CloudWatch Logs Anomaly Detection and Pattern Analysis

With log anomalies and patterns features, you can now detect abnormalities in your service and group log entries with recurring patterns with the CloudWatch data source.

Log Anomalies

You can now list the log anomalies in your service directly in Grafana. CloudWatch Log Anomaly detection is a new query type for this data source that leverages machine-learning and pattern recognition to establish baselines of typical log content:

You can also use the Logs Insights query editor and the anomaly command to define and display log anomalies ad-hoc. See CloudWatch Logs Insights documentation for more on anomaly detection. 

Log Patterns

Group log entry messages (or other string values) matching specific patterns with the patterns command using any CloudWatch Logs query language. The associated diff command compares the log events found in your requested time period with the log events from a previous time period of equal length, returning results that describe the change.

We have added language support for both commands to CloudWatch Logs Insights language editor.  More info can be found in the AWS Log Patterns documentation and the AWS Logs Insights syntax guide.