Virtual Machine and Host High Availability - Getting Started with Red Hat Enterprise Virtualization - page 111

The host's high availability
The Red Hat Enterprise Virtualization manager constantly monitors the status and health of
hypervisor hosts using heartbeat messages over the management network called RHEV-M .
There are two main error states that the hypervisor host will turn on.
If the host state changes to Non Responsive , it clearly indicates that the hypervisor lost
communication with the manager. Under such a scenario, the RHEV Manager tries to re-
cover the nonresponsive host by initiating the fence operation using power management
options. For fencing to work, you need at least one active host in the cluster or in the data
center.
The RHEV Manager does not directly connect to the fencing device to perform power-
management operations. Instead, the RHEV Manager chooses one of the active hypervisors
in the cluster as a proxy in order to perform the fence command.
After the successful fence operation, all the virtual machines that are marked as highly
available and were earlier running on the troubled host will get restarted in other active
hosts in the same cluster.
Once the fenced host turns to the Up state after the reboot operation, the cluster policy
kicks in to load balance the virtual workloads on a newly rebooted host. This is where the
RHEV cluster policy will be greatly helpful for you to load balance your virtual workloads.
If the fenced host does not come up in the prescribed time, then the host status changes to
Non Responsive , and it needs manual intervention for further troubleshooting.
Sometimes, the host state changes to Non Operational ; this is mainly due to the incorrect
configuration of hosts, such as missing logical networks for that host or due to the inability
to access the storage domains or LUNs.
To troubleshoot such scenarios, please refer to the Events tab on the bottom pane of the
Manager Administrator console for detailed logs.