Chapter 5. Power Management and Fencing

The Red Hat Enterprise Virtualization environment is most flexible and resilient when power management and fencing have been configured. Power management allows the Red Hat Enterprise Virtualization Manager to control host power cycle operations, most importantly to reboot hosts on which problems have been detected. Fencing is used to isolate problem hosts from a functional Red Hat Enterprise Virtualization environment in order to prevent performance degradation. Fenced hosts can then be returned to responsive status through administrator action and be reintegrated into the environment.
Power management and fencing make use of special dedicated hardware in order to restart hosts independently of host operating systems. The Red Hat Enterprise Virtualization Manager connects to a power management devices using a network IP address or hostname. In the context of Red Hat Enterprise Virtualization, a power management device and a fencing device are the same thing.

5.1. Power Management

The Red Hat Enterprise Virtualization Manager is capable of rebooting hosts that have entered a non-operational or non-responsive state, as well as preparing to power off under-utilized hosts to save power. This functionality depends on a properly configured power management device. The Red Hat Enterprise Virtualization environment supports the following power management devices:
  • Advanced Lights Out Manager (alom).
  • American Power Conversion (apc).
  • Bladecenter.
  • Dell Remote Access Card 5 (drac5).
  • Electronic Power Switch (eps).
  • Integrated Lights Out (ilo, ilo2, ilo3).
  • Intelligent Platform Management Interface (ipmilan).
  • Remote Supervisor Adapter (rsa).
  • rsb.
  • Western Telematic, Inc (wti).
  • Cisco Unified Computing System (cisco_ucs).
In order to communicate with the listed power management devices, the Red Hat Enterprise Virtualization Manager makes use of fence agents. The Red Hat Enterprise Virtualization Manager allows administrators to configure a fence agent for the power management device in their environment with parameters the device will accept and respond to. Basic configuration options can be configured using the graphical user interface. Special configuration options can also be entered, and are passed un-parsed to the fence device. Special configuration options are specific to a given fence device, while basic configuration options are for functionalities provided by all supported power management devices. The basic functionalities provided by all power management devices are:
  • Status: check the status of the host.
  • Start: power on the host.
  • Stop: power down the host.
  • Restart: restart the host. Actually implemented as stop, wait, status, start, wait, status.
Best practice is to test the power management configuration once when initially configuring it, and occasionally after that to ensure continued functionality.
Resiliency is accomplished through properly configured power management for a Red Hat Enterprise Virtualization environment. Fencing agents allow the Red Hat Enterprise Virtualization manger to communicate with host power management devices in order to bypass the operating system on a problem host, and isolate the host from the rest its environment with a power cycle. The Manager can then reassign the SPM role(see Section 3.2, “Role: The Storage Pool Manager” ), if it was held by the problem host, and safely restart any highly available virtual machines on other hosts.