index=0 [02:06:14] or one in a blue moon [02:06:24] Cached == Load_index between 1 and 19 [02:06:47] right, but load_index 0 is not especially interesting, right? [02:07:02] Well, it's good to know we didn't break anything. [02:07:09] But I can cut it if you'd like. [02:07:14] halfak: oh good point! [02:07:16] i hadn't thought of that! [02:07:21] no, keep it then [02:07:32] OK, so I think that's it [02:07:39] * ori-l hugs halfak [02:07:45] yay!!! [02:09:34] Woot! [02:10:10] I'm glad things worked out. Remember me the next time you wanna look at some data. :D [02:18:29] :) [08:48:33] hi [08:48:57] I'm having troubles uploading a csv file into wikimetrics [08:49:22] is hewiki supported / working [08:49:25] ?? [09:04:51] i keep getting Server error while processing your upload [19:02:39] qchris, I actually ahve 2 hours before I have to go teach, should we deploy new webstatscollector? [19:02:52] ottomata: If you have time ... sure! [19:03:04] \o/ [19:06:30] Having no clue how to deploy webstatscollector ... what can I do? ... since I do not have root I guess it's just sit and watch? [19:06:38] hi, anything i can help with x-ff card #20? [19:06:56] !card #20 [19:06:57] https://mingle.corp.wikimedia.org/projects/analytics/cards/#20 [19:06:58] trynig to do a context load [19:08:10] yurik: card #20 says "Migrate Oozie jobs to generic stats user" for me ... are you really referring to that card? [19:10:33] yurik: Oh ... Scrum-of-Scrum card #20 :-) [19:10:59] qchris, yep, that bot needs fixing ;) [19:11:14] !card s20 [19:11:14] https://mingle.corp.wikimedia.org/projects/analytics/cards/s20 [19:11:34] vs !card a20 [19:11:53] yurik: Mhmmm ... I do not think there is anything you can do right now to help us getting the xff done. [19:12:14] should this card be postponned somehow? [19:12:47] otherwise it sits there, scarring us and confusing others :) [19:12:48] yurik: there is still the discussion about "what's a pageview" ... that somehow entails the xff handling. [19:13:12] yurik: I would not mind taking it down for now... Who would decide such things? [19:13:26] yurik: It still is a dependency for you on us... right? [19:13:44] ok qchris [19:13:45] ummmm [19:13:50] first I will build a .deb [19:20:18] qchris, i guess it should be dan scheduling all that, right? [19:21:31] (PS1) Milimetric: comscore update and small core fix [analytics/reportcard/data] - https://gerrit.wikimedia.org/r/98872 [19:21:48] hm, qchris, i think because the source has been changed, we have to bump the version number, not just the deb version [19:21:48] (CR) Milimetric: [C: 2 V: 2] comscore update and small core fix [analytics/reportcard/data] - https://gerrit.wikimedia.org/r/98872 (owner: Milimetric) [19:22:15] Which Dan? From my point of view, it's up to the wikipedia zero team to decide how/if wikipedia zero is dependent/blocked by analytics :-/ [19:22:27] yurik: But I might be wrong. [19:22:37] dfoy [19:23:01] yurik: Yes, dfoy sounds good to me. [19:23:42] hehe, well, obviously analytics is critical to us! in reality, zero is not about selling the idea to carriers or end users, its about showing good numbers to sue :D [19:23:57] Hahaha :D [19:25:17] (PS1) Ottomata: debian/changelog - Bumping actual version number because source has changed [analytics/webstatscollector] - https://gerrit.wikimedia.org/r/98877 [19:25:42] (CR) Ottomata: [C: 2 V: 2] debian/changelog - Bumping actual version number because source has changed [analytics/webstatscollector] - https://gerrit.wikimedia.org/r/98877 (owner: Ottomata) [19:29:17] (PS1) Milimetric: rolling back core fix [analytics/reportcard/data] - https://gerrit.wikimedia.org/r/98879 [19:29:31] (CR) Milimetric: [C: 2 V: 2] rolling back core fix [analytics/reportcard/data] - https://gerrit.wikimedia.org/r/98879 (owner: Milimetric) [19:32:59] ok qchris, no changes to collector, right? [19:33:06] Right. [19:33:20] Just saw the msg in the ops channel \o/ [19:33:45] i'm going to go ahead and install and restart filter on gadolinium [19:34:02] sorry [19:34:04] collector first [19:34:09] shoudln't change, just want the installed versions to match [19:34:25] Sure. [19:37:10] ok, done, now doing filter on erbium... [19:37:19] Yippie [19:39:00] ok, done! [19:39:11] guess the best way to verify is…wait and see? [19:39:24] I think so... :) [19:39:25] Thanks. [19:39:38] I'll have an eye on the webstatscollector files. [19:40:05] cool [19:40:29] ottomata: usually when I was updating, you would make a stream for me on some port, so I could test the modified webstatscollector with it [19:41:00] yeah, qchris has been testing stuff, this is just final install [19:41:05] I see [20:55:10] (PS1) Milimetric: Starting OAuth implementation [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/98958 [23:52:24] hi [23:52:48] need some help [23:58:58] can anyone help with user metrics