RTF Application Graceful Shutdown Behavior and Configuring shutdownTimeout

Default Behavior

By default, when an RTF application is stopped via the control plane:

1. The Mule runtime receives a shutdown signal.
2. It allows up to 5 seconds (default shutdownTimeout) for in-flight transactions to complete.
3. After the timeout expires, the runtime forces shutdown regardless of transaction state.

Configuring shutdownTimeout (Application Level)

To allow more time for in-flight transactions to complete, configure the shutdownTimeout parameter in your Mule application. Refer to:

Configure Shutdown Timeout — MuleSoft Documentation

Configuring terminationGracePeriodSeconds (Infrastructure Level)

Important: If you set shutdownTimeout to a value greater than 30000 milliseconds, you must also increase the Kubernetes terminationGracePeriodSeconds to match. Otherwise, Kubernetes will forcefully terminate the pod at 30 seconds.

• If using Anypoint Controller for Kubernetes: You can configure terminationGracePeriodSeconds to a value greater than 30 seconds. See: Deploy Mule Apps with Anypoint Controller
• If NOT using Anypoint Controller: Configuring terminationGracePeriodSeconds beyond 30 seconds is not currently supported.

Recommendations

• Set shutdownTimeout to a value that accommodates your longest expected transaction.
• Ensure terminationGracePeriodSeconds is always greater than or equal to shutdownTimeout.
• For critical flows, implement idempotent processing and reliability patterns to handle cases where transactions may be interrupted.