SEPURI-SAI-KRISHNA commented on PR #13126: URL: https://github.com/apache/seatunnel/pull/13126#issuecomment-6910736038
@DanielLeens Thanks, both are fixed in `431b67528a`, with Option A. **Issue 1.** The writer no longer continues without the key. If the table is missing from the driver metadata, it calls `session.refreshSchema()` once; in driver 4.14 that refresh runs even when schema metadata is disabled (`MetadataManager.refreshSchema(null, true, true)`). If the key still cannot be read, the constructor throws `CASSANDRA-06` naming the table and closes the session, so a NULL key can no longer reach the server. `nullPrimaryKeyReachesTheServerWithoutTableMetadata` is replaced by: - `primaryKeyMissingFromTheMetadataIsReadAfterARefresh`: the key is found after the refresh, and a NULL key is rejected before any write; - `writerFailsAndClosesTheSessionWhenThePrimaryKeyCannotBeRead`; - `writerClosesTheSessionWhenThePrepareFails`: a failed `prepare` leaves the session open on `dev` too. Removing the refresh, the failure or the close each fails its test, and removing the key check fails the four NULL key tests. I did not add a failed-future test, because the metadata-unavailable path no longer submits a write. Making a failed async write fail the job in general (Option B) is the separate issue promised in the description; I will file it with a reproduction. **Issue 2.** `primaryKeyColumns` and the package-private constructor now have Javadoc: the set is never empty and why a NULL key is rejected before the write, and the writer owns the session: `close()` closes it, and so does the constructor when it fails. The EN and ZH incompatible-changes entry now says that the sink reads the key from the driver schema metadata when it starts and fails if it cannot. Fork CI on this head passes (https://github.com/SEPURI-SAI-KRISHNA/seatunnel/actions/runs/43797522344): `CassandraIT` runs 27 tests with 0 failures on JDK 11 and 17, and the unit test jobs pass on Linux and Windows. -- 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]
