Amazon ECS adds non-critical daemons: agent failure no longer replaces the instance

On September 3, 2026, Amazon ECS Managed Daemons introduced a per-daemon critical parameter: with critical=false a failed agent (log shipper, metrics exporter) no longer drains and replaces the instance, while the default true keeps full instance replacement. Startup ordering is unchanged — the daemon reaches RUNNING 18–21 seconds before the application.

• The critical parameter is set per daemon and defaults to true
• With critical=true a daemon failure drains and replaces the instance
• With critical=false the instance stays ACTIVE and tasks are untouched
• The daemon starts 18–21 seconds before the application in both modes