dybyte commented on issue #11752: URL: https://github.com/apache/seatunnel/issues/11752#issuecomment-5292585780
> hi [@dybyte](https://github.com/dybyte), I have a couple of questions about the Flink implementation. > > In [#9867](https://github.com/apache/seatunnel/pull/9867), [#10107](https://github.com/apache/seatunnel/pull/10107), I eventually relied on Flink checkpoints to guarantee flushing of records using the old schema, instead of introducing a dedicated flush API/event. > > do you plan to reuse the checkpoint boundary on Flink, or introduce a separate coordination/barrier mechanism? > > Also, for the recovery case where the external DDL has already succeeded but only part of the parallel writers have refreshed their local schema, where do you plan to persist the authoritative evolved CatalogTable? Would this be maintained in operator/coordinator state and replayed to all writers after recovery? > > How will this design interact with the existing Flink schema evolution path? Thanks for the questions. For Flink, I plan to reuse the existing checkpoint boundary instead of introducing a separate flush API. The evolved schema will be maintained in Flink state so it can be restored and propagated to all writers after recovery. The existing schema evolution path will remain unchanged for sinks that do not opt into the new coordinated capability. -- 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]
