Hello Arturo,
what you would like to achieve is not an easy task: a router can be configured to act as a DHCP relay or as a DHCP server or both, but in the latter case the router would compete with the main DHCP server.
Ideally, this kind of things would require the use of a EEM TCL script that should react to the failure of a test on main DHCP server reachability by modifying the configuration of the router to add the DHCP pool.
A simpler solution may be provided by the use of a secondary DHCP server running on a workstation: the router could be configured to perform DHCP relay agent function for both servers and this is achieved by simply having two ip helper-address command on L3 interfaces facing clients.
_The secondary server should be configured to wait some time before answering the received DHCP requests, in this way it would play a role only when the primary DHCP server fails.
Clearly this solution would lack a way to pass DHCP leases between servers so the lite DHCP failover.
Some DHCP servers can be deployed in a clustered mode, for example Cisco network registrar I have seen are able to do this and to have a primary DHCP server per IP subnet.
I guess some other DHCP SW also open source can do this but I haven't direct experience of this.
Between two cisco network devices they can share a DHCP lease database hosted on an external file server (TFP or FTP)
ip dhcp database this is used also by DHCP snooping feature on multilayer switches.
see
http://www.cisco.com/en/US/docs/ios/ipaddr/command/reference/iad_dhc1.html#wp1016399
Edit:
another useful command to have the router DHCP to wait long time should be the following:
http://www.cisco.com/en/US/docs/ios/ipaddr/command/reference/iad_dhc2.html#wp1052493
ip dhcp-server query lease {retries number | timeout seconds}
by using the timeout 60 option the router should wait 60 seconds before answering giving time to primary DHCP server to answer before it.
This may provide a simple solution clearly without state exchange.
Hope to help
Giuseppe
