Summary

Mongolian is more complex than expected. Good read on https://threadreaderapp.com/thread/1407782738413817858.html if you want to follow the kind of challenges we have supporting languages in Search!

Work on metric collection for Search SLO is starting, with already a few metrics collected. We need to get up to speed on Vue.

Search update pipeline is moving forward. We're not confident that we will have a deployment strategy ready by the time the code is complete, please keep your fingers crossed. We found interesting complexity in how we handle search for files: Mediawiki automatically integrates with files on Commons, so when the same file is present both on the local wiki and on commons, strange things might happen. The logic in the current update pipeline is broken in various ways. This seems like a rare enough issue that we will probably just drop the feature from the new pipeline, after validating the rarity of the use case. The project manager for Commons has been notified.

What we've accomplished

Finalize Search SLOs

A number of metrics are collected for the Search SLOs - https://phabricator.wikimedia.org/T335499

Create project plan for WDQS graph splitting

Analysis of the split using simple instanceof rules done - https://phabricator.wikimedia.org/T342111

Operations

Migration of WCQS to Debian Bullseye completed, some more servers need to be migrated for WDQS - https://phabricator.wikimedia.org/T343124

Misc

Our integration tests have flagged a regression in Mediawiki Core related to externallinks. While those tests are a pain to maintain, it seems that they are useful, and not only for our team! - https://gerrit.wikimedia.org/r/c/mediawiki/core/+/945643
Maintenance on our integration tests to unblock upgrading to more recent nodejs version - https://phabricator.wikimedia.org/T300793