SEZ9 commented on PR #12014:
URL: https://github.com/apache/seatunnel/pull/12014#issuecomment-6887793238

   Thanks for the detailed follow-up on 
`3ee352173a6dd251b700eb1112a0cd59c11687c8`, @davidzollo.
   
   **Looks resolved from the notes**
   - **PR12014-F7 / F8 (docs):** distinguishing job submission from the runtime 
failure on an unsupported engine in both the English and Chinese docs, 
switching to the JDBC key `is_exactly_once`, and linking the filter rows to the 
table-operation section covers what I asked for. I'll confirm the wording when 
I re-read the diff.
   - **PR12014-F4 (restore test):** the E2E flow you describe — FK-induced sink 
TRUNCATE failure, committed rows surviving, fault removed, restore from the 
failed job's completed checkpoint, only the post-TRUNCATE row remaining — is 
exactly the at-least-once path I wanted exercised. Not asserting intermediate 
duplicate visibility for append-only targets is fine; please just make sure the 
docs state that duplicates are possible for non-keyed sinks during replay.
   - **PR12014-F5 (fail-fast on Flink/Spark):** I agree a new cross-connector 
submission validator is out of scope here. With the docs now accurate about 
when the failure happens, I'm okay deferring; please open a follow-up so it 
isn't lost.
   
   **Still open — please address or reply**
   - **PR12014-F1 (failure atomicity / exactly-once):** the FK-failure E2E 
answers the "TRUNCATE can fail" half. What happens when `is_exactly_once` is 
enabled on the JDBC sink and a TRUNCATE arrives? Since TRUNCATE is DDL with an 
implicit commit, I'd expect it to either be rejected up front or explicitly 
documented as breaking XA semantics. Which one is it in this head?
   - **PR12014-F2 (`include.schema.changes` forced on):** no mention in the 
notes. Is enabling table-operations still force-opening Debezium 
`include.schema.changes`, and if so, how do unrelated DDL events get filtered 
before reaching the new deserializer paths?
   - **PR12014-F3 (identifier quoting):** please confirm the sink-side TRUNCATE 
statement is built with the dialect's identifier quoting from the DDL-parsed 
table name rather than raw string concatenation, and point me to the test that 
covers a table name with special characters.
   - **PR12014-F6 (EventType ordinals):** please move the new constant to the 
end of the enum (or confirm it already is) so existing `LIFECYCLE_*` ordinals 
are unchanged.
   
   Once the MySQL CDC and JDBC E2E suites have run on the updated build, ping 
me and I'll do the final pass.
   
   <!-- streview-comment:2354 -->


-- 
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]

Reply via email to