1. Related concepts
1.1, HA working principle
When we add a host to a vSphere HA cluster (after HA is enabled), the agent will be uploaded and installed on the host and configured to communicate with the agents on other hosts in the cluster. One host will be selected as the [primary host] in the cluster, and the remaining hosts as [subordinate hosts]. The cluster is enabled with vSphere HA, and all active hosts (hosts that are not in standby or maintenance mode or those that are not disconnected) will participate in the election of the cluster's preferred host.Mounting the largest number of data storage hosts have an advantage in election. Each cluster usually has only one primary host. If the primary host fails, shuts down, is in standby mode, or is removed from the cluster, the HA cluster will conduct a new primary host election.
The primary host monitors the activity of the secondary host in the cluster. This communication is carried out by exchanging network heartbeats every second . When the primary host stops receiving these heartbeats from the secondary host, it will check that the host is active before declaring that the host has failed to determine whether the secondary host is exchanging heartbeats with one of the data stores . Furthermore, the primary host will also check if the secondary host responds to ICMP pings sent to its management IP address . Host network isolation occurs when the secondary host is still running but can no longer monitor traffic from the vSphere HA agent on the management network . If the host stops monitoring this traffic, it will try to ping the cluster isolation address . If this ping also fails, the host will declare that it is isolated from the network.
VMware vCenter selects a set of preferred data save sets for heartbeat. This choice will maximize the number of hosts that have access to the heartbeat data storage, that is, these hosts in the cluster can access the storage set, which means that the storage set is distributed storage or external shared storage, butvSAN data storage cannot be used