May 17, 2007, 11:19 AM —
Send your Windows question to Mitch today! | See other Windows tips
DHCP servers are at the heart of network client availability, and I've seen administrators bend over backwards trying to build some sort of high-availability solution that can kick in when their main DHCP server fails. And believe me, seeing your average administrator try to bend over backwards isn't a pretty sight.
One company I heard about recently apparently does the following: They have two DHCP servers configured with identical scopes, but one of the servers (we'll call it the fallback) has its scopes deactivated. The administrator created a netsh script that is scheduled to run periodically on the main server and which first backs up this server's DHCP configuration and then restores the configuration onto the fallback server.
So is this a good idea? Probably not. For one thing, if the main server suddenly fails the secondary server probably won't respond to DHCP clients trying to renew their address leases. Why not? Because DHCP lease renewal messages initially use unicast and are thus sent directly to the IP address of the DHCP server that leased the client its address. (When a lease reaches 87.5 percent of its lifetime however, the client will try broadcasting to any available DHCP server in order to try and get its lease renewed.) What this means is that the fallback DHCP server will only work if the admin changes its IP address to match the failed server, but this requires manual intervention of course and that means this scenario won't work automatically. And on top of that, the lease info in the DHCP database of the fallback server will only be as accurate as the last backup of the main server, and this could cause problems with lease renewals.
A much better solution to DHCP server availability is to use clustering. Set up a simple two-node active/passive cluster for your DHCP server, and when the active node goes down for some reason, the passive one takes over almost immediately and has up-to-date lease information for all clients on your network. Or you can combine a clustered approach with 80/20 scope splitting to ensure both high availability and fault tolerance. For more information on implementing DHCP clusters using Windows Server 2003, search Microsoft TechNet for "DHCP cluster" and you'll get several useful articles.
Breaking news: In addition to having been lead author for the Windows Vista Resource Kit (a must-have for administrators who will be deploying Vista in their organizations) I also recently finished writing Microsoft's first book about their upcoming Windows Server Code-Name "Longhorn" platform. This book will be published soon by Microsoft Press and you can pre-order it now. The reason I'm plugging my new book here (in addition to hoping sales will fund my retirement home in Florida - lol) is because clustering is going to be a whole lot easier to implement in Windows Server Code-Name "Longhorn" than it is today in Windows Server 2003. So if you've been holding off clustering your DHCP server because you're afraid you'll have to wade through a 50-page whitepaper just to get the clustering part working, wait till my book comes out and you'll see how simple it will be to set up clustered DHCP (or clustered anything else) in Windows Server Code-Name "Longhorn". Plus you'll learn about Windows Server Core, Terminal Services Gateway, Network Access Protection, and a ton of other cool things by reading my book, so pre-order it today!
I'm shameless, aren't I? Well, authors gotta eat.
