wForget opened a new issue, #2761: URL: https://github.com/apache/uniffle/issues/2761
### Code of Conduct - [x] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct) ### Search before asking - [x] I have searched in the [issues](https://github.com/apache/incubator-uniffle/issues?q=is%3Aissue) and found no similar issues. ### What would you like to be improved? We observed severe lock contention when running workloads with a large number of small tasks. Many tasks spend noticeable time creating new shuffle writers, and thread dumps show that these tasks are blocked on the monitor of `RssShuffleManagerBase`. The contention happens in block retry mode when `RssShuffleManagerBase#getRemoteShuffleHandleInfoWithBlockRetry` is called during shuffle writer creation. This method is currently declared as `synchronized`. <img width="1902" height="640" alt="Image" src="https://github.com/user-attachments/assets/2d2845c4-c134-44af-bc4c-c92f3fe087d5"; /> <img width="1914" height="521" alt="Image" src="https://github.com/user-attachments/assets/03056216-5a65-45a8-be66-c836e526e811"; /> `getRemoteShuffleHandleInfoWithBlockRetry` only builds a request, sends an RPC, and deserializes the returned handle. It does not mutate shared state, so the method-level `synchronized` seems unnecessary. ### How should we improve? _No response_ ### Are you willing to submit PR? - [x] Yes I am willing to submit a PR! -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
