Question

A CloudOps engineer configures a CloudWatch alarm on a custom metric that a batch application publishes only when it processes jobs. Between job runs, the application publishes no data points at all. The team notices the alarm frequently enters the INSUFFICIENT_DATA state during idle periods, generating confusing notifications. They want the alarm to remain in the OK state during idle periods and only trigger when actual reported values breach the threshold. What is the most appropriate configuration change?

Accepted Answer

Set the alarm's 'Treat missing data as' setting to 'notBreaching' (treat as good).

Answer

Set the alarm's 'Treat missing data as' setting to 'breaching' (treat as bad).

Answer

Increase the alarm evaluation period to 24 hours to span multiple job runs.

Answer

Enable anomaly detection on the metric so idle periods are learned as normal behavior.