4.5.3 High availability
Load Balancer provides a built-in high availability function. It allows you to configure a backup Load Balancer server; if the primary Load Balancer server fails, the backup server will take over the load balancing for all clusters.
The two Load Balancer servers need connectivity to the same clients and to the same cluster of servers, as well as connectivity between themselves. Both Load Balancer servers must be running the same operating systems, and they must be connected to the same network.
The two Load Balancer servers are referred to as the primary server and the backup server. These are the roles that are associated with each server during the configuration.
Load Balancer servers run in a specific state: one server is active and the other server is standby. This means that the Load Balancer server that is in active state is the one that is distributing the workload.
The Load Balancer server that is in standby state is monitoring the active one. If the active server fails, the standby server performs a failover; it switches to active state and starts load balancing the cluster. So the state of a server changes when a failure occurs, but the roles do not change during a failure.
You can also configure the backup server to respond to other clusters' addresses while it works as a standby server for the clusters on the primary server.
When the primary server is operational again, but in standby state, you can either decide that it automatically forces a takeover and it becomes the active server again, or it stays in standby mode, monitoring the other server for a failure. In this last case, you will have to manually force a takeover if you later want it to become the active server again. This is known as auto recovery or manual recovery.
In order to monitor the health of the active server, heartbeats are sent every half second. The failover occurs when the standby server receives no response from the active server within two seconds. Another possible reason for a failover is when the standby server is able to ping more reach targets than the primary machine. A reach target is a server out of the load balancing environment that is used for an external connectivity test. We usually recommend using the default gateway of the network as the reach target of your configuration.
By default, high availability is only supported for the Dispatcher component, not for the Site Selector.
For a highly available Site Selector, you should configure Dispatcher to host a cluster in front of multiple Site Selectors. The cluster will act as the sitename DNS clients are requesting. The Dispatcher tier can provide the failover mechanism and the Site Selectors will then have no single point of failure. The DNS advisor to be used to detect a failed site selector is included with the product.