VMware Server 1.x and Microsoft NLB/WLBS

While working on a NLB/WLBS cluster for Windows Server 2003 Terminal Services [MS SPTK 70-290 ch 2 lesson 5] I encountered a problem that had me stumped. I use VMware Server 1.0.6 to create my dev machines. I had configured NLB/WLBS clusters before with no difficulty, but could not get the same thing working with my virtual machines.

Each of the two cluster members had two vNICs installed and configured for different virtual switches on different subnets. One vNIC was for management and the other was for the NLB/WLBS cluster. I could configure the first node without problem but was never able to configure the second node: the second node never converged and eventually lost connectivity. On a fluke I discovered that after a reboot of my host and all virtual machines I noticed that the first node was no longer pingable but now the second node was pingable.

Some of the troubleshooing:

• Confirmed that the MAC address of the cluster vNIC was the same on both nodes after the cluster was configured using IPCONFIG on the local machine.
• Used the NLB Troubleshooting Overview for Windows® Server 2003 [http://www.microsoft.com/downloads/details.aspx?FamilyId=9E6F4999-AEEB-40CF-867B-AF68742FFFDC&displaylang=en].
• Confirmed compatibility of vNICs using CHKNIC [http://support.microsoft.com/kb/816910]. The VMware Server vNICs tested positive for compatibility with NLB.
• Verified configuration of Terminal Services [http://support.microsoft.com/kb/280805].
• See http://support.microsoft.com/kb/287672.
• Change default gateway [http://support.microsoft.com/kb/193602].

Google and VMTN finally provided the solution and an explanation: NLB in Unicast mode support? [http://communities.vmware.com/thread/42501]. The answer was to use Multicast, not Unicast for the NLB/WLSB cluster. After making the configuration change to my cluster it work right away.

MS SPTK 70-290= Microsoft Self-Paced Training Kit 70-290, ISBN 0-7356-2289-2