XiaoHongbo-Hope opened a new issue, #987: URL: https://github.com/apache/paimon-rust/issues/987
### Search before asking - [x] I searched the issues and found no similar issue. ### Paimon version Current `main` (`fc27163`) ### Compute Engine PyPaimon native reader / paimon-rust ### Minimal reproduce step Create or load a table with `scan.manifest.parallelism=8`, then run native scan planning for a snapshot containing multiple manifest files. ### What does not meet your expectations? Native scan planning uses a hard-coded `.buffered(64)` for both incremental manifest-list reads and manifest-file reads. The configured value is ignored, unlike Java Paimon and the Python planner. This can produce avoidable object-store request bursts when many processes plan concurrently, because each process may issue up to 64 concurrent manifest reads even when the table config requests a lower bound. ### Anything else? The native planner should honor `scan.manifest.parallelism` and use the process CPU count when the option is absent, matching the documented Java default. Invalid values such as zero should return a configuration error rather than reaching `StreamExt::buffered(0)`. -- 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]
