opentelemetry-java-contrib/jmx-metrics/docs/target-systems/kafka.md

Kafka Metrics

The JMX Metric Gatherer provides built in Kafka metric gathering capabilities for versions v0.8.2.x and above. These metrics are sourced from Kafka's exposed Yammer metrics for each instance: https://kafka.apache.org/documentation/#monitoring

Metrics

Broker Metrics

• Name: kafka.message.count

• Description: The number of messages received by the broker

• Unit: {messages}

• Instrument Type: LongCounterObserver

• Name: kafka.request.count

• Description: The number of requests received by the broker

• Unit: {requests}

• Attributes: type

• Instrument Type: LongCounterObserver

• Name: kafka.request.failed

• Description: The number of requests to the broker resulting in a failure

• Unit: {requests}

• Attributes: type

• Instrument Type: LongCounterObserver

• Name: kafka.request.time.total

• Description: The total time the broker has taken to service requests

• Unit: ms

• Attributes: type

• Instrument Type: LongCounterObserver

• Name: kafka.request.time.50p

• Description: The 50th percentile time the broker has taken to service requests

• Unit: ms

• Attributes: type

• Instrument Type: DoubleValueObserver

• Name: kafka.request.time.99p

• Description: The 99th percentile time the broker has taken to service requests

• Unit: ms

• Attributes: type

• Instrument Type: DoubleValueObserver

• Name: kafka.request.time.avg

• Description: The average time the broker has taken to service requests

• Unit: ms

• Attributes: type

• Instrument Type: DoubleValueObserver

• Name: kafka.network.io

• Description: The bytes received or sent by the broker

• Unit: by

• Attributes: state

• Instrument Type: LongCounterObserver

• Name: kafka.purgatory.size

• Description: The number of requests waiting in purgatory

• Unit: {requests}

• Attributes: type

• Instrument Type: LongValueObserver

• Name: kafka.partition.count

• Description: The number of partitions on the broker

• Unit: {partitions}

• Instrument Type: LongValueObserver

• Name: kafka.partition.offline

• Description: The number of partitions offline

• Unit: {partitions}

• Instrument Type: LongValueObserver

• Name: kafka.partition.under_replicated

• Description: The number of under replicated partitions

• Unit: {partitions}

• Instrument Type: LongValueObserver

• Name: kafka.isr.operation.count

• Description: The number of in-sync replica shrink and expand operations

• Unit: {operations}

• Attributes: operation

• Instrument Type: LongCounterObserver

• Name: kafka.max.lag

• Description: Max lag in messages between follower and leader replicas

• Unit: {messages}

• Instrument Type: LongValueObserver

• Name: kafka.controller.active.count

• Description: Controller is active on broker

• Unit: {controllers}

• Instrument Type: LongValueObserver

• Name: kafka.leader.election.rate

• Description: Leader election rate - increasing indicates broker failures

• Unit: {elections}

• Instrument Type: LongCounterObserver

• Name: kafka.unclean.election.rate

• Description: Unclean leader election rate - increasing indicates broker failures

• Unit: {elections}

• Instrument Type: LongCounterObserver

• Name: kafka.request.queue

• Description: Size of the request queue

• Unit: {requests}

• Instrument Type: LongValueObserver

Log metrics

• Name: kafka.logs.flush.time.count

• Description: Log flush count

• Unit: ms

• Instrument Type: LongSumObserver

• Name: kafka.logs.flush.time.median

• Description: Log flush time - 50th percentile

• Unit: ms

• Instrument Type: DoubleValueObserver

• Name: kafka.logs.flush.time.99p

• Description: Log flush time - 99th percentile

• Unit: ms

• Instrument Type: DoubleValueObserver