Most of your customers’ data centers are configured for redundancy, and to obtain this goal, most use high availability
(HA) within their virtualized environments.
By using virtual machine (VM) HA, you can make sure that if a VM goes down, it’s restarted elsewhere in the cluster. Unfortunately, this often isn’t enough for full vital services protection.
VM HA runs on your virtualization hosts to monitor if your virtual machines are still available. If a VM goes down, or if a host that runs several VMs goes down, the VM HA wakes up and makes sure that all the unavailable VMs are started elsewhere in the environment. This ensures that if there’s a problem, the service discontinuation is short.
VM HA is a solid addition to vital services availability with a customer’s data center, but it's not enough. One thing that VM HA cannot do is monitor individual services within the VM. This is a serious challenge, because in many cases service discontinuation is caused by a service that stops working and not by a complete VM disappearance. With the increased operating systems (OS) stability, complete VM disappearances are more rare.
VM HA cannot monitor services within the OS that run in the VM because the VM HA layer is OS-agnostic. Monitoring individual service availability is a task to be performed in the OS. There are products in both Windows Server 2008 that offer HA clustering, such as the Pacemaker stack on the Linux platform. This type of product monitors individual services and takes action if the service becomes unavailable -- even if the host OS keeps running.
Using an OS HA layer provides additional benefits that cannot be offered by a VM HA solution. The most important of these is service dependencies monitoring, which means an OS HA product can also restart or migrate a service if a dependency is no longer met.
Setting up an OS HA solution requires some additional investment for your customers. First, there is the HA software license cost. Open source solutions such as Pacemaker can be used without paying additional license costs, but service support for enterprise environments does cost money. With regard to hardware, there is no need for extra investment if the server instances are running as VMs. In case the server instances are running on real hardware, the only requirement is that you'll need is a second server to which resources can be migrated if the primary server goes down.
You may think that combining an OS HA solution with a VM HA solution adds unnecessary complexity because you have to manage two HA environments instead of one. But, each solution is using a different approach to realize its protection so having two HA solutions isn't a problem. Also, there normally isn’t too much management involved in having two HA solutions because they basically run themselves. In the end, the additional protection gained by having OS HA is worth it, because it ensures that failed services are automatically restarted.
About the expert
Sander van Vugt is an independent trainer and consultant living in the Netherlands. Van Vugt is an expert in Linux high availability, virtualization and performance and has completed several projects that implement all three. He is also the writer of various Linux-related books, such as Beginning the Linux Command Line, Beginning Ubuntu Server Administration and Pro Ubuntu Server Administration.