To be able to better judge the impact of timeout errors, it would be incredibly useful to aggregate them by their source, i.e. during which operation they occur. While looking at URLs in the Wikibase Timeout dashboard it seems those timeouts occur predominately during API or special page requests. Therefore it would helpful to visualize these in a table that could look like the following:
Problem: The information on which entry-point a timeout happens is encoded in urls like /w/api.php?action=wbeditentities&ids=Q1032%7CQ766%7CQ863%7C.... and needs to be parsed out, probably with a regular expression or some script.
- Investigate if it is possible to parse the URL in such a way into a field in logstash