High Availability cluster: what it is and how it works

Modern business cannot afford the inaccessibility of IT infrastructure for a long time. Often even a minute of downtime entails serious reputational and financial losses. It is for this reason that services and applications, especially those that are sensitive to availability loss, are deployed on high availability cluster-based server solutions.

Let us start with the terminology.

High Availability cluster (HA-cluster, high availability cluster) is a group of several servers designed in accordance with methods of high availability and guaranteeing minimum system downtime thanks to hardware redundancy.

In simple terms, if one server in the cluster fails, its tasks are distributed to machines whose performance is not affected. This allows not only to avoid inaccessibility of systems in case of equipment failure, but also to carry out maintenance or any other routine work without stopping the IT systems.

Let us imagine that a failure occurred on the server where the business application is deployed.
If there is no cluster, the business application will be unavailable until the server is restored. If the company has taken the effort to set up an HA cluster, the application will automatically restart on intact nodes.

Obviously, at least two nodes are required to set up an HA cluster. This configuration is accordingly called a two-node configuration. However, in practice there are different implementation schemes of High Availability clusters. Let us consider the most common ones.

This is one of the most popular HA cluster implementation schemes. If one of the servers in an active/active cluster becomes unavailable, its traffic will be redistributed among the remaining nodes. This method is relevant in situations where the cluster nodes are homogeneous in terms of hardware and software and perform the same tasks.

Each node in the cluster is redundant. The redundant node is activated only in case of failure of the main node. It is no surprise that the active/passive scheme requires considerable redundancy.

The cluster has one backup node, which takes over the tasks of the failed server. After restoring the equipment, the node becomes backup again. The N+1 scheme can be implemented in a cluster with heterogeneous tasks, and the "spare" node will have to be able to replace any machine in the cluster.

If one backup server is not enough due to the heterogeneity of services and tasks that the servers in the cluster perform, you can use the N+M scheme. In contrast to the previous one, it implies that there will be more than one backup machine in the cluster. This way of building an HA cluster makes it possible to achieve a compromise between the cost of the solution and the required level of reliability.

The most expensive of the schemes discussed above is the active/passive scheme. In this way of implementing a High Availability cluster, the company literally has to pay double the cost of running equipment.

So, what does it take to build an HA cluster?

1. Resource redundancy
Free computing resources are required to move the task from the failed server to a functional one. A simple example: if a cluster consists of 5 servers, the load in the cluster should not be more than 4 machines can handle.

2. Sharing the data of each node in the cluster
The key principle of a High Availability cluster is to distribute the tasks of a failed server among the remaining servers. Accordingly, when implementing an HA cluster, it is necessary to provide not only duplication of capacity, but also redundancy of channels to access data and end users.

3. Software capability
Not every application will be able to run on an HA cluster. This feature should be incorporated at the stage of building the architecture and software development. In particular, applications that are planned to be deployed within an HA cluster must be able to:

ISPsystem's VMmanager 6 platform allows you to create HA clusters. If one of the cluster nodes becomes unavailable, VMmanager will automatically move virtual machines to the intact nodes, stop the VM on the failed node and isolate it.

Learn more about creating HA clusters in VMmanager and their logic in the platform documentation.