Essential Monitoring Metrics

Critical JMX metrics every Kafka team must monitor: • UnderReplicatedPartitions > 0 → replication problem • ActiveControllerCount ≠ 1 → controller issue • OfflinePartitionsCount > 0 → data unavailability • RequestHandlerAvgIdlePercent < 0.3 → broker overloaded • NetworkProcessorAvgIdlePercent < 0.3 → network overloaded • BytesInPerSec + BytesOutPerSec → throughput monitoring • Consumer group lag → processing health

Key configuration

broker / client configurationproperties

# Add JMX exporter for Prometheus:
KAFKA_JMX_OPTS='-javaagent:/opt/jmx_exporter.jar=9101:/etc/jmx-kafka.yaml'

Key configuration

Try it in the simulator