Thanks Jose for your feedback. I have updated the KIP accordingly. "How are you planning to implement this? Kafka can't control how often the metric gets measured and we don't want to schedule an event every 30 seconds to measure a metric."
My plan is to use the existing TimeRatio implementation, which is why I specified the metric type as TimeRatio in the KIP. The mention of "30 seconds" in the original description was an error on my part. I was initially confusing the measurement interval with the sampling frequency, but after re-reading that part, I have a clearer understanding. "How about just calling it AvgIdleRatio?" I think it's better to avoid using "Avg" because the metric is not a true average, which could be misleading. We have a similar confusion with "poll-idle-ratio-avg" metric. Although its description was updated in this PR <https://github.com/apache/kafka/pull/13207/files#diff-c332d2f026c35319f91aa0fefbb449d0721af93eb44578a701058ee990908947R139>, the name itself still causes confusion. I have changed the name to ThreadIdleRatio. Let me know what you think. Thanks, Mahsa On Fri, Aug 15, 2025 at 11:32 AM José Armando García Sancio <jsan...@confluent.io.invalid> wrote: > Hi Mahsa, > > I see that you are proposing to add the metrics > ControllerThreadIdleRatio to the > kafka.controller:type=ControllerEventManager group. The word > controller is already part of the group and namespace. How about just > calling it AvgIdleRatio? That better matches the name used by the > kraft actor (poll-idle-ratio-avg) and the name used by the request > handler (RequestHandlerAvgIdlePercent). > > Thanks, > -- > -José >
