[07:11:26] <darthvader>	 hi there
[07:11:29] <darthvader>	 everyone
[11:09:07] <Reedy>	 Are the germans +1'ing https://youtrack.jetbrains.com/issue/WI-24808 ? :P
[11:14:07] <Reedy>	 FYI, they don't care about votes
[11:14:22] <Reedy>	 See also https://youtrack.jetbrains.com/issue/WI-473
[13:22:02] <harej>	 Here's an idea: is there any way to get API request data for Wikidata? While page views is one sort of metric, I feel like for Wikidata that API access is a more interesting piece of information.
[13:34:21] <addshore>	 harej: so you mean see what items have been accessed through the api?
[13:34:29] <harej>	 yes, and how many times
[13:35:02] <addshore>	 interesting, I mean, it is defintly data that we have just not in a nice way
[13:37:41] <harej>	 is it something that would be found on labs somewhere?
[13:43:04] <Reedy>	 no
[13:43:13] <Reedy>	 WMF have api logs and stats available... but not public
[13:43:34] <harej>	 what would it take to make this information publicly available? obviously with the PII redacted
[13:44:25] <Reedy>	 Probably a lot of effort
[13:44:42] <Reedy>	 Depends what exactly you're warning
[13:44:51] <Reedy>	 You could probably get analytics to write some reports for it
[13:45:17] <harej>	 I mean, would it require someone with access to nonpublic information?
[13:45:47] <Reedy>	  I guess
[13:47:42] <harej>	 This information is important to me; I'm going to try to get funding to do this. If I find a person and pay them, could I get them through the process?
[13:49:31] <Reedy>	 Depends on the WMF
[13:49:48] <Reedy>	 But someone appropriately NDA'd should be able to get access
[13:49:53] <Nemo_bis>	 harej: I assume you already checked the Research committee suggestions?
[13:49:53] <Reedy>	 Like I say, talk to analytics first
[14:00:48] <harej>	 Nemo_bis: I hadn't thought of that; I wanted to ask if the report existed first
[14:00:51] <harej>	 But sure, I can talk to analytics
[15:58:44] <a_>	 Please remove gujarati language wikipedia of "Anwar Agewan"
[15:59:12] <a_>	 hello
[17:22:54] <harej>	 so I asked Analytics and they're kind of working on it? but i don't know if it's exactly the thing I want.
[17:35:38] <harej>	 is Lydia_WMDE around?
[17:38:28] <DanielK_WMDE__>	 harej: Lydia_WMDE has gone home for the day
[17:39:09] <harej>	 you might know, in any case: is anyone currently working on metrics for access to wikidata's data by means other than visiting the website
[17:39:25] <harej>	 i can get page view data for wikidata but i don't think that's very interesting compared to machines using wikidata
[17:39:48] <hoo|busy>	 addshore was at some point poking at dump access data
[17:39:53] <hoo|busy>	 but even that is incomplete
[17:40:03] <hoo|busy>	 there are several mirrors for which we don't have access logs
[17:40:11] <DanielK_WMDE__>	 yea, i was also going to suggest talking to addshore
[17:40:20] <harej>	 addshore said he is not working on that precise thing
[17:40:38] <DanielK_WMDE__>	 there is actualyl two things to consider:
[17:40:45] <DanielK_WMDE__>	 access to the query endpoint
[17:40:51] <DanielK_WMDE__>	 and access to the linked data interface
[17:40:59] <harej>	 WDQS and Special:EntityData, respectively?
[17:41:07] <DanielK_WMDE__>	 they are on different serves, using different urls and different infrastructure
[17:41:19] <DanielK_WMDE__>	 harej: yes, exactly
[17:41:24] <hoo|busy>	 For special:entitydata, see https://grafana.wikimedia.org/dashboard/db/wikidata-special-entitydata
[17:41:33] <DanielK_WMDE__>	 and both of them may use the varnish caches for some things
[17:41:38] <hoo|busy>	 but the blazegraph instances also access that
[17:41:55] <DanielK_WMDE__>	 we have some stats for that, yea
[17:42:12] <harej>	 so that graph gives me *all* the hits to entitydata; i'd like to see a breakdown by item
[17:42:48] <hoo|busy>	 don't think we have that :/
[17:42:49] <DanielK_WMDE__>	 that's oprobably not possible for WDQS
[17:43:10] <harej>	 WDQS does not capture which items are served in queries?
