How Kafka handles partition overflow?

Question

I'm getting started with Kafka and trying to understand exactly what would happen in a case when a partition of a topic in a Kafka cluster fills up beyond its limit. I understand that a partition resides in the same node but different partitions reside in different nodes. If I use a custom partition logic in the Kafka producer where a certain key always goes to a certain partition, then what happens when that partition becomes full? Are messages with this key sent to a random partition?

Answer 1

The partition is just cleaned and all the messages are deleted so that new messages can be sent to the same partition. The "retention" policy is handled by all the "log.retention.*" broker parameters.