Events in Motion: Kafka & Messaging for Java Developers

09:00 – 10:15

See why services should stop waiting for each other

synchronous vs asynchronous, events vs commands, where messaging belongs

Takeaway A cascading failure demonstrated in a chain of REST calls, and the same flow sketched as events, with the failure contained.

10:30 – 11:45

Stand up Kafka and read its anatomy

brokers and topics, partitions and offsets, first contact from the CLI

Takeaway Kafka running on your laptop with messages produced and read from the command line: the moving parts seen bare before any framework.

12:00 – 13:00

Produce events deliberately

producers and keys, partitioning behaviour, acks and delivery guarantees

Takeaway Ordering proved to follow the key, by watching same-key events land in one partition and mixed keys spread.

14:00 – 15:15

Consume as a group

consumer groups, offsets and commits, rebalancing

Takeaway A rebalance watched live: a consumer killed mid-stream and its partitions reassigned, with nothing lost.

15:30 – 17:00

Design the event itself

event payload design, schema versioning, what must never ride in an event

Takeaway An event schema that survives its first breaking change, because you versioned it before you had to.

09:00 – 10:15

Wire it into Spring

Spring Kafka listeners, serialisation, configuration that matters

Takeaway Two Spring services talking through a topic instead of a REST call: the morning's CLI knowledge now running as code.

10:30 – 11:45

Handle failure without losing messages

retries and backoff, error handlers, poison pills

Takeaway A failing consumer retrying with backoff instead of spinning, and the message still there when it recovers.

12:00 – 13:00

Send the poison pill somewhere safe

dead-letter topics, alerting on the DLT, replaying after a fix

Takeaway A poison message parked on a dead-letter topic, the fix deployed, and the message replayed: the full recovery loop, done once by hand.

14:00 – 15:15

Survive being delivered twice

duplicate delivery, idempotency keys, exactly-once claims read honestly

Takeaway A consumer handed the same event twice on purpose, and the payment it guards charged exactly once.

15:30 – 17:00

Assemble the flow and defend it

an end-to-end order flow, consumer lag read, review

Takeaway An order flow across two services demoed end to end, with lag, retries and the dead-letter path explained to the room.