Stephen0421 opened a new pull request, #9417: URL: https://github.com/apache/paimon/pull/9417
### Purpose Fix bucket-aware state recovery for the multi-table CDC sink. Previously, restored writer state was not filtered by the same channel ownership rule used for record routing. This could assign restored bucket state to an incorrect subtask after rescaling. This change: - Uses the same database, table, partition, and bucket routing rule for both records and restored state. - Adds the required internal shuffle for the database-aware multi-table CDC sink. - Preserves the legacy constructor behavior and historical Flink topology compatibility. - Waits briefly for newly created tables to become visible before computing their channels. - Fails after a bounded 5-second table lookup retry period instead of using a temporary, potentially incorrect channel. ### Tests - Added tests verifying that record routing and restored-state ownership use the same channel. - Added rescaling coverage for restoring state from parallelism 1 to 3. - Added compatibility coverage for legacy union-state restoration and sink topology. - Added coverage for delayed table visibility and cross-database record validation. -- 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]
