Product SiteDocumentation Site

3. High Availability - Host Initiated Reboot

At this stage you now have six running virtual machines, three of which are highly available. This section simulates an event in which a user error has caused a temporary host failure, after which the host recovers, but the virtual machines which were running on the host are terminated. In this case, the host is manually fenced before being placed into maintenance mode.
To demonstrate virtual machine high availability when host is incorrectly fenced
  1. On the Tree pane, select VMs. The available virtual machines display on the Virtual Machines tab.
    Virtual machines running on hosts
    Figure 46. Virtual machines running on hosts

    In this example, there are currently two virtual machines running on the Atlantic host; and another four on Pacific. You can check whether a virtual machine is highly available by selecting it looking at the General tab of the details pane. Here, on the Atlantic host RHEL6Thames is set to be highly available, while RHEL6Congo is not.
  2. On the Tree pane, select Hosts. The available hosts display on the Hosts tab. Select the Atlantic host, click the Power Management drop-down menu and select Restart.
    Restart host
    Figure 47. Restart host

    The Restart Host(s) dialog displays, click OK to confirm and proceed. The host's status changes to Reboot, then Non-Responsive.
    Non-responsive host
    Figure 48. Non-responsive host

    You have now simulated an environment where a host is manually fenced before it was placed into maintenance. Since power management has been configured for this host, it will automatically reboot after a short period.
  3. While the host is being restarted, observe what has happened to the virtual machines which were running on it. On the Tree pane, click VMs to display the Virtual Machines tab. Notice that both the virtual machines running on Atlantic were shut down as soon as the host was restarted.
    Virtual machines starting on another host
    Figure 49. Virtual machines starting on another host

    The highly available virtual machine, RHEL6Thames, is automatically restarted. Its status changes from Down to Wait for Launch, and then to Powering Up. It runs on the Pacific host in the interim period while the Atlantic host is still rebooting. In contrast, RHEL6Congo remains turned off, its status displays as Down.
You have just run a demonstration where a host was manually fenced before the virtual machines on it were properly shut down, causing the virtual machines to crash. When virtual machines are not properly stopped, only the highly available ones are restarted on another host in the cluster. In contrast, the non-highly available machines remain powered down until they are manually restarted.