The first sentence in the slide notes states:
Datastores are used as a backup communication channel to detect virtual machine and host heartbeats.Correct:
The highlighted portion of the above sentence is very ambiguous. To me, it seems to imply that the virtual machines heartbeat mechanism (provided by the VMware Tools) can failover to a datastore heartbeat-like capability.
This is - AFAIK - a nonsense because:
- Virtual machine heartbeats are "raised" by the VMware Tools in the guest OS and "catched" internally by hostd in the ESXi host running that VM, without using the network. Thanks to this capability, virtual machine heartbeats - and VM Monitoring, in vSphere HA - work with any VM, regardless there is a configured vNIC or not.
As there is no networking resource failing here, there is also no need to have a backup mechanism based on the datastore(s).
- I've never heard anything about such a feature anywhere, neither on the official VMware documentation, nor from any VCDX's or vExpert's blog, not even a rumor since when virtual machines heartbeats where introduced ages ago.
The above mentioned sentence is likely to be simply a misleading statement.
In vCenter Server 4.0, VMware introduced an additional disk I/0 status check, after the virtual machine heartbeat fails, to prevent accidental VM resets. This check ensures to wait for 120 seconds (by default) of no disk I/O activity before the reset takes place. The das.iostatsInterval advanced HA attribute can further customize the time the host is going to wait before actually resetting the VM.
This disk I/0 status check might be what the slide notes are referring to as "a backup communication channel to detect virtual machine heartbeats".
VMware Knowledge Base articles.