The refreshLInkRecommendations.php maintenance script tracks detailed statistics of its run to prometheus, and also has the capability of adding a lot of data to the logs of the run that are stored on the server. The latter contributed to a recent issue of a maintenance host running out of space (T392834) and subsequently its verbose logging was disabled: mediawiki: Make refreshLinkRecommendations job less verbose.
This task is about reconsidering these aspects:
- What data do we want to track in Prometheus, and how do we want to display it? Do we need to review it regularly or is it for on-demand debugging?
- What data do we want to be present in the logs at all times for debugging, and which data should be possible to enable on demand (for debugging or development)?
- note that we could have multiple levels of verbosity, like we added for the FixLinkRecommendationData maintenance script