You are managing a Kubernetes cluster where one of the pods running a critical application occasionally crashes due to an unexpected bug. The application must be restarted automatically whenever it fails. However, in certain cases, the application should not be restarted, such as when the pod's container has completed successfully. You want to define a proper restart policy for this scenario. Which restart policy would be most appropriate to ensure the pod is restarted only when it fails and not when it completes successfully?