Metrics added / replaced for Kinesis Data Streams / Step Functions AWS Integration and more

Hello. Mackerel team CRE Inoue (id:a-know) here with the last update announcement of 2020. Thank you for choosing Mackerel this year.

Metrics added / replaced for Kinesis Data Streams AWS Integration

The following metrics have been added / replaced for Kinesis Data Streams.


  • Records
    • GetRecords.Records (kinesis.records.get_records)
    • IncomingRecords (kinesis.records.incoming)
    • PutRecords.Records (kinesis.records.put_records)


  • Records
    • GetRecords.Records (kinesis.records.get_records)
    • IncomingRecords (kinesis.records.incoming)
    • PutRecords.SuccessfulRecords (kinesis.records.put_records_successful)
    • PutRecords.TotalRecords (kinesis.records.put_records_total)
    • PutRecords.FailedRecords (kinesis.records.put_records_failed)
    • PutRecords.ThrottledRecords (kinesis.records.put_records_throttled)

The metric previously recorded as PutRecords.Records (kinesis.records.put_records) is now PutRecords.SuccessfulRecords (kinesis.records.put_records_successful). Three new types of records data are now integrated as well. The name change from kinesis.records.put_records to kinesis.records.put_records_successful should be noted in particular because it may affect things such as monitors.

Also, the maximum number of metrics per host has been increased from 31 to 34.

Metrics added for AWS Integration with Step Functions

  • Service Integrations
    • ServiceIntegrationsFailed (states.service_integrations.SERVICE_INTEGRATION_NAME.failed)
    • ServiceIntegrationsScheduled (states.service_integrations.SERVICE_INTEGRATION_NAME.scheduled)
    • ServiceIntegrationsStarted (states.service_integrations.SERVICE_INTEGRATION_NAME.started)
    • ServiceIntegrationsSucceeded (states.service_integrations.SERVICE_INTEGRATION_NAME.succeeded)
    • ServiceIntegrationsTimedOut (states.service_integrations.SERVICE_INTEGRATION_NAME.timedout)
  • Service Integration Time
    • ServiceIntegrationRunTime (
    • ServiceIntegrationScheduleTime (states.service_integration_time.SERVICE_INTEGRATION_NAME.schedule)
    • ServiceIntegrationTime (states.service_integration_time.SERVICE_INTEGRATION_NAME.all)

The service integration name replaces SERVICE_INTEGRATION_NAME.

With this addition, the maximum number of metrics per host is now calculated using the formula 7 + 9 × (number of activities) + 8 × (number of Lambda functions) + 8 × (number of service integrations). Please be sure to take note of the number of metrics as this may affect your micro hosts.

The same applies to the above-mentioned Kinesis Data Streams integration. If necessary, you can select integration targets using the optional selection function.

Network interface metrics can now be excluded with mackerel-agent

You can now exclude network interface metrics by writing the following in the mackerel-agent.conf.

ignore = "bond.*"

Using a regular expression, metrics from the specified interface will not be collected. For Windows Server, specify the interface name that corresponds to that in Device Manager.

See the help page below for more information.

Along with this update, the mkr command line tool has been updated from 0.42.0 to 0.43.0.

Official support for mackerel-agent in CentOS6 environments has ended

With CentOS6 ending their support on November 30th of this year, Mackerel too has officially ended support for CentOS6 environments.

Even if an environment is kept the same, surrounding environments are always changing (for example, managing old versions of TLS communication and host CA certificate renewal etc.), and so we believe that the best course of action is to update and switch environments. We apologize for the inconvenience and thank you for your cooperation.