junrao commented on pull request #8657: URL: https://github.com/apache/kafka/pull/8657#issuecomment-666753554
@ijuma : I think the proposal is to complete all delayed operations in a separate thread pool. My concerns for that approach are the following: (1) Configuration: How many threads should be used? Should that be dynamically configurable? Should the queue to this thread pool be unbounded or fixed size? If it's the latter, how do we configure the queue size? (2) Monitoring: How do we monitor the utilization of the new thread pool and the potential back pressure it creates (if the queue is fixed size)? (3) Quota: Do we need to account for the resource utilization of this new thread pool per client/user for quota purpose? (4) Debugging: If there is a latency issue, how do we know whether this is due to delays in the new thread pool? So, overall, I feel that we should be careful with adding another thread pool used in the common code path since it adds other complexity. Given that this issue is isolated in GroupCoordinator, I am not sure adding a thread pool is an overall simpler solution. @chia7712 : I saw that you consolidated the commit history. Were there significant changes since the last review? ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
