As a WMDE staff working on Wikidata and Wikimedia projects integration I want to be able to see the influence of our work on the performance of Recent Changes, so that I can understand the impact of our work and take necessary measures when needed.
Wikidata changes are likely to result in Recent Changes entries on wikis that make use of Wikidata data in their articles.
Recent Changes queries might become unnecessarily slow if there’s a high number of Wikidata-related rows that are not relevant for the most of Recent Changes queries.
A suggested (by @Ladsgroup ) rule of thumb/heuristic to distinguish between “intended” and “not intended” state seems to be to look at the ratio of entries in Recent Changes table at a given point in time: number of number entries related to Wikidata vs entries related to Mediawiki (article) edits. There’s no absolute “good” figure but general expected trend is to keep the ratio low. i.e. low ratio is intended, high ratio indicates there’s likely a high number of Recent Changes “noise” that’s not relevant for most/any editors making use of Recent Changes
Just observing the ratio of Recent Change entries per wiki would only provide a view on part of the "truth", it would describe the content of Recent Change table but itself it wouldn't express anything on how the database query performance is affected. Therefore that kind of metric should be observed together with a relevant metric signifying Recent Changes query performance/slowness.
To be defined further/refined
- what metrics exactly should be observed (in particular, what database query metric to look at)
- how to observe those (e.g. a dashboard)