A virtualization infrastructure requires maintenance independent of scheduled, physical hardware reboots, so virtualization...
administrators should create an IT maintenance plan to ensure the regular deployment of essential updates and reboots.
After implementing new technology, admins must shift from installation and configuration to maintenance. Physical server maintenance plans revolve around the hardware's scheduled reboots. The care of a virtual environment, as opposed to a physical one, calls for a different IT maintenance plan.
Hardware-based OSes crash more often without regular patches and reboots due to the sheer number of drivers necessary to connect the OSes to hardware platforms. Virtualization replaces a lot of those drivers with a streamlined set of fundamental drivers that aren't hardware-dependent. This often causes VMs to run longer between reboots.
Running longer changes how admins look at their virtual environments. Once admins incorporate high availability and vMotion, it becomes temptingly easy to delay maintenance -- but that's a disaster waiting to happen.
Include hosts in an IT maintenance plan
Hypervisor code is lean -- not nearly as large as a traditional OS -- so they require less maintenance. The importance of a hypervisor, however, is much greater than in the traditional OS, so it demands the attention of a detailed IT maintenance plan that includes a strict reboot schedule.
A weekly reboot isn't necessary, but a reboot every few months isn't out of the question. Part of the reason for this is to address memory leaks and runaway processes. Though the hypervisor is streamlined code, all code can still have minor issues. The host reboot ensures that memory leaks or other minor issues don't turn into major problems that cause outages.
A reboot schedule isn't a software patch installation window. Though admins could apply necessary patches in this window, the purpose of the reboot schedule is to keep the hypervisor running at peak efficiency. Admins should design a schedule based on that need and not on software installations. Without planning, patching can replace the true intention of the reboot schedule.
Host reboots can help keep your hosts in shape, but it's also important to maintain the virtual guests and the tools installed by the hypervisor. Best practice is to reboot virtual guests fairly often based on traditional patching for the OS. This might mean a reboot every month or every few months for a full Windows instance.
Maintain virtualization tools and virtual hardware levels
The hypervisor has two guest-facing software aspects when it comes to VMs. The first piece of software is the virtual hardware layer. This isn't the drivers, but the version of virtualized hardware the VM sits on top of. The virtual hardware platform typically changes with each new major revision of the hypervisor and needs to be updated before the tools or drivers.
The downside to the hardware update process is it often requires the VM to be off. The drivers installed in the guest OS for the hardware functions of the VMs typically consist of keyboard, video, mouse and other standard drivers that help the VM operate smoothly. The driver version moves up in version as the hypervisor updates.
Unlike non-disruptive hypervisor updates, each guest needs to be updated to the new tools and virtual hardware. This is time-consuming and may require a guest reboot. Admins should plan the virtual tools and hardware alongside traditional maintenance reboots of the virtual guest to minimize outages.
Many organizations don't consider host and guest reboots critical aspects of the virtualized data center. Maintenance, done correctly and in a timely manner, will go unnoticed. When organizations delay or skip maintenance, minor problems can become outages. It's all too easy to forget how quickly this can occur without an IT maintenance plan that includes hosts and guests.