Search expressions

Search expressions are a type of math expression that you can add to CloudWatch graphs. Search expressions enable you to quickly add multiple related metrics to a graph. They also enable you to create dynamic graphs that automatically add appropriate metrics to their display, even if those metrics don’t exist when you first create the graph.

For example, you can create a search expression that displays the AWS/EC2 CPUUtilization metric for all instances in the Region. If you later launch a new instance, the CPUUtilization of the new instance is automatically added to the graph.

Search expression syntax

SEARCH(' {Namespace, DimensionName1, DimensionName2, ...} SearchTerm', 'Statistic', Period)

Example search expression

SUM(SEARCH('{ECS/ContainerInsights,ClusterName} MetricName="TaskCount"', 'Average', 300))

Fetching a specific metric under a dimension

Take a look at the following Search expression. This will select the metric called pod_cpu_utilization under ClusterName, Namespace, PodName dimension from the ContainerInsights namespace.

SEARCH('{ContainerInsights,ClusterName, Namespace, PodName} MetricName="pod_cpu_utilization" ', 'Average', 300)

This is extremely useful in high cardinality scenarios like containers where you might be deploying new application pods whose name you may not be aware while creating a dashboard widget. When you use Search expressions like the one above, the metric gets automatically added to the graph as soon as the criteria is matched.

Searching across metrics, dimensions and namespaces

You could also use an expression such as the one below. This will show the metric pod_cpu_utlization from all dimensions in the Namespace along with matching namespaces, dimensions that match the string.

SEARCH(' "pod_cpu_utilization" ','Average',120)


Using partial matches

The following search expression fetches values of MetricNames that contain the string pod_cpu under the dimension mentioned within {} curly braces.

SEARCH(' {ContainerInsights,ClusterName, Namespace, PodName}  MetricName=pod_cpu ','Average',120)