An event records something that happened
A producer publishes an event, such as order.created. The record contains a key, a value, and a timestamp, with optional headers.
Event streaming with Kafka
How producers, partitions, and consumers share a durable event log.
Process overview
- Produce
- Partition
- Store
- Consume
- React
Steps
5 stepsThe record enters a partition
Kafka appends events to a topic’s partitions. Ordering is guaranteed within a partition; a topic does not have one global event order.
The log retains the event
The topic stores events according to its retention configuration. Reading an event does not remove it, so another consumer can read it again.
Consumers read at their own pace
Consumers subscribe to topics and process records. Producers and consumers are decoupled, so each application can evolve around its own responsibility.
One fact can power many outcomes
The same order event can feed analytics, inventory, and notifications. Each consumer turns a shared fact into work for its own domain.
Keyboard: ←→
Scope
A basic Kafka event stream. Delivery guarantees, offset commits, consumer groups, retries, and transactions require additional design.
