SEZ9 commented on issue #12983: URL: https://github.com/apache/seatunnel/issues/12983#issuecomment-6407797899
Thanks for filing this follow-up. To scope it, could you confirm which schema-dependent state each adapter owns on restore: for Spark, whether InternalRowCollector's InternalRowConverter must be rebuilt from the restored CatalogTable list before the first restored record; and for Flink, whether FlinkRowCollector needs any per-table state at all or whether downstream type info is the actual concern? Suggest splitting into two independent PRs (Flink, Spark) each with a checkpoint-restore e2e test: run a CDC job with schema-change enabled, apply a DDL change, take a checkpoint/savepoint, restore, and assert the first restored records are emitted with the evolved schema. Without such a test, the no-op inheritance alone doesn't prove a runtime defect, as noted. Note for contributors: please keep this scoped to restoreSchema on the translation collectors; the broader Flink checkpointed schema-evolution protocol belongs in #11958 and should not be folded in here. <!-- streview-comment:2046 --> -- 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]
