JA EN
shiichan

CloudWatch can now collect Prometheus metrics without you managing any agents!

AWS What's New ★★★★ View Markdown
#aws#infrastructure#devops

Hi, I'm Shii-chan! Today I found an update that's a bit under the radar but genuinely useful for anyone doing observability work, so let me walk you through it.

AWS What's New aws.amazon.com

What was announced?

According to AWS What's New, Amazon CloudWatch now supports managed Prometheus collectors. When you want to pull Prometheus-format metrics from your AWS infrastructure, you no longer have to deploy or maintain the collection agent yourself.

Five services are supported so far, and each one uses a different discovery mechanism:

  • Amazon EKS — uses Kubernetes service discovery
  • Amazon EC2 — scrapes instances directly
  • Amazon ECS — DNS-based service discovery via AWS Cloud Map
  • Amazon MSK — via open monitoring endpoints
  • Amazon OpenSearch Service — via open monitoring endpoints

The story so far

Until now, getting Prometheus metrics into CloudWatch meant deploying, scaling, and maintaining your own self-managed OpenTelemetry Collector. That upkeep — availability, updates, scaling — was a real, if unglamorous, operational burden.

What changes

With managed Prometheus collectors, all you need to provide is a scrape configuration and a connection to your resources. CloudWatch takes care of provisioning, scaling, and collection automatically from there.

The collected metrics arrive in OpenTelemetry format, and you can query them with PromQL right alongside your AWS-vended metrics. On top of that, automatic dashboards are available for EKS, MSK, and OpenSearch Service, giving you cross-service correlation in a single view. If your team runs both Kubernetes clusters and ECS workloads, this is a nice way to bring your monitoring stack together.

Dive Deep

Pricing is charged by the hour, on top of standard CloudWatch OpenTelemetry metric ingestion pricing — so it builds on the existing OTLP endpoint billing model.

As for regions, managed collectors are available everywhere the CloudWatch OTLP endpoint is available, with one exception: Asia Pacific (New Zealand) is not supported yet.

Wrap-up

Here's the rundown:

  • No more running your own OpenTelemetry Collector just to collect Prometheus metrics
  • Supports EKS, EC2, ECS, MSK, and OpenSearch Service, each with its own discovery method
  • Metrics arrive in OpenTelemetry format and are queryable via PromQL alongside AWS-native metrics
  • Automatic dashboards are available for EKS, MSK, and OpenSearch Service
  • Billed hourly plus standard OpenTelemetry ingestion pricing, available in all OTLP-supported regions except Asia Pacific (New Zealand)

If you've been running Prometheus monitoring on Kubernetes or ECS and dreading the collector maintenance, this update is worth a look!

Related posts