beawolf - Fotolia
Host server maintenance windows are an important consideration for a heavily virtualized organization. Like any other type of server, virtualization hosts require periodic maintenance, and a maintenance window gives IT staff the opportunity to take care of any pressing maintenance tasks. However, setting up a maintenance window is somewhat of an art form.
Plan for maintenance windows
There are a number of considerations that should be taken into account when planning for a maintenance window, but most of these considerations are in at least some way related to the organization's service-level agreements. Simply put, administrators must schedule maintenance windows to avoid business disruption.
One such consideration is whether maintenance windows will occur at regularly scheduled intervals or if they will be created on an ad hoc basis. Both approaches have their advantages. Ad hoc servicing usually results in fewer maintenance windows. This approach might be less disruptive to the end user.
Conversely, regularly occurring maintenance windows can be a good option because they're predictable. If everyone in the company knows, for example, that IT does maintenance at 9 p.m. the first Sunday of each month, then users will likely avoid scheduling critical tasks at a time that coincides with the maintenance window.
Proper maintenance window use
Although scheduling is probably the most obvious consideration related to hypervisor maintenance windows, it isn't the only consideration. Another important consideration is how to use the maintenance window. On the surface, this seems simple. After all, the maintenance window is a block of time during which IT is free to perform required maintenance on the host servers. Even so, IT must still use the maintenance window in a responsible manner.
Suppose, for instance, that a particular maintenance task required individual host servers to be taken offline for a period of time. Although the maintenance window grants the IT department the freedom to perform the required maintenance, it would be irresponsible to shut down all of the host servers at the same time.
A mass shutdown would affect mission-critical workloads that need to remain online. Just for the sake of argument, however, let's pretend that there aren't any workloads that have to remain online. Even in this situation, performing a maintenance task simultaneously across all host servers is a bad idea. If something were to go wrong during the maintenance operation, then every host server could potentially be affected by the problem, and there might not be enough time to fix it before the end of the maintenance window.
A better approach would be to perform maintenance on a subset of the host servers. In doing so, you should ideally be able to leave enough hosts online to continue running important workloads and be able to perform a failover should one of the remaining hosts experience an unrelated problem during the maintenance window.
Avoid disruption by reserving maintenance windows
Some organizations have actually created formalized policies that stipulate the activities that should and shouldn't be performed during a normal maintenance window. Such an organization might, for example, reserve the normal maintenance window for minimally disruptive tasks, such as patch management, storage volume expansion or making changes to load balancer settings.
Organizations might need to perform more time-consuming or potentially risky tasks at a time that has been set aside specifically for those tasks. Examples of such tasks might include dynamic disk storage reclamation, major architectural changes or version upgrades to the hypervisor or to support software. These tasks might not be performed during the normal maintenance window because these types of tasks tend to be more time-consuming than the more basic maintenance tasks. Rushing through a major maintenance task in an effort to complete the task during a maintenance window, which is normally relatively short, can lead to mistakes being made.
Ultimately, there's no right or wrong way to set up a virtualization maintenance window. Each organization has its own requirements, and the maintenance window needs to be defined in a way that meets those requirements, while also giving the IT department the ability to complete key maintenance tasks.
Follow this server maintenance checklist
Use System Center Virtual Machine Manager maintenance mode to apply updates
Oversee application maintenance