A throttle-time metric reports how long a Kafka broker delays a client after the client exceeds a quota. Brokers use it to detect and manage quota violations while protecting shared cluster resources. section 13.3 section 13.4
Use the metric to reason about quota behavior:
- A rising value indicates that a client is being delayed rather than receiving unrestricted broker capacity. section 13.3
- Interpret it alongside produce, consume, and request quotas, which can constrain bytes per second or broker CPU time. section 13.3
- Treat unexpected throttling as an operational signal: aggressive limits can disrupt service and contribute to cascading failures. section 13.3 section 13.4
Misconfigured producer
A producer that rapidly generates data can consume disproportionate broker capacity. A quota limits its activity, and throttle-time metrics reveal that the broker is delaying the client to protect other Kafka users. section 13.3 section 13.4
