Q. What is the problem with Windows Server 2008 cluster network names?

A. Server 2008 lets a network name resource using OR logic have multiple IP addresses. Therefore, the network name is available as long as one of the IP addresses is available, which is vital if you have cluster nodes on different IP subnets.

The concern is the amount of time it might take cluster clients to obtain a network name IP address. The problem occurs when cluster clients access the resource through the network name, which DNS resolves. DNS is cached on the client for as long as 20 minutes, which means it could take up to 20 minutes for cluster clients to get the correct IP address to connect to a network name. To resolve this problem, reduce the network name cluster service names Time-to-Live (TTL) to five minutes.

Use the cluster.exe command-line utility to make the TTL change, as in the following example:

cluster res /priv HostRecordTTL=300

After making the change, the client will wait only five minutes before contacting DNS for an updated IP address.

But that’s not the end of the story. With the TTL change, five minutes is the maximum time a client will wait before asking DNS for a new IP address—but only if DNS is up to date.

One possible solution is using Active Directory (AD)-integrated DNS, in which AD stores DNS content and replicates the content according to the AD schedule. One problem with this approach, however, is that if you have multiple AD sites with domain controllers (DCs) at each site, replication could take 15 minutes or longer. When the remote-site client queries DNS, the DC might not have received the updated network-name record. There’s no simple solution. It's a consideration when planning clusters that span different subnets using multiple IP addresses.