How do you prevent host isolation during network maintenance?

It's common to encounter host isolation problems when performing network maintenance, but you can avoid this issue by reconfiguring your VMware high availability cluster settings.

The hosts in your high availability-enabled cluster use a heartbeat network to determine the presence of hosts. If the master host can no longer send heartbeats and the slaves aren't receiving heartbeats or master-election traffic, then the hosts consider themselves isolated.

To prevent host isolation during network maintenance, ensure the hosts are aware of and temporarily stop the heartbeat process. There is a toggle switch for this in the cluster's settings that enables and disables host monitoring. Disable this setting during network maintenance.

Unwanted host isolation might also result if your hosts can't send a ping -- an internet control message protocol (ICMP) request -- to the isolation address or receive a response. The hosts use the default address of the default gateway for the VMkernel.

When that router doesn't respond to ICMP requests, your hosts will consider themselves isolated, even if they're still connected to the network. In this case, you need to configure the VMware high availability cluster to specify another address as an isolation address.

VMware documents a variety of parameters and possible values for a VMware high availability cluster.

It's a best practice to not rely on a single isolation address. Instead, configure an additional address on each of your management networks. This will minimize the chance your host will report host isolation.

It's also a good idea to make sure the heartbeat network itself is redundant. The easiest way to do this is to combine network interface card teaming with at least two adapters for your management VMkernel port.

You can also create an additional VMkernel network port on a second virtual switch. With this last option, the high availability feature will send additional heartbeats on that network.