SEZ9 commented on issue #12267:
URL: https://github.com/apache/seatunnel/issues/12267#issuecomment-5724332088
Thanks @1362227089, the minimal config and DDL are exactly what was needed
to move the `Duplicate key` investigation forward. A few follow-ups on the same
point:
1. **Replica-identity / catalog result as text, please.** The screenshot is
hard to work with. Could you paste the query output as text, and extend it
slightly so we can see whether the catalog itself returns `test_a` more than
once:
```sql
SELECT oid, relname, relkind, relreplident, relnamespace,
relnamespace::regnamespace
FROM pg_class WHERE relname = 'test_a';
SELECT oid, nspname FROM pg_namespace WHERE nspname = 'testdb';
```
If either query returns more than one row (e.g. a second schema entry, a
partition, or another relkind with the same name), that would be a plausible
reason two `CREATE` records for `highgo.testdb.test_a` end up in
`PostgresIncrementalSource.tableChanges()`.
2. **Config key spelling.** In the block you pasted, `schema‑changes.enabled
= false` appears to contain a non-ASCII hyphen rather than a regular `-`. Could
you confirm whether that is only a copy/paste artifact or is actually present
in the job file? We'd like to rule it out before relying on this config as the
reproduction baseline.
3. **Startup mode.** The block has no startup mode set, so I'll assume the
default. Please confirm that is intentional.
Once we have the text output from (1), we can confirm whether the duplicate
comes from the HighGo catalog or from the connector's table-discovery path, and
then design the regression that feeds duplicate table-change records through
this path (deterministic dedup for identical definitions, descriptive failure
for divergent ones). As before, please keep `REPLICA IDENTITY`, port and driver
unchanged for this run so the results stay comparable.
<!-- streview-comment:1133 -->
--
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]
