3.5 techniques for safer message handling

Nils Larsgård

Short talk - in English

A lot can go wrong in a distributed event-based system. Underlying services are failing and becomes unresponsive. Your kafka-cluster goes down from time to time because of upgrades, full disks or just plain old crashes. Unexpected rebalancing of topics creates trouble for your kafka-consuming applications. Your micorservice is maybe healthy, but how do you handle that the rest of the world is falling apart?

In this talk I will talk about how you can process your kafka-messages in a safer way. I have destilled some practices from working with architectures at Finn.no and Reuters to 3.5 techniques that can become handy in any event-based system, not just limited to kafka.