NLB can be somewhat tricky to configure. For certain, the best method would be using NLB in IGMP Multicast mode. But "IGMP Multicast" is a switch-assisted method, which would require you to configure the switch. And that is total different story.
You are currently using NLB in Unicast mode, which is a non-switch-assisted method, which requires no configuration on a switch. Sounds easy. But there are a few problems with it, like these:
- NLB in Unicast mode causes a well know issue called 'switch-flooding'.
- Some low-budget L3 switches do not support NLB in unicast mode.
- NLB in unicast mode is not supported on VMware. Multicast is required, unless you configure a work-around.
- Because the NLB nodes use the same unicast MAC Address (which causes switch-flooding); in some situation NLB members 'cannot' communicate. This is why a second NIC is required for communication. This is why you get a warning when you open NLB manager.
Anyway. I don't know you scenario, but allow me to ask a few questions:
- You added this to the High Availability (Cluster) forum. Are you using NLB on top of Failover Cluster Manager? You don't, right?
- Are you running the NLB members as a Virtual Machine? If so, make sure the vNIC is allowed to spoof its MAC Address.
- If you are using VMs; On which Hypervisor are you hosting them? Before you configure, please move the VMs to the same host before you configure it, so they are connected to the same vSwitch.
Sometimes if NLB in Unicast doesn't work well. Try to add the NLB member from the other side as well. Of course it ain't nice, but it can work.