High Request Queue Size Indicates Broker Overload
warningWhen request queue size grows, it indicates the broker cannot process incoming requests fast enough, leading to increased latency and potential client timeouts.
Monitor kafka.request.queue_size and kafka.response.queue_size trending upward. Alert when queue size exceeds 100 sustained for 5+ minutes. Cross-reference with kafka.request.handler_idle_percent to confirm handler saturation.
1. Increase num.io.threads: Add handler capacity. 2. Check broker CPU: Scale vertically if CPU-bound. 3. Optimize request patterns: Reduce request frequency from clients. 4. Monitor request types: Use kafka.request.produce_rate and kafka.request.fetch_consumer_rate to identify dominant request type. 5. Scale horizontally: Add brokers to distribute load. 6. Review slow storage: Check if disk I/O is causing slow request processing.