I'm exploring how the PG connector works in case of concurrent schema changes and updates and it seems the connector stops to work in the following scenario:
- Set a break point to PostgresSchema#refresh()
- Run this DDL: alter table customers add column middle_name varchar(255);
- Run this DML: update customers set first_name='Mary' where id = 1004; this will hit the break point set above. Wait there.
- Run this DDL: alter table customers add column another_middle_name varchar(255);
- Continue after the break point
There will be no update for the DML statement in the Kafka topic. I reckon the connector runs into some kind of error, it doesn't log anything, though.
I suspect the sequence above poses an issue either way: it seems the schema-update upon processing the DML would read the second DDL which isn't valid yet at this point, though.