[00:24:14] <stevemunene>	 o/ volans missed your ping earlier. the ticket should be T390171. The drive replacement is done and we are readding the two hosts an-worker1177 and an-worker1156 to the cluster 
[00:24:16] <stashbot>	 T390171: Upgrade an-worker hard drives from 4TB to 8TB (group 4 - rack F3) - https://phabricator.wikimedia.org/T390171
[06:30:59] <volans>	 stevemunene: ack, does puppet still need to be disabled?
[08:54:18] <stevemunene>	 volans: The hosts should be done onboarding today with puppet reenabled
[10:16:28] <dcausse>	 lunch
[14:02:15] <gehel>	 ebernhardson: would you be ok with participating in the interview for the Wikidata Platform tech lead? Amy is asking for your participation.
[14:10:49] <gehel>	 pfischer: for next DPE staff meeting (June 2), would you like to demo the work done on maven build process?
[14:11:53] <pfischer>	 gehel: Yes, should be done and stable by then.
[14:12:04] <gehel>	 Cool! I'll add it on the list
[14:16:15] <pfischer>	 gehel: Sonatype responded. If you approve, we can migrate the org.wikimedia namespace to their new central platform.
[14:19:40] <MichaelG_WMF>	 we seem to have an issue with search for link- and image-recommendations: https://phabricator.wikimedia.org/T394416
[14:20:06] <MichaelG_WMF>	 In particular: `hasrecommendation:link` seems to return 0 results everywhere
[14:22:46] <MichaelG_WMF>	 Are you aware of anything on your end that might cause this?
[14:23:15] <dcausse>	 MichaelG_WMF: looking
[14:23:22] <MichaelG_WMF>	 thank you!
[14:25:04] <dcausse>	 hm, the keyword is not even interpreted...
[14:37:27] <dcausse>	 MichaelG_WMF: I suspect it's caused by https://gerrit.wikimedia.org/r/c/mediawiki/extensions/CirrusSearch/+/1143958, working on a fix but currently in meetings so could take time so feel to revert 
[14:37:48] <stephanebisson>	 Hey search team, as of this morning I'm noticing that our usual "articletopic:architecture" and "articlecountry:ita" queries don't work anymore but similar queries "topic:architecture" and "country:ita" seem to work... It this intentional?
[14:37:52] <MichaelG_WMF>	 looking
[14:38:14] <dcausse>	 I think we should revert
[14:38:38] <dcausse>	 stephanebisson: yes same cause: T394416
[14:38:38] <stashbot>	 T394416: Link and Image recommendation tasks missing on all production wikis - https://phabricator.wikimedia.org/T394416
[14:40:12] <dcausse>	 reverted at https://gerrit.wikimedia.org/r/c/mediawiki/extensions/CirrusSearch/+/1146642
[14:42:20] <dcausse>	 backport https://gerrit.wikimedia.org/r/c/mediawiki/extensions/CirrusSearch/+/1146643
[15:00:21] <ebernhardson>	 \o
[15:00:37] <ebernhardson>	 apparently need some new integration tests in cindy
[15:00:49] <dcausse>	 yes...
[15:00:51] <dcausse>	 o/
[15:04:31] <ebernhardson>	 plausibly, this is what desktop users did may 10th: https://phabricator.wikimedia.org/F60020160
[15:04:37] <ebernhardson>	 not 100% sure it's correct, but maybe mostly
[15:06:31] <ebernhardson>	 i also have a session normalized variant, but the normalization is wrong (the flows don't all add up anymore)
[15:38:41] <ebernhardson>	 i've seen a few other oddities of rendering, but moving nodes around gets it to do different things. Still have to try loading the results into superset
[15:40:16] <ebernhardson>	 the numbers for session normalized are 1.45M start -> autocomplete,  where page -> serp is 14k, start -> target page is 5.86k,  so pretty small 
[15:40:57] <ebernhardson>	 this does at least suggest fulltext is a significant part of the system, even if autocoplete is generally more successful
[15:43:13] <dcausse>	 yes
[15:43:44] <dcausse>	 I'd be for dropping the tiny links that are non-intuitive
[15:44:00] <dcausse>	 like session start -> session end
[15:44:15] <ebernhardson>	 yea thats probably reasonable,  those are just kinda artifacts of the data colletion method
[15:58:05] <dcausse>	 meh... thought there was CI issues but turns out it's my fault, pretty hard to find what's wrong in https://gitlab.wikimedia.org/repos/data-engineering/airflow-dags/-/jobs/511643
[15:58:49] <ebernhardson>	 wow thats an awkward output
[15:59:56] <dcausse>	 ERROR: /builds/repos/data-engineering/airflow-dags/tests/conftest.py Imports are incorrectly sorted and/or formatted which is nice it tells me the file but no clue what's wrong there...
[16:02:45] <ebernhardson>	 dcausse: i think you just need to run isort
[18:34:34] <ebernhardson>	 i'm not sure how actionable it is, but i was curious and tried to classify all the search paths people took in april: https://phabricator.wikimedia.org/T394450
[18:42:08] <dcausse>	 direct success there's no search at all?
[18:42:35] <ebernhardson>	 dcausse: it is, i just dont check which one. it means two steps and the last was a TargetPage, its probably AutoComplete -> TargetPage, but could be SERP -> TargetPage
[18:42:48] <dcausse>	 ok