SEZ9 commented on PR #11206:
URL: https://github.com/apache/seatunnel/pull/11206#issuecomment-7025820722
Thanks for the update. The quarantine lock boundary change does not by
itself close the points from the previous review, so they are still open.
Please address them together in the next commit:
**Code**
- **Single initial table on Zeta:** the runtime `CreateTableEvent` is
registered only when the source produced type is a `MultipleRowType`. A job
that starts with exactly one matched table therefore never learns the new
table's row type.
- **JDBC capability check:** `JdbcSinkFactory.supportsNewlyCreatedTable()`
returns `true` regardless of config. An exactly-once or custom-`query` sink is
rejected only after a writer is built, and under `CONTINUE_OTHER_TABLES` the
table is quarantined instead of failing the job. Please make the check
configuration-aware.
- **Retries at the schema-change barrier:** writer creation is retried with
`TimeUnit.SECONDS.sleep` while every queue worker is parked. One failing new
table then stalls all tables and checkpoints for the whole retry window.
- **DROP followed by re-CREATE:** a second `CREATE TABLE` for an already
registered path is ignored. A re-CREATE with a different schema leaves stale
converters and a stale sink schema.
- **Runtime admission bound:** there is no upper bound and no removal path,
so sink writers and checkpoint state can grow without limit.
**Docs (`MySQL-CDC.md`)**
- Explain how the runtime option relates to `schema-changes.enabled`, and
that it overrides a user-supplied `debezium { include.schema.changes }`.
- Document that `scan.binlog.newly-added-table.enabled` is rejected at
startup when `table-names` is configured, in both the option row and the "Newly
Added Tables" section.
- State which engines support runtime table registration, since the wiring
is engine-specific.
When you push, please add a short note on how each item was handled. For the
single-initial-table and DROP/re-CREATE cases, please also say which test
covers the behaviour. If you plan to handle any of these differently, such as
the admission bound or the retry strategy, describe the approach here so we can
agree on it first.
<!-- streview-comment:2474 -->
--
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]