[17:43:39] <DanielK_WMDE__>	 no. it doesn't know about items at all, actually
[17:43:47] <DanielK_WMDE__>	 it only knows about triples and resource uris
[17:44:04] <DanielK_WMDE__>	 it's an rdf triple store. it knows nothing of wikidata
[17:44:10] <harej>	 can one infer items from triples and resource URIs, or are we out of luck there
[17:44:25] <DanielK_WMDE__>	 you can, yes
[17:44:40] <DanielK_WMDE__>	 but you would have to log the full query and full response
[17:44:47] <harej>	 (i take it the triple is item/property/value?)
[17:45:04] <DanielK_WMDE__>	 some triples are like that 
[17:45:31] <DanielK_WMDE__>	 but some are item/property/statement and some are statement/qualifier/value 
[17:45:40] <DanielK_WMDE__>	 and some are value/field/literal
[17:45:42] <DanielK_WMDE__>	 etc
[17:45:54] <DanielK_WMDE__>	 the rdf mapping is rather complex. 
[17:46:30] <harej>	 i can see how it would be expensive to get individual items out of that
[17:46:40] <harej>	 so, let's disregard WDQS analytics for now. if i found you a volunteer and navigated them through the NDA process, could that person come up with entitydata access metrics on a per-item basis?
[17:47:31] <DanielK_WMDE__>	 harej: that should be possible, yes. and analytics hopefully has the raw data for this. but i don't know the details
[17:48:15] <hoo|busy>	 Yeah, we have the data in question in hive
[17:48:30] <hoo|busy>	 just takes someone to crunch it
[17:48:55] <harej>	 How much time should it take for an experienced engineer to produce such a dashboard?
[17:51:37] <hoo|busy>	 not sure how that part currently works
[17:51:46] <hoo|busy>	 so dunno how much work it would really be
[17:51:54] <hoo|busy>	 but I presume not that much
[17:58:12] <addshore>	 harej: the api stuff is only very recently in hadoop / hive
[18:11:12] * aude waves
[18:11:47] <hoo|busy>	 hey aude :)
[18:14:30] <hoo|busy>	 Just landed?
[18:18:31] <aude>	 hoo|busy: last night
[18:18:49] * aude assumes the train is happening today
[18:19:01] <hoo|busy>	 I guess so
[18:19:12] <hoo|busy>	 hope we can see a huge drop in UPDATEs on tendril
[18:19:21] <aude>	 thanks for sorting out the deployment stuff
[18:19:27] <aude>	 :)
[18:21:36] <hoo|busy>	 :)
[19:12:59] <hoo|busy>	 aude: https://tendril.wikimedia.org/host/view/db1018.eqiad.wmnet/3306 Look at that
[19:13:04] <hoo|busy>	 "Query Write Traffic -- 24h / 5m"
[19:16:27] <aude>	 \o/
[19:30:39] <hoo|busy>	 also https://tendril.wikimedia.org/host/view/db1052.eqiad.wmnet/3306
[20:15:17] <multichill>	 Lydia_WMDE: Did you see the fun javascript hacks we did on Wikidata? ;-)
[20:15:55] <multichill>	 https://www.wikidata.org/w/index.php?title=Wikidata:Europeana_Art_History_Challenge/Content_Translation&withJS=MediaWiki:EuropeanaContentTranslation.js&itemid=Q2258948&destlang=nl
[20:16:41] <multichill>	 Javascript magic, feels like the Wiki Loves Monuments preparation
[21:24:47] <jzerebecki>	 DanielK_WMDE__: I wonder if it is worth it to deploy https://gerrit.wikimedia.org/r/#/c/283555/ now, to understand https://phabricator.wikimedia.org/T132645 
[21:29:12] <quiddity>	 REMINDER: about next week's data center migration. There will be NO EDITING for about 30 minutes on both Tuesday, 19 April at 14:00 UTC and Thursday, 21 April at 14:00 UTC. This will affect ALL of the wikis. You will be able to read the wikis and to log in, but you will not be able to edit. Read https://meta.wikimedia.org/wiki/Tech/Server_switch_2016 for more information. Please share this news.
[21:29:13] <quiddity>	 If you find unexpected problems (e.g., can't read for more than a few minutes), then please report those in #wikimedia-tech, which will be monitored by members of the Operations team. #wikimedia-operations will be moderated until the switch is finished and everything is stable again.