srdo commented on PR #17139: URL: https://github.com/apache/kafka/pull/17139#issuecomment-2562702398
I've implemented the above idea in the latest commit, as input to the discussion. It's not quite mergeable, but it should be good enough to show that the idea can work. Rather than doing approximation like TimeRatio, it instead divides time into fixed-size windows, and submits one exact sample to the metric per window. As far as I can tell, the benefits over TimeRatio is that this metric is exact and doesn't have any edge cases related to 0-wide polling intervals. The drawback is that the metric will be slightly behind, since the measured time within a sampling window is only known once the window closes. For example, if you have a sampling window size of 10 seconds, the metric will be showing values for polls between 0 and 10 seconds in the past, depending on how far through the window you are. If we decide that we'd like to go with this type of metric over the TimeRatio approximation, I'd be happy to replace TimeRatio with the new fixed-window version, and adapt the tests as necessary. -- 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: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
