I think it is, just not as sensitively. The sampled stats are dependent on the number of samples limit (default=2) and the sample age limit (default=30s iirc). As long as neither of these limits are reached during the test, these stats will be 100% accurate.
Without looking deeply into the implementation, I would think that the rate, on the other hand, would depend on exactly how many seconds elapse during the measurement period. Regardless, you raise a good point, and I'll consider how I can make the test more deterministic. [ Full content available at: https://github.com/apache/kafka/pull/5742 ] This message was relayed via gitbox.apache.org for [email protected]
