I have set the DHCP failover clustering up, using iSCSI target 3.3 and win2008R2 servers (3) The whole instalallation was smoothly and i have been able to publish DHCP service and final users are working with it succesfully. The problem now is the testing, it looks like the service is only hosted in one of the two DHCP servers because DHCP1 server has it's own IP address and the Virtual cluster IP address, and DHCP2 has it's own IP, and cluster's admin IP. So, if i put DHCP2 offline the service runs ok. But if DHCP1 is offline nothing works, i mean, final users don't receive IP, and i can't open the failover clustering admin tool from the DHCP2 saying that the cluster is not reachable. From the admin failover cluster i have changed the actual owner to DHCP2, but the results are the same now when DHCP2 is down. So, i'm not sure if i understood as well the failover clustering concept, if this is the normal behavior, or if may be will be necessary to do more configurations to get the full failover functionality. What i'm looking for is to give final users DHCP high availability in the case of a disaster for one of the two DHCP servers. I appreciate any guidance.

I would have recommended two DHCP servers splitting scopes in a 50/50 design, but since you proceeded with the clustering setup...it sounds like a cluster configuration issue. Just manually taking the resource offline is not going to force a fail-over the resources in a cluster. You probably should have done some live tests with this cluster prior to going into production such as simulating network connectivity loss for one system, etc... You may want to post specific questions with regard to the clusting configuration in the HA forum. Clustering Forum http://social.technet.microsoft.com/Forums/en-US/exchangesvravailabilityandisasterrecovery/threads Other resources... Load Balancing DHCP Using Split Scopes (50/50) http://www.anitkb.com/2011/05/load-balancing-dhcp-using-split-scopes.html Clustering is great and it has its specific use. However, I always caution the introduction of complexity into the design. Many times, complexity will decrease reliability. Visit: anITKB.com, an IT Knowledge Base.

Thanks JM, one of the most important reasons to decide not to use the 50/50 is that customer works with 25 scopes and each one has more than 15 reservations by MAC and some exclusions per each scope. And this is not very practical in case of have to switch from DHCP server. I will reach the links above hoping to find the best solution for my current scenario. Thanks for the help.

Hello Fernando, The ultimate design should most definately meet the business need and of course there are going to be trade-offs. In the scenario you described, you are correct...there would be additional managment costs associated by having to duplicate the exlcusions and reservations on both servers. What I have done in the past is manage that via dotNET code where all reservations are created usign a front end application that writes back to the DHCP scopes. I am not against the clustering of DHCP...its a solid solution and if it meets your business needs at an acceptable risk level and cost, then its should be implemented. If you want to test the fail-over, schedule a change window and during that scheduled time, pull the networking cable from one of the nodes and see if the resources fail-over. The impact is going to be very minimal as you will not be able to service DHCp clients during that time. If you have a reasonable lease period, no hosts should really be affected. You may want to pick up a good book on clustering. I think you would benefit from it. Visit: anITKB.com, an IT Knowledge Base.