Understand Kafka as a distributed, partitioned, replicated commit log — not a traditional queue.

Control which partition a record lands in — the lever that governs ordering and load balance.

Scale consumption horizontally while each message is processed once per group.

Survive broker failures without losing committed data.

Choose between at-most-once, at-least-once, and exactly-once — and know what each really costs.

Decide how long data lives — by time/size, or by keeping only the latest value per key.

Know exactly where Kafka promises order — and where it does not.

Let producers and consumers evolve message formats independently without breaking each other.

Recognise the workloads Kafka is built for — and the ones it is wrong for.

Apache Kafka