Assume the following setup:
DHCP Server 1: 10.10.3.1/24
DHCP Server 2: 10.10.3.2/24
Note that in the above image, DHCP failover is already configured. The names chosen for this example were for simplicity. The naming convention doesn't matter as long as they are identical under each service (the service names will vary). Be sure to name the failover objects to something meaningful. Later on, when monitoring logs, tracking down issues or watching the failover occur, the troubleshooting and monitoring will be much easier to follow. Each service object contains a subnet of 10.10.3.0, and a pool (called test_pool) that match each other. This is a necessary step for any pool to be included in the failover process. If it is not included in the failover process, then it isn't necessary. For example, if I include subnet 192.168.1.0 in myService2, I don't have to have it under myService1 if it is not to failover.
- Create two new SERVICE objects. The objects must be unique in name. In the above example they are myService1 and myService2. When prompted to assign a primary server, I chose serverA for myService1 and serverB for myService2.
- Highlight each SERVICE object and create a subnet and pool (sometimes referred to as range). Be sure to make an identical subnet and pool under the second service object. The object names must be the same. In the above example, the subnet under myService1 is called 10.10.3.0. It is identical under myService2. The pool name under both subnets (again, which fall under different SERVICES) is test_pool. They are identically configured in options in addition to the naming convention.
- Create a FAILOVER object under each service. In the above example they are both called myFailoverObject. You can name them whatever you like but they must be exactly the same as each other just like in the example above. While configuring the FAILOVER object options, configure both to be identical (especially when specifying the primary and secondary servers).
- More information can be found on creating a Failover peer by referencing section 7.1.11 of the DHCP documentation.
- The same port can be used for the primary and secondary servers. As the RFC does not currently specify a reserved port for DHCP Failover Peer setups, specify a port(s) of your choice that is/are unused. Be sure that the firewall is not blocking the chosen ports.
- Edit each of the pools and choose the only available FAILOVER PEER from the drop down list.
- Restart DHCP and monitor /var/log/messages for any errors. Stop and start the dhcp server in different orders on the two servers. Release and renew addresses from clients. Watch how the communication between the two servers behave and how the failovers occur. This is where the naming convention of the failover objects pays off.