Agent for Envoy - Amazon App Mesh
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Agent for Envoy

Important

End of support notice: On September 30, 2026, Amazon will discontinue support for Amazon App Mesh. After September 30, 2026, you will no longer be able to access the Amazon App Mesh console or Amazon App Mesh resources. For more information, visit this blog post Migrating from Amazon App Mesh to Amazon ECS Service Connect.

The Agent is a process manager within the Envoy image that's vended for App Mesh. The Agent ensures Envoy remains running, stays healthy, and reduces downtime. It filters Envoy statistics and ancillary data to provide a distilled view of the Envoy proxy’s operation in App Mesh. This can help you troubleshooting related errors quicker.

You can use the Agent to configure the number of times that you want to restart the Envoy proxy in the event that the proxy becomes unhealthy. If a failure occurs, the Agent logs the conclusive exit status when Envoy exits. You can use this when troubleshooting the failure. The Agent also facilitates Envoy connection draining, which helps make your applications more resilient to failures.

Configure the Agent for Envoy using these variables:

  • APPNET_ENVOY_RESTART_COUNT – When this variable is set to a non-zero value, the Agent attempts to restart the Envoy proxy process up to the number that you set when it deems the proxy process status unhealthy on polling. This helps reduce downtime by providing faster restart compared to a task or pod replacement by the container orchestrator in the case of proxy health check failures.

  • PID_POLL_INTERVAL_MS – When configuring this variable, the default is kept to 100. When set to this value, you allow for faster detection and restart of the Envoy process when it exits compared to task or pod replacement through container orchestrator health checks.

  • LISTENER_DRAIN_WAIT_TIME_S – When configuring this variable, consider the container orchestrator timeout that's set for stopping the task or pod. For example, if this value is greater than the orchestrator timeout, the Envoy proxy can only drain for the duration until the orchestrator forcefully stops the task or pod.

  • APPNET_AGENT_ADMIN_MODE – When this variable is set to tcp or uds, the Agent provides a local management interface. This management interface serves as a safe endpoint to interact with the Envoy proxy and provides the following APIs for health checks, telemetry data and summarizes the operating condition of the proxy.

    • GET /status – Queries Envoy stats and returns server information.

    • POST /drain_listeners – Drains all inbound listeners.

    • POST /enableLogging?level=<desired_level> – Change Envoy logging level across all loggers.

    • GET /stats/prometheus – Show Envoy statistics in Prometheus format.

    • GET /stats/prometheus?usedonly – Only show statistics that Envoy has updated.

For more information about Agent configuration variables, see Envoy configuration variables.

The new Amazon App Mesh Agent is included in App Mesh-optimized Envoy images starting from version 1.21.0.0 and requires no additional resource allocation in customer tasks or pods.