Various systems and methods can provide high availability of an application executing in a highly-available virtual machine environment. One method involves receiving information indicating a state of an application executing in a virtual machine from a monitoring agent executing in the virtual machine. In response to receiving the information, the method involves determining whether the virtual machine should be restarted. Based upon that determination, the method then determines whether the monitoring agent should send a heartbeat message to a virtualization controller prior to expiration of a timeout interval. The virtualization controller is configured to restart the virtual machine if the virtual machine does not send the heartbeat message prior to expiration of the timeout interval.