I'm trying to set up NLB on a couple of Windows 2008 R2 virtual servers running on top of Hyper V R2.
The servers each have a single vNIC for LAN access (and a second vNIC for SAN access). I'm setting up the cluster to use Multicast mode. The vNICs are each set to allow MAC spoofing.
Essentially I'm finding that i can add SERVER1 as a host and it will pick up and respond to the cluster IP from a remote subnet. If I then 'stop' the node in NLB manager it still responds when I would expect it to stop answering on that IP.
If I recreate the cluster and add SERVER2 as the first host, the wizard completes correctly and an IPCONFIG on the server shows that it now has the cluster IP but I can't ping the cluster IP from a remote subnet but I can from another machine on the same subnet.
As a final test – with both servers in the cluster, pinging from another machine on the same subnet I still get a response from the cluster IP when both nodes are stopped according to the NLB manager.
The two VMs are sat on the same physical blade and are built up exactly the same as they'll be used as SharePoint web front end servers. I'm at a loss as to what could be wrong with the second VM that prevents it taking on the address just as the sole node in the cluster, never mind the strange behaviour of the cluster when I stop/start nodes.
Best Answer
Typically cross-subnet issues like this are networking equipment related, not necessarily NLB. Take a look at this blog post from the Microsoft Enterprise Networking team, it goes into the problems and possible solutions.