3. Kafka Producers: Writing Messages to Kafka

Source: Kafka: The Definitive Guide, 2nd Ed., Ch. 3 Learning goal: the producer API is ~10 lines. Everything that matters is (a) what happens on the internal threads, and (b) how ~15 configs interact to define durability, ordering, latency, and throughput. Most production Kafka bugs are producer misconfigurations.

Sections

3.01
The motivating example (keep this in your head)A credit-card transaction processing system:
3.11
How does it work internally? Producer architecture
3.2
Constructing a producerBuilt in: ByteArraySerializer ("which doesn't do much"), StringSerializer, IntegerSerializer, and much more.
3.32
The three ways to sendEverything that goes wrong at or after the broker is invisible.
3.42
Configuring producers — every important knobA logical identifier for the client and the application it's used in.
3.53
Message delivery time — the timeout configs and how they interactHow long the producer may block when calling send() and when explicitly requesting metadata via partitionsFor().
3.62
⚠️ ORDERING GUARANTEES — the reordering bug
3.71
Remaining configsControls the size of a produce request. It caps both:
3.82
SerializersThat last point is the organizational killer — the same lockstep-deploy problem Ch. 1 identified, now embedded in your byte layout.
3.94
PartitionsThat's a genuine, deliberate availability-for-consistency trade.
3.10
HeadersThat last clause is the compelling one: headers let an intermediary route/trace data it is not authorized to read.
3.11
InterceptorsProducerInterceptor has two key methods:
3.122
Quotas and throttlingVia kafka-config.sh or the AdminClient API:
3.13Failure catalog
What actually breaks in production — Ch. 3 consolidated
3.143
Deploy / monitor / scale — the producer's operational surfaceThere is no producer-side backup, but there is a producer-side durability boundary, and knowing where it sits is what matters:
3.15Self-test
Self-test