Metrics collected by Application Signals

Application Signals collects both standard application metrics and runtime metrics from the applications that you enable it for.

Standard application metrics relate to the most critical parts of service performance, latency and availability.

Runtime metrics track application metrics over time, including memory usage, CPU usage, and garbage collection. Application Signals displays the runtime metrics in the context of the services that you have enabled for Application Signals. When you have an operational issue, observing the runtime metrics can be useful to help you find the root cause of the issue. For example, you can see if spikes in latency in your service relate to spikes in a runtime metric.

Standard application metrics collected

Application Signals collects standard application metrics from the services that it discovers. These metrics relate to the most critical aspects of a service's performance: latency, faults, and errors. They can help you identify issues, monitor performance trends, and optimize resources to improve the overall user experience.

The following table lists the metrics collected by Application Signals. These metrics are sent to CloudWatch in the ApplicationSignals namespace.

The Availability metric displayed on Application Signals dashboards is computed as (1 - Faults/Total)*100. Total responses includes all responses and is derived from SampleCount(Latency). Successful responses are all responses without a 5XX error. 4XX responses are treated as successful when Application Signals calculates Availability.

Dimensions collected and dimension combinations

The following dimensions are defined for each of the standard application metrics. For more information about dimensions, see Dimensions.

Different dimensions are collected for service metrics and dependency metrics. Within the services discovered by Application Signals, when microservice A calls microservice B, microservice B is serving the request. In this case, microservice A emits dependency metrics and microservice B emits service metrics. When a client calls microservice A, microservice A is serving the request and emits service metrics.

Dimensions for service metrics

The following dimensions are collected for service metrics.

When you view these metrics in the CloudWatch console, you can view them using the following dimension combinations:

Dimensions for dependency metrics

The following dimensions are collected for dependency metrics:

When you view these metrics in the CloudWatch console, you can view them using the following dimension combinations:

Running on Amazon EKS clusters

Runtime metrics

Application Signals uses the AWS Distro for OpenTelemetry SDK to automatically collect OpenTelemetry-compatible metrics from your Java and Python applications. To have runtime metrics collected, you must meet the following pre-requisites:

Runtime metrics are not collected for .NET or Node.js applications.

Runtime metric are charged as part of Application Signals costs. For more information about CloudWatch pricing, see Amazon CloudWatch Pricing.

Java runtime metrics

Application Signals collects the following JVM metrics from Java applications that you enable for Application Signals. All runtime metrics are sent to CloudWatch in the ApplicationSignals namespace, and are collected with the Service and Environment dimension set.

Python runtime metrics

Application Signals collects the following metrics from Python applications that you enable for Application Signals. All runtime metrics are sent to CloudWatch in the ApplicationSignals namespace, and are collected with the Service and Environment dimension set.

Disabling the collection of runtime metrics

Runtime metrics are collected by default for Java and Python applications that are enabled for Application Signals. If you want to disable the collection of these metrics, follow the instructions in this section for your environment.

Amazon EKS

To disable runtime metrics in Amazon EKS applications at the application level, add the following environment variable to your workload specification.

env:
    - name: OTEL_AWS_APPLICATION_SIGNALS_RUNTIME_ENABLED 
      value: "false"

To disable runtime metrics in Amazon EKS applications at the cluster level, apply the configuration to the advanced configuration of your Amazon CloudWatch Observability EKS add-on.

{
  "agent": {
    "config": {
      "traces": {
        "traces_collected": {
          "application_signals": {
            
          }
        }
      },
      "logs": {
        "metrics_collected": {
          "application_signals": {
            
          }
        }
      }
    },
    "manager": {
      "autoInstrumentationConfiguration": {
        "java": {
          "runtime_metrics": {
            "enabled": "false"
          }
        },
        "python": {
          "runtime_metrics": {
            "enabled": "false"
          }
        }
      }
    }
  }

Amazon ECS

To disable runtime metrics in Amazon ECS applications, add the environment variable OTEL_AWS_APPLICATION_SIGNALS_RUNTIME_ENABLED=false in the new task definition revision and redeploy the application.

EC2

To disable runtime metrics in Amazon EC2 applications, add the environment variable OTEL_AWS_APPLICATION_SIGNALS_RUNTIME_ENABLED=false before the application starts.

Kubernetes

To disable runtime metrics in Kubernetes applications at the application level, add the following environment variable to your workload specification.

env:
    - name: OTEL_AWS_APPLICATION_SIGNALS_RUNTIME_ENABLED 
      value: "false"

To disable runtime metrics in Kubernetes applications at the cluster level, use the following:

helm upgrade ... \
--set-string manager.autoInstrumentationConfiguration.java.runtime_metrics.enabled=false \
--set-string manager.autoInstrumentationConfiguration.python.runtime_metrics.enabled=false