zentol commented on PR #17834: URL: https://github.com/apache/flink/pull/17834#issuecomment-1397207313
So gauges are a bit annoying. Ideally we would only have Number gauges because really nothing else is properly supported by a majority of systems. Boolean gauges are usually a mistake from the get-go as well. Lets take the `isBackpressured` metric. This metric tells you whether the task is back-pressured right now at this very moment. That's a terrible metric to make any decision, and you should rather use `backPressuredTimeMsPerSecond` because it's not susceptible to bad luck. `isBackpressured` is only accurate if you are either 100% or 0% back-pressured; for everything in-between it's quite inaccurate (especially since the sampling interval is the reporting interval, aka typically in the order of seconds) (edited) That's why the PR didn't receive any attention. In a way it'd only enable users to rely on bad metrics. Sure, consistency across reporters isn't a bad argument, but this consistency should still provide some real value. Mapping booleans to ints isn't necessarily sound as well, because aggregating them isn't obvious. If we really wanted to supported gauges we'd ideally map them to a distribution I guess. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@flink.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
