Here is the list of metrics based on MBeans exposed by Kafka broker.
Broker metrics:
Metric Name | Type | Attributes | Description |
---|---|---|---|
kafka.message.count | Counter | The number of messages received by the broker | |
kafka.request.count | Counter | type | The number of requests received by the broker |
kafka.request.failed | Counter | type | The number of requests to the broker resulting in a failure |
kafka.request.time.total | Counter | type | The total time the broker has taken to service requests |
kafka.request.time.50p | Gauge | type | The 50th percentile time the broker has taken to service requests |
kafka.request.time.99p | Gauge | type | The 99th percentile time the broker has taken to service requests |
kafka.request.queue | UpDownCounter | Size of the request queue | |
kafka.network.io | Counter | direction | The bytes received or sent by the broker |
kafka.purgatory.size | UpDownCounter | type | The number of requests waiting in purgatory |
kafka.partition.count | UpDownCounter | The number of partitions on the broker | |
kafka.partition.offline | UpDownCounter | The number of partitions offline | |
kafka.partition.underReplicated | UpDownCounter | The number of under replicated partitions | |
kafka.isr.operation.count | UpDownCounter | operation | The number of in-sync replica shrink and expand operations |
kafka.lag.max | Gauge | The max lag in messages between follower and leader replicas | |
kafka.controller.active.count | UpDownCounter | The number of controllers active on the broker | |
kafka.leaderElection.count | Counter | The leader election count | |
kafka.leaderElection.unclean.count | Counter | Unclean leader election count - increasing indicates broker failures |
Log metrics:
Metric Name | Type | Attributes | Description |
---|---|---|---|
kafka.logs.flush.count | Counter | Log flush count | |
kafka.logs.flush.time.50p | Gauge | Log flush time - 50th percentile | |
kafka.logs.flush.time.99p | Gauge | Log flush time - 99th percentile |