The requirements for configuring oVirt high availability are minimal, but critical. Without power management, there can be no high availability.
As the open source platform
What to expect from oVirt high availability
More resources on oVirt 3.1
Getting your oVirt 3.1 environment ready for production
What's the difference between oVirt and RHEV?
High availability in oVirt 3.1 ensures you can easily restart virtual machines when the host platform goes down. This lets administrators build an infrastructure with minimal down time. To build oVirt high availability, you need a minimum of two hosts, as well as a shared storage platform. You also need to configure power management on the hosts.
In a high availability environment, you must be able to guarantee the integrity of VMs. The high availability stack monitors all hosts. If it detects an unreachable host, then it will start VMs on one of the remaining hosts.
But oVirt must first ensure that the unreachable host has actually failed. In the event a network cable between hosts breaks and the storage infrastructure remains intact, two hosts may end up running the VMs simultaneously. This would result in data corruption if the VMs perform different actions on the attached storage. Configuring power management on every host will prevent this scenario.
No high availability without power management
Power management is often an external mechanism that the virtualization cluster accesses to shut down a failing host. In this case, oVirt 3.1 communicates with hardware. This could be a blade center, an APC power switch or embedded management cards in the server, such as HP iLO Management Engine or Dell Remote Access Card. OVirt sends a command to the management card, and the cluster can hard-shut down a failing host.
If you don't configure your server for power management, you cannot switch on high availability. The hosts in the HA cluster also must have enough resources to run all VMs in the event one host fails. You need to calculate the amount of RAM needed by all VMs and ensure that the total amount of RAM is available after a host fails.
If you need 40 GB of RAM to run all VMs, and you have three hosts, then you should have 32 GB of RAM per host. This ensures that all VMs have enough resources to run if one host goes down. The same calculation applies to CPU resources: The remaining hosts must have enough CPU cycles available to accommodate all VMs.
With power management enabled, you can easily configure the VMs for high availability by clustering hosts together and adding VM files on shared storage. Select the VM's highly available property upon creation, or use the edit button to choose high availability once you create the VM.
Enabling high availability also enables automatic monitoring. If a VM or a host fails, the HA stack will quickly restart all affected VMs on another host.
When you design an oVirt 3.1 virtual network, be mindful of high availability requirements. You should have a dedicated network for HA and other management traffic, which exists as the Management Network. You should also isolate HA protocol traffic from users accessing VMs. This ensures that the HA stack continues operating, even if user traffic to VMs reaches a peak.
This was first published in May 2013