Creating a Medium Availability Cluster.

Some time back my colleagues and I needed to build an example usage for server automation product Cassatt Collage (since renamed Cassatt Active Response ). It needed to be relatively portable, and yet show off as many of the bells and whistles as possible.

What we decided to do was build a “Medium Availability Cluster“.
(It appears that “Medium Availability Cluster” appears to be a term we coined.)

A Medium Availability Cluster is just a term we used to describe a configuration of servers that is able to automatically recover from a failure in hardware (or software) with minimal downtime automatically. This is not a system that never skips a beat, this is a compromise solution that will fail briefly whilst it recovers.

The idea of the build was to show the power and benefits of server automation. We were not trying to show the power management or load balancing functionality that the Cassatt products offer. It was a simple example of how Disaster Recovery could be automated by software.

So how to create a medium availablity cluster…

To start with, build up a server and install your server management/automation software (Cassatt, OpenQRM, Puppet, etc.)

Once this is up and running use the included tools to create your “images” for your MySQL, Apache tiers.

In the example build we created, it was 2 MySQL servers, 2 Apache servers and a load balancer.
MySQL was configured to be a MySQL Cluster, Each Apache server was a basic Apache build with a simple PHP application running. We added the load balancer also to give a single point of failure.

A more sophisticated configuration that is particularly relevant to smaller sites, is to utilise Virtualisation via VMware, Xen, KVM, etc. Then each server in the tiers is in fact a virtual server on a physical host. You can include the host server (such as VMware ESX), inside your automation software configuration also.

Depending on how you do all this, you will want some spare physical servers included in the mix so you can cope with the hardware failures.

Okay, once it is built, time to play!