DHCP failover peers operate within server states. These states tell the DHCP failover server how to interact or not interact with its peer server, manage normal server operations, and manage operations when the two servers cannot communicate.
- Normal State—This is the standard operational state for DHCP failover servers. In this state, both servers can communicate with each other.
- Communication-Interrupted State—In this state, the servers can no longer
communicate with each other and neither server is aware of the state of its
peer. In this state, all operations assume that the other DHCP server could also
be live and issuing address leases.
After a server has entered the communication-interrupted state, it changes the way that it assigns address leases. Clients initially attempting to renew existing leases receive a new lease for the remainder of their regular lease time with the maximum client lead time (MCLT) value added. Subsequent leases are only handed out for the MCLT and clients are never given a lease renewal; they always receive a lease for a new address instead. If a client releases an address lease manually, then that address is abandoned until the server returns to the normal state.
The disadvantage of the communication-interrupted state is immediately apparent. If clients are given short (MCLT) lease times and their leases are not renewed, the address pool could quickly become depleted and there would be an increased level of network traffic.
- Partner-Down State—When a failover peer is down, you need to manually change the working peer to the Partner-Down State using the dhcp force-partn-down-state command from the Additional Configuration mode of the DNS/DHCP Server Administration Console. The remaining server then becomes the primary server, whether or not it was the primary before you use the dhcp force-partn-down-state command. The working server continues to hand out leases for the MCLT, but renews the leases. This server also reclaims all of the expired, reset, and released leases and is able to use the entire free address pool for allocations. When its partner comes back online, this server remains the primary server.