I was going to reply in this topic because it’s related but decided to start a different one because the other one anyway seems closed.
I’m experimenting some issues with the way that records end up in the DLQ.
I seem to see that the only options we have in Kafka to handle the connector errors is [none, all] as seen in the confluent [documentation]( Kafka Connect Sink Configuration Properties for Confluent Platform | Confluent Documentation)
So my problem or what I want to achieve is that only failures during the deserialization (valueConverter) end up in the DLQ. Any error in the Sink stage should make the connector fail.
What I experience so far is that for example if my JDBC Sink is going to a target where there is problems with the tablespace size all the messages end up discarded into the DLQ. I wanted to avoid this and ensure that in those cases the connector would fail.
I also experience similar problem when using for example the S3 Sink when connected to a target that uses for example MinIO or some sort of on-premises object storage solution.
Is there a way to achieve this?
I could not find any reference even if in the linked discussion it talks about “behavior.on.malformed.documents” for a Elasticsearch sink I was wondering what is the solution for JDBC for example.