Structure & Function

At minimum, you require two cluster nodes. For our example, using SteelEye LifeKeeper, these nodes do not need to be the same as far as their hardware configuration is concerned.

To achieve the planned level of high availability, it's necessary to use stable hardware components that can withstand a 24x7 duty cycle. The systems should have redundant components for all system critical points, especially fans, power supplies, hard disks and so on. Every node should have, at minimum, two LAN interfaces. It is also important to have a stable solution to fix all the cables, power lines and all the small peripherals to reduce the likelyhood of crashes due to damage / removal of these components.

After mounting the systems, the first step will be to install and to configure the operating system. This includes inserting the nodes into the LAN environment, and introducing them to to the name servers and all other network services. After that, the core component of the cluster software will be installed on all systems.

After the configuration of the systems, the next step will be the connection to the shared storage or the SAN. This connection should be redundant. LifeKeeper supports many multi path configurations. It is very important for the cluster software to recognize the real name of the volumes using different ways to attach to it. This job is done by special `multi path` drivers. This part of the configuration should be an important point for the support assessment. During the installation, access to the shared storage will be able from all nodes at the same time.

The next step will be the installation and configuration of the heartbeat connections. The heartbeat is necessary to exchange cluster information between all members of the cluster. The heartbeat connection is designed to be redundant. LifeKeeper will accept LAN or serial connections, and a separate volume as a heartbeat.

It is typical to create a private LAN connection between the cluster nodes as the first connection. It's recomended to install a second connection with the serial interface, or with the use of the public LAN. The advantage of a serial connection is that it uses a different communication stack, and brings a higher level of safety. Because there isn't much information, the bandwidth of a serial connection is enough.

If all heartbeat connections are out of order, a critical state known as the "split brain" would occur. During this state, the cluster nodes would bring the same applications in service at the same time. LifeKeeper's cluster solution includes roles to solve this critical situation however.

After installation and configuration of the cluster nodes, the next step will be to connect the clients to the cluster. An important point is to be sure that from the clients, all cluster nodes will have a connection that is redundant, and is free from communication problems. To use an application, the client has to connect to a logical address which is different from application to application. From the view of the clients, it isn't possible to identify the physical node they are working with. The cluster works transparent to the clients, so specialised client software isn't necessary.