DanielLeens commented on issue #10203: URL: https://github.com/apache/seatunnel/issues/10203#issuecomment-6093465002
Classification: A / Connector-V2 CDC restore and dynamic-table lifecycle. @ryanmeowy, thank you for writing down the proposed boundaries and test plan. I rechecked current `dev`, the relevant source path, commits, and open PRs before replying. The proposed attach point already runs during restoration: `IncrementalSource.restoreEnumerator()` discovers the current collections from the configured source, and `restore()` computes `capturedTables - checkpointCapturedTables` before appending newly discovered tables to remaining snapshot work. The MySQL factory already exposes `table_pattern` as the alternative to explicit `table_names`. A new default-off Phase 1 gate at this point would therefore change existing restore behavior rather than add the requested capability. The unresolved correctness work at this boundary is #12567 for #12565. It carries the per-table restore watermarks needed when a job is restored after matching tables are added. That PR is not merged and its Build is still ACTION_REQUIRED, so please do not open a second PR that changes table admission or configuration while the state and exactly-once path is awaiting review. If you would like to help after that work is resolved, please start with a separate design for an actually missing live-runtime discovery behavior, rather than restart or restore discovery: trigger cadence, enumerator ownership, persisted state, split reassignment, checkpoint failure and recovery, duplicate suppression, and the exact MySQL versus OceanBase scope. Existing `table_pattern` restore semantics must remain backward compatible. -- 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]
