[06:37:11] <wikibugs>	 10Scoring-platform-team, 10DBA, 10MediaWiki-Database, 10Blocked-on-schema-change, and 2 others: Schema change for rc_this_oldid index - https://phabricator.wikimedia.org/T202167 (10Marostegui)
[08:21:26] <wikibugs>	 10Scoring-platform-team, 10DBA, 10MediaWiki-Database, 10Blocked-on-schema-change, and 2 others: Schema change for rc_this_oldid index - https://phabricator.wikimedia.org/T202167 (10Marostegui)
[08:21:30] <wikibugs>	 10Scoring-platform-team, 10DBA, 10MediaWiki-Database, 10Blocked-on-schema-change, and 2 others: Schema change for rc_this_oldid index - https://phabricator.wikimedia.org/T202167 (10Marostegui) s7 eqiad progress  [] labsdb1011 [] labsdb1010 [] labsdb1009 [] dbstore1002 [] db1125 [] db1116 [] db1101 [] db109...
[13:28:43] <Amir1>	 o/
[14:02:50] <wikibugs>	 10Scoring-platform-team (Current): Prototype a bot framework that utilizes newcomerquality - https://phabricator.wikimedia.org/T211434 (10notconfusing) updates: * using a local mysql database * using a (dev with ssh tunnel)  still todo: [ ] exponential backoff for each stage in hostbot_ai.py - https://github.com...
[14:18:53] <icinga2-wm>	 PROBLEM - puppet on ORES-worker02.experimental is CRITICAL: CRITICAL: Catalog fetch fail. Either compilation failed or puppetmaster has issues
[14:38:47] <thresholdT>	 Hi everyone, I'm Tom from Freie Universität and I'm looking into User Centered Threshold Optimization for ORES. Right now I'm trying to understand all the metrics/parameters for the itemquality model. Is there some kind of documentation for the metrics that are shown when I call the API with https://ores.wikimedia.org/v3/scores/wikidatawiki/?models=itemquality&model_info=statistics ? I don't fully understand filter_rate, mat
[14:38:59] <halfak>	 Hi Tom!
[14:41:09] <halfak>	 filter_rate and match_rate are names I made up.  I don't know if there's a good standard term for them.  They refer to the proportion of observations matched/not-matched (filtered). 
[14:41:32] <halfak>	 E.g. if you set a threshold of zero, then match_rate should be 1.0 (or 100%)
[14:42:24] <halfak>	 If you set a threshold to 1.  Or actually, let's say 1.01 (impossibly high), the match_rate will be zero and the filter_rate will be 1.0 or 100%
[14:43:38] <halfak>	 Otherwise, the rest of the terminology should be standard.  The only hiccup is the "recall" vs "!recall".  Any metric named "!<metric>" is the same metric for the negative class. 
[14:44:20] <halfak>	 So "recall" TP/TP+FN.  "!recall" is TN/TN+FP
[14:44:53] <icinga2-wm>	 RECOVERY - puppet on ORES-worker02.experimental is OK: OK: Puppet is currently enabled, last run 38 seconds ago with 0 failures
[14:46:10] <halfak>	 Let's say you wanted to find all of the items that are not "E" class.  Then you would be looking at the "!recall" for "E" class. 
[14:47:13] <halfak>	 thresholdT, ^ 
[14:47:31] <thresholdT>	 Oh alright, thank you. Just to be sure (and sorry if I'm a bit slow on this one): would it be correct to say that match_rate = TP/total?
[14:48:18] <halfak>	 Right! 
[14:48:25] <halfak>	 Oh wait. 
[14:48:26] <halfak>	 No
[14:48:31] <halfak>	 TP+FP/Total
[14:49:07] <thresholdT>	 Ah, ok! 
[14:50:19] <thresholdT>	 And in that notation, what would the filter_rate be?
[15:02:21] <wm-bot4>	 Technical Advice IRC meeting starting in 60 minutes in channel #wikimedia-tech, hosts: @Thiemo_WMDE & @chiborg - all questions welcome, more infos: https://www.mediawiki.org/wiki/Technical_Advice_IRC_Meeting
[15:24:26] <halfak>	 thresholdT, sorry I missed your last message.  Filter_rate == 1 - Match_rate
[15:24:53] <halfak>	 So TN+FN/Total
[15:25:07] <thresholdT>	 No worries, ok cool! Thanks a lot!
