Enable vSphere HA on a cluster to ensure high availability of a VMware vSphere 6.7 infrastructure

11. Testing vSphere HA (isolated host - no HA response configured)

For this test, we'll show you what happens when a host is considered isolated.
However, initially, we will not configure a response from vSphere HA when this isolation issue occurs.

In summary, you will notice that the virtual machine located on the isolated host will still be functional, despite the host not being accessible (via its management network).

For this test, we therefore used 3 standard virtual switches:

• vSwitch0: for the host management network and vSphere HA agent communication (since these use the host management network).
• vSwitch1: for network access of the host's virtual machines.
• vSwitch2: for access to shared storage located on an iSCSI server accessible by all hosts in our cluster.

Thus, we will be able to cut off the network access for the host management network without cutting the network connection of the virtual machines nor for the connection for access to the iSCSI server.
Since the management network will no longer function, the host will be considered isolated by vSphere HA.

If you wish to reproduce the test environment for access to iSCSI storage used in this case, refer to our tutorial: VMware vSphere 6.7 - Configure port binding (iSCSI traffic).

As before, our "VCSA" virtual machine is running on our "esxi1" host.

As before, our "Win 10 x64 v2004" virtual machine is running on our "esxi2" host.
Thus, we will be able to cause a host isolation problem on this host "esxi2" without having a problem with the vSphere Client hosted by our virtual machine "VCSA" running on host "esxi1".

Important : if you want to perform this test at home, make sure you can ping this virtual machine.
If not, make sure Windows is using a private network or that ping (ICMP) is allowed in its firewall.

On our 2nd host we also used 3 standard virtual switches for the same reasons.

For this test, we disconnect the network adapters used by our virtual switch "vSwitch0" (used for the host management network).

In the vSphere Client, we see that the vSphere HA status of our host "esxi2" is "HA Agent Unreachable".

Then, messages will appear for this host.

Plain Text

vSphere HA host status.
vSphere HA has detected that this host is in a different network partition than the master to which vCenter Server is connected.

You will also see that the vSphere HA status of this host is "Partitioned Network".

If you deploy the "vSphere HA Status" node, you will see in particular that the vSphere HA agent of this host will apply the isolation response for each of its VMs if this is configured on your cluster and on them (in the case where a VM replacement is configured for them).

A little later, the vSphere HA status may change to "Isolated Network".

On your virtual machine page, you will see that the affected cluster and host are in error.

Note that the current IP address of our Windows 10 virtual machine is "10.0.0.60".

On your cluster page, you will see several messages appear.
Including problems for the host concerned, given that its management network no longer works.
You will also see issues with insufficient failover resources, as vSphere HA will no longer be able to keep your virtual machines available in the event another host goes down.

As you can see, although the host management network where our Windows 10 virtual machine is located is inaccessible, we can still ping it.
Which proves that this virtual machine continues to operate normally, despite the failure of its host's management network.

A little later, the messages for the host whose management network you cut off may change.
You will see this one in particular.

Plain Text

vSphere HA has detected that this host is network isolated from the cluster.

Reconnect the network adapters used for this host's management network.

Wait a bit and the error and warning messages will disappear.
As expected, our host "esxi2" is working again.

The same goes for our cluster.

Our virtual machine continues to run on our host "esxi2" even though it was previously considered isolated.
Indeed, vSphere HA did nothing, because no action had been selected in our cluster settings for response in the event of a host isolation.