This topic provides answers to frequently asked questions (FAQs) about metrics that are used in Message Queue for Apache Kafka.

What metrics do I need to monitor?

In most cases, you need to monitor only the following metrics:
  • InstanceMessageInput
  • TopicMessageInput
  • MessageAccumulation

The values of specific metrics, such as TopicMessageNumInput, TopicReqsInput, TopicReqsOutput, InstanceMessageNumInput, InstanceReqsInput, and InstanceReqsOutput, may be inaccurate when your business traffic is low or the version of your client is outdated. You do not need to focus on these metrics when your service traffic is normal.

Why are the values of specific metrics inaccurate?

Inaccuracies in metric values may be caused by one of the following causes:
  • Your business traffic is low. The system calculates the value of each metric based on a specific formula. If the business traffic is low, the deviation of the calculation result is large.
  • The version of your client is outdated. Earlier versions of the client do not pass the corresponding parameters. This results in deviation of monitoring data. We recommend that you upgrade your client to the latest version.
  • Data is compressed. Producers compress data to meet specific transmission requirements or storage requirements. This results in deviation of monitoring data.

Why is the value of the InstanceMessageOutput or TopicMessageOutput metric 0 when the value of the InstanceReqsOutput or TopicReqsOutput metric is greater than 0?

This issue is not caused by exceptions or errors. This issue occurs because the consumers remain active and attempt to pull messages from the broker when no new messages are published to the broker. As a result, the number of consumed messages is 0 and the number of message consumption attempts increases. Monitoring metrics