A consumer group is a set of consumers that cooperatively split the partitions of one or more topics among themselves, so the group as a whole can process a topic's data in parallel while each partition is still consumed by exactly one member at a time.
Key Points: • Each partition in a subscribed topic is assigned to exactly one consumer within the group, so work is naturally load-balanced. • Adding more consumer instances to the group lets Kafka rebalance and hand them a share of the partitions, increasing throughput up to the partition count. • Multiple independent consumer groups can each read the entire topic at their own pace without interfering with one another. • If a consumer in the group fails, its partitions are automatically reassigned to the remaining members, which also supports fault tolerance alongside scalability.
Example: An analytics pipeline processing a high-volume "events" topic with 20 partitions might start with 4 consumer instances and later scale up to 20 as volume grows, each addition automatically absorbing a share of the remaining partitions during rebalance.
Interview Tip: A concise interview answer is:
"Consumer groups improve scalability by letting Kafka automatically split a topic's partitions across multiple consumer instances, so adding more consumers to the group increases processing parallelism up to the number of partitions, while Kafka handles rebalancing the work whenever group membership changes."