voonhous opened a new pull request, #20649: URL: https://github.com/apache/hudi/pull/20649
### Describe the issue this Pull Request addresses Closes #20628. Part of #20625. Stacked on the benchmark PR for #20626; only the top commit is new. Every merge-path split built its own meta client and resolved the table schema on the worker, reading `hoodie.properties` once and the timeline twice per split. ### Summary and Changelog - `HudiTableState`: table config plus, for table versions below 8, the committed instants; captured once on the coordinator. - `HudiTableHandle` carries it together with the lazily resolved schema for MOR tables. - `HudiPageSourceProvider` passes both to the file group reader and no longer creates a meta client per split. - The benchmark's `schema_in_handle` variant is removed, since it is now the only behavior. Local benchmark, scan `_rt`, 48 partitions, default injected latency: | | before | after | |---|---|---| | `hoodie.properties` reads | 50 | 2 | | timeline reads | 97 | 2 | | total requests | 723 | 533 | | wall | 4.22 s | 3.59 s | ### Impact Fewer storage requests per MOR split. Two behavior changes: the schema is resolved once per query on the coordinator rather than per split, and `hoodie.table.create.schema` is not shipped in the handle. ### Risk Level medium. The handle grows by the table config. Covered by handle serialization tests, a schema-change test, and v6/v8 snapshot read file-operation tests. ### Documentation Update none ### Contributor's checklist - [x] Read through [contributor's guide](https://hudi.apache.org/contribute/how-to-contribute) - [x] Enough context is provided in the sections above - [x] Adequate tests were added if applicable -- 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]
