Sensors

Cruise Control Metrics

Cruise Control uses dropwizzard metrics to report its own status.

Cruise Control metrics are useful to monitor the state of Cruise Control itself. There are metrics for the following sensor types:

• Executor
• LoadMonitor
• UserTaskManager
• AnomalyDetector
• GoalOptimizer
• MetricFetcherManager
• KafkaCruiseControlServlet

Executor Sensors

DESCRIPTION	MBEAN NAME
The number of replica action in progress	kafka.cruisecontrol:name=Executor.replica-action-in-progress
The number of leadership action in progress	kafka.cruisecontrol:name=Executor.leadership-action-in-progress
The number of replica action pending	kafka.cruisecontrol:name=Executor.replica-action-pending
The number of leadership action pending	kafka.cruisecontrol:name=Executor.leadership-action-pending
The number of replica action aborting	kafka.cruisecontrol:name=Executor.replica-action-aborting
The number of leadership action aborting	kafka.cruisecontrol:name=Executor.leadership-action-aborting
The number of replica action aborted	kafka.cruisecontrol:name=Executor.replica-action-aborted
The number of leadership action aborted	kafka.cruisecontrol:name=Executor.leadership-action-aborted
The number of replica action dead	kafka.cruisecontrol:name=Executor.replica-action-dead
The number of leadership action dead	kafka.cruisecontrol:name=Executor.leadership-action-dead
Has an ongoing execution in kafka_assigner mode	kafka.cruisecontrol:name=Executor.ongoing-execution-kafka_assigner
Has an ongoing execution in non-kafka_assigner mode	kafka.cruisecontrol:name=Executor.ongoing-execution-non-kafka_assigner
The number of (all) execution stopped	kafka.cruisecontrol:name=Executor.execution-stopped
The number of execution stopped by user	kafka.cruisecontrol:name=Executor.execution-stopped-by-user
The number of execution started in kafka_assigner mode	kafka.cruisecontrol:name=Executor.execution-started-kafka_assigner
The number of execution started in non-kafka_assigner mode	kafka.cruisecontrol:name=Executor.execution-started-non-kafka_assigner
Per broker cap on inter-broker partition movements	kafka.cruisecontrol:name=Executor.inter-broker-partition-movements-per-broker-cap
Per broker cap on intra-broker partition movements	kafka.cruisecontrol:name=Executor.intra-broker-partition-movements-per-broker-cap
Global cap on leadership movements	kafka.cruisecontrol:name=Executor.leadership-movements-global-cap

LoadMonitor Sensors

DESCRIPTION	MBEAN NAME
The number of monitored windows	kafka.cruisecontrol:name=LoadMonitor.total-monitored-windows
The number of partitions that is valid but require extrapolations	kafka.cruisecontrol:name=LoadMonitor.num-partitions-with-extrapolations
The number of topics	kafka.cruisecontrol:name=LoadMonitor.num-topics
The metadata factor, which corresponds to (number of replicas) * (number of brokers with replicas) ^ exponent	kafka.cruisecontrol:name=LoadMonitor.metadata-factor
The number of valid windows	kafka.cruisecontrol:name=LoadMonitor.valid-windows
The monitored partition percentage	kafka.cruisecontrol:name=LoadMonitor.monitored-partitions-percentage
Cluster model creation time in ms	kafka.cruisecontrol:name=LoadMonitor.cluster-model-creation-timer
The cluster has partitions with ISR > replicas (0: No such partitions, 1: Has such partitions)	kafka.cruisecontrol:name=LoadMonitor.has-partitions-with-isr-greater-than-replicas

UserTaskManager Sensors

DESCRIPTION	MBEAN NAME
The number of active sessions	kafka.cruisecontrol:name=UserTaskManager.num-active-sessions
The number of active user tasks	kafka.cruisecontrol:name=UserTaskManager.num-active-user-tasks

