A consumer assignment strategy determines how Kafka distributes topic partitions among consumers in a group. It supports parallel processing while defining which consumer reads each partition. section 4.2.3 section 4.7
During group startup, consumers join a broker-selected coordinator. One consumer becomes group leader, calculates assignments using the configured strategy, and returns them to the coordinator, which informs the group members. Consumers then fetch data directly from partition leaders. section 4.2.3
Design and governance considerations include:
- Configure the client-side partition.assignment.strategy property to select an available strategy. section 4.2.3
- Implement the PartitionAssignor interface when a custom distribution strategy is required. section 4.2.3
- Document the assignment strategy alongside consumer services, expected groups, partition counts, and consumption models. section 10.1.3
- Distinguish dynamic subscription-based assignments from explicit partition assignment when targeted processing is needed. section 4.7
Consumer-group startup
Several consumers join a group, one is elected leader, and the leader calculates a partition distribution. The coordinator distributes the result; each consumer polls only its assigned partitions, processes messages, and commits offsets. section 4.2.3 section 4.2
Figure 4.9 Distributing the assignments
The assignment-distribution diagram makes the coordinator, group leader, consumers, and partition allocation relationships concrete. section 4.2.3