[15:28:18] <icinga-wm>	 PROBLEM - ORES worker production on ores.wikimedia.org is CRITICAL: HTTP CRITICAL: HTTP/1.1 500 INTERNAL SERVER ERROR - 6490 bytes in 5.050 second response time
[15:28:30] <halfak>	 uh oh
[15:28:52] <halfak>	 Working for me.  Connecting from MN. 
[15:29:02] <halfak>	 Checking grafana
[15:30:38] <icinga-wm>	 RECOVERY - ORES worker production on ores.wikimedia.org is OK: HTTP OK: HTTP/1.1 200 OK - 923 bytes in 0.130 second response time
[15:31:45] <halfak>	 Grafana changed and many of our plots have "no data points" WTF
[15:33:02] <halfak>	 A refresh got it. 
[15:35:30] <halfak>	 I don't see anything concerning. 
[15:35:35] <halfak>	 Maybe it was a fluke?
[15:36:10] <halfak>	 I don't like the new "response time" graph.  It can fluctuate depending on how many batch requests we are getting.  
[15:36:12] <halfak>	 https://grafana.wikimedia.org/d/000000255/ores?refresh=1m&panelId=43&fullscreen&orgId=1
[15:42:45] <Amir1>	 halfak: oh I have seen this, just change time span, it would help in some cases
[15:43:29] <Amir1>	 btw. I will be working mostly on wikidata stuff today as I was working on ores yesterday (the presentation)
[15:45:05] <halfak>	 How did the presentation go? 
[15:52:18] <wm-bot4>	 Technical Advice IRC meeting starting in 10 minutes in channel #wikimedia-tech, hosts: @Thiemo_WMDE & @chiborg - all questions welcome, more infos: https://www.mediawiki.org/wiki/Technical_Advice_IRC_Meeting
[15:53:35] <Amir1>	 halfak: it was good, specially the workshop
[16:06:11] <halfak>	 Cool.  Any notes or outcomes you can point to? 
[16:08:24] * halfak fights with his monitor & docking station again for the 11 millionth time. 
[16:11:03] <halfak>	 ahhh!
[16:31:56] <Amir1>	 someone should fill my spot for SOS
[16:59:28] <wikibugs>	 10ORES, 10Scoring-platform-team, 10Documentation: Rework "get support" page for ORES - https://phabricator.wikimedia.org/T211794 (10awight)
[17:04:49] <halfak>	 Amir1, I didn't get the ping, so I guess that didn't happen. 
[17:04:59] <halfak>	 Amir1, still joining for our staff meeting? 
[17:05:13] <Amir1>	 halfak: no, doing wikidata stuff 
[17:05:32] <Amir1>	 Nothing much to discuss either. We were all on offsite :P
[17:05:37] <halfak>	 Current work. 
[18:10:59] <awight>	 Just grepping here, https://etherpad.wikimedia.org/p/scoring_fy19Q2_offsite-todos
[18:11:51] <harej>	 good call, thank you
[18:17:17] <awight>	 :)
[18:17:50] <awight>	 What collaborative drawing thing would you recommend?
[18:18:28] <harej>	 Drawing?
[18:20:50] <awight>	 https://docs.google.com/drawings/d/1hkPkDSXKe7tIS6SnVlDxpSeNnwbzFsaSYEGFk9ax9dE/edit
[18:21:40] <harej>	 Interesting.
[18:35:16] <awight>	 rrgh, rebooting due to OS update jacking
[18:49:05] <awight>	 https://docs.google.com/drawings/d/1ybck5xyfNVwfKw55T4WDX246wc0C213t6bpEge7IuTE/edit?usp=sharing
[19:10:01] <awight>	 harej: What do you think about readers vs. consumers, in the logic model?
[19:25:13] <awight>	 This is plotting curation generally, but I'm struggling to say how Jade changes the steps here.
[19:41:19] <wikibugs>	 10Jade, 10Scoring-platform-team: Explore alternative names for Jade data - https://phabricator.wikimedia.org/T200365 (10awight)
[19:43:20] * halfak is watching the research showcase.  "Why we read Wikipedia"
[19:43:25] <halfak>	 See #wikimedia-research
[19:52:58] <awight>	 halfak: Thanks for announcing!
