kirktrue commented on PR #17700: URL: https://github.com/apache/kafka/pull/17700#issuecomment-2578784450
> @kirktrue : Thanks for the experimental results. What's the reason for the high eviction rate at the beginning of all tests? Four warm-up runs are executed before the test, and these warm-up runs create many consumers. When the actual test is started, the pre-test consumers are expired from the session cache since their caches contain very few partitions each. > Could you do some consumer latency test to see if there is any noticeable impact with option 4? Yes, I pulled the average fetch latency metric out of the same test runs as before: `CLASSIC` group protocol, option 1: <img width="1619" alt="CLASSIC, option 1, fetch latency avg" src="https://github.com/user-attachments/assets/69dc3305-f563-4ccb-b354-a06dd0f2c5ce"; /> `CLASSIC` group protocol, option 4: <img width="1616" alt="CLASSIC, option 4, fetch latency avg" src="https://github.com/user-attachments/assets/eb86c009-81eb-4404-8e49-38d08283207b"; /> ------------------- `CONSUMER` group protocol, option 1: <img width="1618" alt="CONSUMER, option 1, fetch latency avg" src="https://github.com/user-attachments/assets/a41d13b2-7c15-4e25-b84d-6032afdd2059"; /> `CONSUMER` group protocol, option 4: <img width="1619" alt="CONSUMER, option 4, fetch latency avg" src="https://github.com/user-attachments/assets/7a55dd1d-502b-4d4b-85f0-704778a1617b"; /> As I see it, there does not appear to be any noticeable difference with respect to the fetch latency with the change. Thanks! -- 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
