Quotas and Resource Isolation

A reporting team starts replaying several days of orders to rebuild a database. The consumer has permission to read those records, but its traffic competes with the live order producer for broker resources. Soon, an authorized background job is affecting an application that needs predictable latency.

Kafka quotas help contain this kind of interference by limiting supported categories of client resource use. They let a shared cluster slow an expensive workload while continuing to serve others.

This chapter explains what quotas measure, which clients share a limit, how throttling works, and where stronger isolation is necessary. Examples assume Apache Kafka 4.3.1 in KRaft mode, authenticated clients, and Kafka's default quota handling without a custom quota callback.