[20:59:58] <awight>	 Thing missing from the logic model is the whole point of why I started making it: what are the "upstream" motivations for our implementation choices.
[21:04:09] <halfak>	 awight, not sure what you're thinking about.  What would an example of an "upstream" motivation be? 
[21:18:11] <awight>	 halfak: this task was in the context of our team discussion https://etherpad.wikimedia.org/p/scoring_fy19Q2_offsite-topic_2
[21:18:41] <halfak>	 Right.  Trying to think of what you mean by "upstream"
[21:18:48] <awight>	 I guess "upstream" is a bad term for what I mean
[21:18:59] <awight>	 We wanted to map the causal links
[21:19:43] <awight>	 so, * what is causing us to make implementation choices, and * what are the desired effects of these choices?
[21:20:30] <awight>	 e.g. for editquality, one cause is that review work overwhelms the curation community.
[21:20:35] <halfak>	 "Robust and productive wiki communities" <-- "Good social dynamics" <-- "Good newcomer socialization" <-- "People have time and energy to spend on newcomers" <-- "Quality control is efficient and easy" <-- "Machines support quality decision processes" <-- "The AI team has resources"
[21:20:42] <awight>	 A desired impact is that the workload is decreased
[21:20:53] <awight>	 :) great, thanks
[21:21:00] <halfak>	 Is that on the right track?
[21:21:04] <awight>	 exactly
[21:21:08] <halfak>	 Got it
[21:21:14] <awight>	 here's my stub, https://docs.google.com/drawings/d/1hkPkDSXKe7tIS6SnVlDxpSeNnwbzFsaSYEGFk9ax9dE/edit
[21:22:30] <awight>	 Grantmakers will be one of our audiences for this diagram.
[21:25:36] <awight>	 Not quite sure what that implies for our choice of scope, i.e. whether we should show the broadest social impacts of our technologies if applied generally.
[21:35:26] <halfak>	 awight, not following your diagram exactly. 
[21:36:39] <awight>	 I'm not particularly good at these.
[21:54:07] <halfak>	 It seems like there are many paths that could lead to an "upsteam" goal. 
[21:54:20] <halfak>	 Maybe we should make exactly one path together. 
[21:54:27] <halfak>	 Is that what you are aiming for?
[21:54:43] * halfak curses at graphite and it's slowness. 
[21:55:07] <halfak>	 I'm trying to make a graph of the rate of requests for various batch sizes. 
[21:55:15] <halfak>	 Been lightly hacking on this in meetings all day. 
[21:55:38] <awight>	 great idea!
[21:55:41] * Hauskatze points that holly oaks grow slower than graphite works
[21:55:53] <halfak>	 lol too true
[21:56:25] <halfak>	 "error 503 backend failed"
[21:56:26] <halfak>	 arg!
[21:57:42] <awight>	 halfak: I think that isolating a few of the paths would be helpful here, for example to highlight the motivations for and potential impact of Jade.
[22:09:19] <halfak>	 https://grafana.wikimedia.org/d/000000255/ores?refresh=1m&panelId=56&fullscreen&edit&orgId=1&from=now-24h&to=now-1m&tab=legend
[22:16:13] <awight>	 oh wow, I think this query killed Grafana 
[22:17:40] <Hauskatze>	 PROBLEM - graphite.wikimedia.org on graphite1004 is CRITICAL: CRITICAL - Socket timeout after 10 seconds
[22:17:49] <Hauskatze>	 you took the service down lol
[22:18:08] <awight>	 I was able to load a different graph though.
[22:18:22] <Hauskatze>	 (there's a recovery message from a few minutes ago though)
[22:19:08] <awight>	 lemme try that first graph again...
[22:29:58] <wikibugs>	 10Jade, 10Scoring-platform-team: Explore alternative names for Jade data - https://phabricator.wikimedia.org/T200365 (10awight) p:05Normal>03High There are several vetoes against "judgment", so we'll block deployment until this can be resolved.  Some thoughts about what the name should communicate: * We're...
[22:31:37] <wikibugs>	 10Jade, 10Scoring-platform-team: Explore alternative names for Jade data - https://phabricator.wikimedia.org/T200365 (10jrbs) >>! In T200365#4818766, @awight wrote: > There are several vetoes against "judgment", so we'll block deployment until this can be resolved. >  > Some thoughts about what the name should...
[23:17:14] <awight>	 gtg for now