0xffmeta opened a new issue, #6865: URL: https://github.com/apache/iceberg/issues/6865
### Feature Request / Improvement When we use a flink `upsert` mode to generate the v2 format, we found that the table scan part is very slow due to massive delete files need to be loaded into `DeleteFilter`. We checked the `FileScanTask` and ovserved that the delete files might be read again and again for the data files within the same partition, e.g. one partition with several commits: - data1, data2 - data3, data4, data5 with deleteA, deleteB, deleteC - data6 with deleteD When to scan the eariler data file(e.g. data1, data2), the delete files commited later(e.g. deleteA to deleteD) are needed to be loaded into memory, and for each scan(data1 and data2), those delete files will be loaded each time. And this could slow down the reading performance if there are a lot of writers with frequent commit. We can introduce a `LoadingCache` in `DeleteFilter` to cache the delete file content. ### Query engine None -- 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]
