Virtual machine monitoring can determine whether a VM is not functioning through a heartbeat exchange between the host server and the virtual machine’s VMware Tools instance. This is performed every 30 seconds. If this exchange does not take place, the virtual machine is rebooted. From the service provider perspective, this might not be something you want to occur. When enabled, a virtual machine is declared failed if the heartbeat inside the guest operating system tools is no longer received, and there is no network or storage I/O observed.

To detect an application crash, VMware provides an application recognition API that allows software vendors and in-house application developers to develop solutions that are monitored for health and that tell vSphere HA to reboot a virtual machine if a running application fails. For more information about the VMware application recognition API, see the VMware vSphere Guest SDK Documentation at https://www.vmware.com/support/developer/guest-sdk/.

Enable virtual machine monitoring only if the application or service running on the virtual machine is able to recover after a virtual machine restart. Disable virtual machine monitoring on a virtual machine when it is already protected by either VMware vSphere Fault Tolerance or Microsoft Cluster Services.