Chapter 4. Configuration
This chapter introduces the Turbolinux Cluster LoadBalancer 10 configuration tools and provides instructions on how you can configure your cluster. Due to the flexibility of Turbolinux Cluster LoadBalancer 10 and the different needs that it can address, there is no standard or default configuration. Each cluster will be configured differently, depending upon how it is used and the resources available to build it. This chapter will teach you how to use the configuration tools and provide the detailed knowledge necessary to design and configure a working cluster.
The topics covered in this chapter are:
Planning the design
Configuration tool overview
Services
Servers
Advanced Traffic Managers
Clusters
Global Settings
 | After cluster system configuration, there are two daemons which need to be started: clusterserverd and cmcd. To start/stop a daemon, execute /etc/init.d/clusterserverd and /etc/init.d/cmcd in the same method that you start other daemons. |
4.1. Planning the Design
It is very important to plan the design of your cluster before you begin configuring it. You will need to decide which systems will be your ATMs and which will be your nodes. You may also need to decide which services you want to run on the cluster, although that decision has probably already been made before the purchase of the Turbolinux Cluster LoadBalancer 10 software.
One helpful tool that may assist you in designing your cluster is to draw a diagram of the cluster. Typically you put ATMs at the top of the diagram, with nodes hanging down from the ATMs. This represents the fact that traffic must flow through the primary ATM before reaching the cluster nodes. You may want to include host names and IP addresses in your cluster diagrams to make it clear which server is which. (A single system can be both an ATM and a node. It's best to put these dual-purpose systems at the top of the diagram when trying to conceptualize the cluster.)
It is recommended that you start out with a simple cluster and work your way up to a more complex configuration, if that is what you require. A smaller cluster is easier to troubleshoot, and you can more easily debug any problems that come up. After you have set up the small cluster and tested it, you can incrementally add ATMs and nodes until you reach your desired configuration. Turbolinux Cluster LoadBalancer 10 provides you with this flexibility due to its scalability features.
4.1.1. Typical Scenarios
Most clusters will probably be used as web servers. Web traffic has increased substantially in the past several years, and many sites need added reliability, availability, and scalability. While web servers are a prime candidate for clustering, other services can benefit just as well. The design of a cluster is based more on performance considerations and hardware availability than the particular service that will be running on the cluster.
This chapter will walk you through the steps required to configure a cluster. The cluster used as an example is of medium complexity, covering all the basic configuration options. It will not be a typical configuration, because we need to cover every available option.
4.1.1.1. Small Cluster
The smallest cluster possible consists of only two systems. One system will act as the primary ATM and as a cluster node. The other system will act as a node and probably as a backup ATM. By having each system work as both an ATM and a node, you make the most effective use of the resources available.
In the example below, we have configured one system as the ATM and another as a node. Note that the node can send reply traffic without going through the ATM.
4.1.1.2. Larger Cluster
Larger clusters will dedicate one or two systems as ATMs, with all other systems acting as dedicated cluster nodes. The primary ATM will handle all the traffic forwarding and nothing else. The backup ATM will sit idle unless it detects that the primary ATM has gone down. The cluster nodes will handle only the services.
Below is an example of a larger diagram. Note that Node 4 must use tunneling, because it is not on the same subnet as the ATM.
4.1.1.3. Complex Cluster
You can come up with some pretty complicated cluster designs. Turbolinux Cluster LoadBalancer 10 allows you to define multiple cluster addresses and put different sets of cluster nodes in each virtual cluster address. You can even run multiple clusters that share nodes but use different sets of ATMs.
It is recommended that you read through this manual before designing a complex cluster. You should also start by configuring a simple cluster to get a feel for your options. You don't want to start out with something too difficult and get in over your head.