SEZ9 commented on issue #11402: URL: https://github.com/apache/seatunnel/issues/11402#issuecomment-5149901166
This scope touches API, engine coordination, connectors, and restore semantics, so it should be captured as a formal design/SEIP document with a state-transition diagram for the schema change lifecycle (observed -> barrier/flush -> sink apply -> checkpointed) rather than resolved only in issue comments. A key point to settle early: sink-side schema mutations (e.g. ALTER TABLE) are externally visible and often non-transactional, so the contract must require idempotent replay of SchemaChangeEvents after restore when the DDL succeeded but the checkpoint did not complete. Proposals should state per-connector how idempotency is achieved (e.g. IF NOT EXISTS semantics, schema comparison before apply). Suggest defining the minimum contract as: (1) per-table total order of SchemaChangeEvents relative to data records, (2) mandatory flush of old-schema records before sink apply, (3) independent progress across tables allowed, (4) fail-fast for any component that does not declare schema-change support. A shared E2E test matrix covering the failure points listed (before/during/after apply, during restore) should accompany the final design. <!-- streview-comment:22 --> -- 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]
