We've had 2 recent incidents where high edit rates on wikidata lead to overload of the public WDQS endpoint. We should at least investigate rate limiting the edits consumed by wdqs-updater to match what the different clusters are able to consume without too much impact on the read traffic. Note that this will cause increase update lag, but that's probably a better solution than a complete loss of service.
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
Yes, right now adding the query service lag to max lag seems like the first sensible step.