Discovery/Status updates/2017-10-16

This is the weekly update for the week starting 2017-10-16

Highlights

Wikidata Query Service had a partial outage on 18 Oct 2017: a single client started to run an unusually high number of queries on WDQS and the overload was not prevented by our current throttling nor was it automatically detected and isolated—detailed information is here: https://lists.wikimedia.org/pipermail/wikidata/2017-October/011280.html
Discovery and Legal chatted on Oct 18 about supporting only CC0 data in the .map data namespace on Commons, in response to a community conversation; a detailed response will be prepared and released next week. (task T154071)

We added more namespaces to the search results for Bengali Wikisource (Special:Search—add) (task T178041)
David fixed an issue where users couldn't click on svwikisource search results when running a crosswiki search on svwiki; this will be in production during the week of Oct 24th (task T177913)
David was also working on finishing up deployment-prep elasticsearch cluster to 5.5.x (task T178413)
David launched the A/B test for testing the relaxing of the retrieval query filter using two different configurations (the test is expected to last a week) (task T177502)
We wrapped up a series of adding content namespaces to wikis that were accidentally excluded from the default results of Special:Search (thanks to TheDJ for discovering this bug and to David for his quick fixing of the issues once the community consultation was completed by Chris K) (task T170473)
Erik deployed a fix for when the rebuildtextindex fails when searchindex on InnoDB (no support for MyISAM) and it went into production on Oct 19 (task T177477)

Mikhail finished up tasks related to establishing the baseline of metrics for SDoC—looking at querying databases (task T177356)

The Wikipedia.org portal page stats (task T128546) and translations (task T142582) were updated on Oct 19, 2017

MaxSem and Gehel fixed a weird bug where tileshell wasn't exiting after after being notified of updates, effectively stopping the OSM data in production from being regenerated; data updates are now back and running once a day (task T177389)

--