The High Availability feature of the LoadMaster guarantees the availability of your server farm. HA is achieved by a hot-standby, failover mechanism. Two identical LoadMaster units are integrated into the network as a HA pair. One machine serves as the active LoadMaster and the second one remains in a standby, idle state, always prepared to take over the activities from the active server. This HA pair appears as a single logical unit to the DMZ side and to the server farm side connections.

With a HA pair, each network interface has an individual IP address and one shared IP address which is shared with the partner unit. The shared IP address is identical for both LoadMaster appliances, though it is associated with only the active LoadMaster at any given time.

Note: If the LoadMaster is set as the default gateway of the servers, remember to use the shared address of the HA pair since this address will always be available.

During normal operation each node periodically sends health check messages over any interface that has the Use for HA checks option enabled to verify the availability of the peer appliance. Health checks are enabled unless they have been explicitly disabled. In the event that the active LoadMaster should fail, the standby appliance will become active and take over the task of balancing.

The topology for HA single arm looks like this:

LoadMasters HA1 and HA2 use eth0 to connect to the network (firewall) and to the servers and have one shared IP address between the two ports. Optionally, eth1 on each unit is directly connected using a patch cable - the port is auto-sensing so it makes no difference if the cable is straight or reversed - and is used exclusively for additional HA health checking.

The topology for HA dual arm looks like this:

Both HA1 and HA2 use eth0 to connect to the network (firewall) and eth2 for connection to the servers. The two eth0 ports have one shared IP address and the two eth2 ports have a different shared IP address. Health checking between the two LoadMasters occurs between both eth ports. Optionally, eth1 on each unit may be directly connected using a patch cable for added HA health checking though it is quite unnecessary since there is already 2 health check routes between the HA pair.

Note: HA1 and HA2 must be on the same subnet with the same default gateway and be located within the same physical site. They must not be separated by an intra-site link and must use the same gateway to return traffic.
Note: Multiprotocol Label Switching (MPLS) networks are not supported in a HA setup.

Running HA spanned across multiple subnets will not provide hardware redundancy in the event of a failure of the link between them. If traffic balancing between multiple sites is required, the GEO LoadMaster is the correct solution. This is Progress Kemp's DNS-based appliance that employs health checking to avoid site outages. For further details, refer to the GEO Feature Description.