Amazon CloudWatch adds managed Prometheus collectors, retiring the self-hosted scraper

Amazon CloudWatch now offers fully managed Prometheus metric collectors, so teams no longer need to deploy and maintain a self-managed OpenTelemetry Collector just to get Prometheus-format metrics into AWS. Point CloudWatch at EKS, EC2, ECS, MSK, or OpenSearch Service resources with a scrape configuration, and AWS handles provisioning, scaling, and collection automatically. Metrics arrive in OpenTelemetry format and are queryable with PromQL right alongside native CloudWatch metrics, giving unified dashboards, alarms, and cross-service correlation instead of a bolted-on Prometheus stack. Service discovery is built in: Kubernetes-native for EKS, AWS Cloud Map for ECS, and direct or open-endpoint scraping for EC2, MSK, and OpenSearch. It's billed hourly on top of standard OTLP ingestion pricing, and it's live everywhere the CloudWatch OTLP endpoint runs except Asia Pacific (New Zealand). For anyone running Prometheus infrastructure purely to feed CloudWatch, this quietly removes a whole layer of ops overhead.

This is a summary by our content curator. Read the original at AWS What's New: https://aws.amazon.com/about-aws/whats-new/2026/07/cloudwatch-managed-collectors/.