To view a broker's metrics in detail, click the arrow for its host name in the list at the bottom of the broker metrics overview screen. Both broker metrics and node metrics are included. Broker metrics are data collected from the Kafka broker, and node metrics are data collected from the node or pod where the Kafka broker is running.
Broker Summary
Displays summary information of the selected broker.
- Partitions. Number of topic partitions stored in the broker.
- PRODUCING. The number of message generated bytes per second for that broker.
- CONSUMING. The number of message bytes consumed per second for this broker. Includes Consumer and Follower.
- CPU Load. CPU utilization of that broker.
- Disk usage. Average disk usage for that broker. Average disk usage by major mounts.
- File descriptor. Number of open files and maximum number of open files in the corresponding broker.
Broker Metrics tab
Displays data collected directly from the Kafka broker.
Throughput
- Production. History of the number of message creation bytes per second for the corresponding broker.
- Consumption. History of message bytes consumed per second for that broker. Includes Consumer and Follower.
- Error Count. Number and types of errors that occurred during the production/consumption process.
Latency
- Producer. Latency metric history when requesting message creation from Producer.
- FetchConsumer. Latency metric history when requesting message fetch from Consumer.
- FetchFollower. Latency metric history for Message Fetch requests from Partition Follower.
The history displayed is the time taken to process 99% of the requests (99th percentile).
Latency details
You can view detailed information by clicking on a specific location in the latency history graph.
- request_queue time. How long the request stayed in the request queue.
- response_queue time. Time the response stayed in the response queue.
- request_local time. The time the request was processed locally (that is, the time used to write to the write buffer).
- request_remote time. The time the request was processed remotely (that is. the time it took the follower to replicate).
- response_send time. Time taken to send response.
Queue status
- Request Queue. History of how long a request has been in the request_queue.
- Response Queue. History of how long a response stayed in response_queue.
Network idle rate
- Percent. Network idle rate history
Zookeeper connection status
- Disconnection. History of the number of times ZooKeeper has been disconnected.
- Syncconnection. History of number of times connected to ZooKeeper.
- Expiration. History of the number of times ZooKeeper connection has expired.
Normally, the number of connections is always 1 greater than the number of disconnections.
If the number of disconnections continues to increase, a network failure may be suspected.
Connection expiration occurs when network latency is severe or problematic.
Node metrics tab
Node metrics display data collected from nodes or pods on which the Kafka broker operates. In a container environment, node metrics can display data from worker nodes, depending on the situation.
Disk usage
- Percent. Disk utilization history by major mount points.
- Bytes. Disk usage history by major mount points.
Memory usage
- Percent. Memory usage history.
- Bytes. Memory usage history.
CPU usage
- Percent
- CPU usage history