AndrewJSchofield commented on PR #14111: URL: https://github.com/apache/kafka/pull/14111#issuecomment-1699383628
@junrao We have done some performance tests with leadership changes and there does not appear to be any performance degradation. The testing used `kafka-producer-perf-test.sh` connected to an Apache Kafka cluster with 6 brokers running on EC2. It used 1 topic with 100 partitions, `linger.ms=0` and `acks=1`. In order to cause leadership changes, the partitions led by one of the brokers were reassigned to other brokers during the performance test. Then, the latency for the reporting period immediately after the leadership change were captured as follows: * AK without this PR: 1.0 ms avg latency, 125.0 ms max latency. * With this PR: 0.9 ms avg latency, 97.0 ms max latency. Of course, this is tricky, because the whole point of the PR is to make repeated metadata requests slower to reduce broker load when there are many clients. So, the fact that these tests do not show a degradation is a good thing. -- 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