AWS CloudWatch Metrics

Overview

You can create a CloudWatch metric stream and direct it to an Amazon Data Firehose delivery stream. Firehose can then deliver the metrics to Dash0 through an HTTP endpoint. This guide will teach you how.

Create a Firehose delivery stream

Follow AWS documentation to create a Firehose delivery stream from AWS console:

1. Choose source and destination

Choose "Direct PUT" as the source and "HTTP Endpoint" as the destination of your Firehose stream.

2. Firehose stream name

Give your Firehose stream a name.

3. Destination settings

In HTTP Endpoint URL, enter AWS CloudWatch Metrics (JSON) via Firehose HTTP endpoint, which can be found in your Dash0 organization settings under Endpoints:

Create a Dash0 authentication token for the Firehose stream. It is recommended to grant only the ingestion permission to this token.

In the authentication section, you can either choose "Use access key" option and enter the authentication token directly as the Access Key:

or use AWS Secrets Manager to retrieve the token programmatically.

By default, data will be sent to the default dataset. To route data to a specific dataset, you have two options:

• Via Firehose parameters (recommended): In the Parameters section, add a parameter with key dash0-dataset and your dataset identifier as the value. Make sure the authentication token has ingest permissions for that dataset.
• Via token restriction: Restrict the authentication token's access to the specific dataset you want to target.

You can keep the rest of the settings as default. Optionally, you can enable GZIP encoding to compress the data.

4. Backup settings

Create or choose an existing S3 bucket to store data in case of delivery failures.

Create a CloudWatch metric stream to send metrics to Firehose

Follow AWS documentation to create a CloudWatch metric stream that sends selected metrics in JSON to your Firehose delivery stream previously created.

1. Choose destination

Choose Custom setup with Firehose and select the Amazon Data Firehose stream created in the previous step.

2. Change output format

Choose JSON as the output format.

CloudWatch supports exporting metrics in both JSON and OpenTelemetry 1.0 format. We have tested sending metrics in both formats to Dash0 and found that metrics exported in JSON format have more structured resource attributes that are easier to query. Therefore, we require using the JSON format. Choosing OpenTelemetry 1.0 format will lead to delivery failure in Firehose stream.

3. Select metrics to be streamed

Select the metrics you want to stream to Dash0.

4. Custom metric stream name

Give your CloudWatch metric stream a unique name.

Explore metrics in Dash0

Once everything is set up, you can find CloudWatch metrics in Dash0 Metrics Explorer by filtering with aws.cloudwatch.metric_stream_name or with your cloud.account.id.

You will see metrics grouped under aws metric namespace. The metric names have been normalized according to OpenTelemetry naming conventions. For example, CPUUtilization of EC2 instances becomes aws.ec2.cpu_utilization.

Use CloudWatch Metrics for Alerting

You can create check rules using CloudWatch metrics imported via Firehose. However, be aware that CloudWatch and Firehose introduce buffering delays when delivering metrics to Dash0. These delays may cause alerting issues: when a check rule is evaluated, the most recent metric data may not yet be available in Dash0.

To account for this delay, add an offset clause to your check rule expression. This shifts the query time window backward, ensuring you query data that has already arrived:

The offset value (e.g., 5m) should be adjusted based on the observed delay in your environment.