Amazon Managed Service for Prometheus now supports out of order sample ingestion

This article announces new features for Amazon Managed Service for Prometheus to handle out-of-order metric sample ingestion and improve alerting accuracy.

• Out-of-order sample ingestion with default 1-minute time window for late-arriving metrics
• Configurable time window to match ingestion patterns or disable feature entirely
• Workspace-level rule query offset introduces delay before rule evaluation executes
• Reduces data loss and improves alerting for distributed collectors and variable latency
• New CloudWatch metrics track ingestion patterns for tuning settings
• Available in all AWS regions where Prometheus service is generally available

These features help preserve metric completeness and prevent missed alerts when samples arrive out of order due to network delays or batched exports.