AnomalyDetector Sensors

DESCRIPTION	MBEAN NAME
The number of self healing started	kafka.cruisecontrol:name=AnomalyDetector.number-of-self-healing-started
The ongoing anomaly duration in ms	kafka.cruisecontrol:name=AnomalyDetector.ongoing-anomaly-duration-ms
Whether broker failure self-healing is enabled or not	kafka.cruisecontrol:name=AnomalyDetector.broker_failure-self-healing-enabled
Whether goal violation self-healing is enabled or not	kafka.cruisecontrol:name=AnomalyDetector.goal_violation-self-healing-enabled
Whether disk failure self-healing is enabled or not	kafka.cruisecontrol:name=AnomalyDetector.disk_failure-self-healing-enabled
Whether metric anomaly self-healing is enabled or not	kafka.cruisecontrol:name=AnomalyDetector.metric_anomaly-self-healing-enabled
Whether topic anomaly self-healing is enabled or not	kafka.cruisecontrol:name=AnomalyDetector.topic_anomaly-self-healing-enabled
Whether goal violation detector identified goals that require human intervention (e.g. cluster expansion)	kafka.cruisecontrol:name=AnomalyDetector.GOAL_VIOLATION-has-unfixable-goals
Balancedness score (100 = fully-balanced, 0 = fully-unbalanced, -1 = has dead-brokers / disks in cluster)	kafka.cruisecontrol:name=AnomalyDetector.balancedness-score
Whether the cluster is under-provisioned or not	kafka.cruisecontrol:name=AnomalyDetector.under-provisioned
Whether the cluster is over-provisioned or not	kafka.cruisecontrol:name=AnomalyDetector.over-provisioned
Whether the cluster is right-sized or not	kafka.cruisecontrol:name=AnomalyDetector.right-sized
Mean time to start a self-healing fix	kafka.cruisecontrol:name=AnomalyDetector.mean-time-to-start-fix-ms
Broker failure rate	kafka.cruisecontrol:name=AnomalyDetector.broker-failure-rate
Goal violation rate	kafka.cruisecontrol:name=AnomalyDetector.goal-violation-rate
Metric anomaly rate	kafka.cruisecontrol:name=AnomalyDetector.metric-anomaly-rate
Disk failure rate	kafka.cruisecontrol:name=AnomalyDetector.disk-failure-rate
Topic anomaly rate	kafka.cruisecontrol:name=AnomalyDetector.topic-anomaly-rate
The number of brokers that are metric anomaly suspects, pending more evidence to conclude either way	kafka.cruisecontrol:name=AnomalyDetector.num-suspect-metric-anomalies
The number of brokers that have recently been identified with a metric anomaly	kafka.cruisecontrol:name=AnomalyDetector.num-recent-metric-anomalies
The number of brokers that continue to be identified with a metric anomaly for a prolonged period	kafka.cruisecontrol:name=AnomalyDetector.num-persistent-metric-anomalies
The cluster has partitions with RF > the number of eligible racks (0: No such partitions, 1: Has such partitions)	kafka.cruisecontrol:name=AnomalyDetector.has-partitions-with-replication-factor-greater-than-num-racks
The time taken by goal violation detection	kafka.cruisecontrol:name=AnomalyDetector.goal-violation-detection-timer
The time taken to generate a fix for self-healing for broker failures	kafka.cruisecontrol:name=AnomalyDetector.broker_failure-self-healing-fix-generation-timer
The time taken to generate a fix for self-healing for maintenance events	kafka.cruisecontrol:name=AnomalyDetector.maintenance_event-self-healing-fix-generation-timer
The time taken to generate a fix for self-healing for disk failures	kafka.cruisecontrol:name=AnomalyDetector.disk_failure-self-healing-fix-generation-timer
The time taken to generate a fix for self-healing for metric anomalies	kafka.cruisecontrol:name=AnomalyDetector.metric_anomaly-self-healing-fix-generation-timer
The time taken to generate a fix for self-healing for goal violations	kafka.cruisecontrol:name=AnomalyDetector.goal_violation-self-healing-fix-generation-timer
The time taken to generate a fix for self-healing for topic anomalies	kafka.cruisecontrol:name=AnomalyDetector.topic_anomaly-self-healing-fix-generation-timer
The rate at which automated rightsizing with actions are taken	kafka.cruisecontrol:name=AnomalyDetector.automated-rightsizing-rate

GoalOptimizer Sensors

DESCRIPTION	MBEAN NAME
Proposal computation time in ms	kafka.cruisecontrol:name=GoalOptimizer.proposal-computation-timer
Has unfixable proposals optimization	kafka.cruisecontrol:name=GoalOptimizer.has-unfixable-proposal-optimization

MetricFetcherManager Sensors

DESCRIPTION	MBEAN NAME
The rate of partition metric sample fetch failures	kafka.cruisecontrol:name=MetricFetcherManager.partition-samples-fetcher-failure-rate
The time taken by each round of partition sample fetch	kafka.cruisecontrol:name=MetricFetcherManager.partition-samples-fetcher-timer
The rate of training sample fetch failures	kafka.cruisecontrol:name=MetricFetcherManager.training-samples-fetcher-failure-rate
The time taken by each training sample fetch	kafka.cruisecontrol:name=MetricFetcherManager.training-samples-fetcher-timer

KafkaCruiseControlServlet Sensors

DESCRIPTION	MBEAN NAME
Service time of a successful request in ms for each endpoint	kafka.cruisecontrol:name=KafkaCruiseControlServlet.-successful-request-execution-timer
Request rate for each endpoint	kafka.cruisecontrol:name=KafkaCruiseControlServlet.-request-rate

---

Metrics Collection

You can easily expose Cruise Control metrics by enabling Java Management Extensions (JMX), and collect them by using your favourite JMX collection tool (e.g - jmx_exporter).