Hyper-V problem with Network Load Balancing (NLB) in virtual machines running Windows 2008

I recently had a problem with virtualized Exchange 2007 SP1 with CAS role installed and NLB enabled. Exchange is installed on top of Windows Server 2008 and two network cards were added to configuration. One network card for Outlook Anywhere on port 443 in NLB and the other card for other traffic and server management.

After some time one of the virtualized Exchange 2007 servers started to reboot unexpectedly without any useful information in event logs.

Since I had many problems with NLB in my career I thought that maybe Hyper-V and NLB are not working well together.

During the search for error I ran into Rob Whitehouse’s blog post in which he posted the link to Microsoft’s knowledge base article with exactly the same error/problem I got.

Microsoft provided a hotfix for this error that was described with following symptoms:

• You are running a Windows Server 2008 virtual machine in the Hyper-V environment on a Windows Server 2008 host computer.
• You try to create a new Network Load Balancing (NLB) cluster on this virtual machine. Or, you try to add this virtual machine to an existing NLB cluster.

In this scenario, the NLB host on the virtual machine does not converge as expected. Additionally, if a large volume of traffic occurs in the NLB cluster, the guest system on the virtual machine may restart unexpectedly, and you may receive the following Stop error message:

 

STOP: 0x000000E1 (parameter1, parameter2, parameter3, parameter4)

In my case hosts were converged properly (after the reboot) just that one of the servers was constantly unexpectedly rebooting.

Reference:

Windows Server 2008 Hyper-V virtual machines generate a Stop error when NLB is configured or when the NLB cluster does not converge as expected