SEZ9 commented on PR #11206:
URL: https://github.com/apache/seatunnel/pull/11206#issuecomment-6888625262
Thanks for the update on `da13705e6b08fde27653138a4cb0183563c69726`. The
lifecycle fix for the empty-wrapper case (registering the guarded parent action
at construction so the engine error-handler wrapper isn't replaced later)
sounds like the right shape, and good to see the two-replica regression plus
the MySQL wildcard test running against two sink replicas.
A few things I still need to close out the earlier review points:
1. **Partial registration / quarantine (F5)** – Could you describe what the
post-barrier quarantine does with rows for a table whose writer creation
failed? Specifically: is creation retried, is it idempotent on retry, and can
quarantined rows ever be silently dropped rather than failing the job or being
replayed once the writer exists?
2. **Unsupported sink path (F2)** – With `supportsNewlyCreatedTable()`
defaulting to `false`, what happens when a runtime `CreateTableEvent` reaches a
sink that doesn't support it? I'd like that behavior stated in the interface
contract and covered by a test, so we don't end up with silent row drops or a
null-writer NPE in the multi-table dispatch.
3. **Checkpoint compatibility (F6)** – Since runtime-registered tables are
now persisted in `MultiTableState`, please confirm how restore from a
pre-change checkpoint behaves, and what happens on version rollback.
4. **Default of `scan.newly-added-table.enabled` (F3)** – Is there a reason
to ship this as `true`? Changing restore semantics for existing wildcard MySQL
CDC jobs on upgrade is something I'd rather have opt-in, or at least called out
explicitly in the docs.
5. **Identifier handling (F1, F4, F8)** – Please confirm the identifier
allowlist is enforced in code (not just documented) before names reach
generated sink SQL / the `${database_name}_${table_name}` template, and that
the sink only creates tables within the configured target scope. Also note the
documented regex `[A-Za-z_][A-Za-z0-9_$]*` rejects legal MySQL names starting
with a digit.
6. **JDBC sink docs (F7)** – The `exactly_once = false` / `generate_sink_sql
= true` constraints still need to land in the JDBC sink documentation, not only
in the MySQL-CDC source page.
Since no local functional tests were run, let's get the applicable GitHub CI
running on this head and look at the results before another pass. Happy to
re-review once the above is addressed.
<!-- streview-comment:2360 -->
--
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]