DanielLeens commented on issue #12923: URL: https://github.com/apache/seatunnel/issues/12923#issuecomment-6415642387
Thanks @taozhiyi13 for the detailed revision. The new inventory now names the Zeta action graph, `RecordSerializer` row field, CDC checkpoint tables, and Flink/Spark planning boundaries, and the row-level SRID rule is clear. Those address the *inventory* questions from my previous review. I checked current `dev` (`4f5e3293740`): `LogicalVertex.writeData()` serializes the action with `writeObject`, and `RecordSerializer` writes each `SeaTunnelRow` field with `writeObject`, so a new geometry class really does cross both of those boundaries. Two compatibility decisions still need to be explicit before I can support acceptance. First, changing generic MySQL JDBC `GEOMETRY` from the existing `BYTES` exposure to a new `GEOMETRY` type is a user-visible break for jobs that consume the old bytes. Please call it incompatible, give the `ST_AsWKB(column)` migration path in both language docs and `incompatible-changes.md`, and include a before/after job-schema test. Second, “upgrade every Zeta node first” is an operational rule, not yet a failure contract: an old worker cannot deserialize the new row field class. State how a spatial job is rejected or held before dispatch when client/master/worker versions differ, or document a strict full-cluster upgrade boundary and test the actual failure/restore behavior. A late task deserialization error must not be presented as graceful plan-time rejection. This STIP remains **Under Review**. Please add those two decisions and bring the cross-module API/upgrade contract to `[email protected]` for consensus. After that, the proposed API, source, and sink slices can be reviewed against the stated tests. I added the `STIP` and `api` labels for tracking. -- 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]
