VMware HA Incorrectly Handles Temporary Host Network Isolation

In VMware HA, host network isolation detection occurs within 15 seconds of host isolation. After a failure is detected, all virtual machines are failed over to other hosts. If the network connection is restored within 12 seconds, other hosts in the cluster do not treat this as a host failure, and the virtual machines remain powered on (on their original host). After 12 seconds, the clustering service on the isolated host shuts down and the virtual machines are powered off.

 

If the network connection is restored shortly after 12 seconds, the virtual machines are not started on other hosts because the original host is not considered to be isolated. Thus, the virtual machines are powered off but not failed over.