NLB Cluser of two Hyper-V Virtual Servers fails under load

I have two Physical Servers, each running Server 2008 R2 Enterprise SP1 and each is running the Hyper-V role.
they both have 4 NICS, two of the Nics are reserved for VM's and they each have several Virtual Machines hosted on them. A NLB cluster has been created with 1 virtual machine from each of the two physical servers. The question is involving the stability of the cluster under load.
Both VM's are Server 2008 Std SP2 and configured identically:
2 network adapters (not legacy)
1 has static ip with dynamic mac, the other is used for NLB and has static mac and mac spoofing enabled.
each has two logical processors and 4 Gigs of RAM and integration services components have been updated.

The two VM's run a classic ASP IIS application that when tested separately work fine and when tested as a cluster work totally fine.  For testing the cluster I have an internal dns host record "" pointing to the cluster ip address. Within an hour of changing DNS to point the production record "" to the cluster ip address and traffic starts to hit it, the cluster stops responding.
The cluster is configured as unicast with ports 80 and 443, and the cluster is reconfigured with the production host name and converged successfully before changing the DNS record.  Nothing of consequence shows in the event logs.
The only thing I can think of is switch flooding but would like to get this resolved and stable.
Who is Participating?
jbvernejConnect With a Mentor Commented:
What kind of NLB are using in your settings ?
"Unicast" NLB or "MultiCast" NLB or even "MultiCast IGMP"  ?

In my point of view: y
ou don't need to use "STATIC MAC"  in the VM settings on NLB NIC, you only need to have "mac address Spoofing" enabled and you should stay in Dynamic Mac Address.

As far I know, NLB layer must can change MAC address onto the NIC to override native NIC MACaddress .It replace it with it own dedicated macaddress. Without it, it can't work correctly.
Wwith your configuration, I am not sure that NLB finally change the macaddress (and if it can do it in the VM, the hyperVisor override it again with another macaddress)

Could you try without Static MacAddress on NLB ?
seaytechAuthor Commented:
Any ideas experts?  I'm surprised no one has commented yet.

what do you mean exactly by " the cluster stops responding." ?
- one VM hangs, the 2 VMs hang (need to be "hard reseted")?
- one or 2 host HyperV hangs ?
- The nlb Virtual IP is pingable ?
- the nlb vm are pingable ?

In the NLB, if you stop one member of the cluster, does it work on the only one node ?
seaytechAuthor Commented:
Until load is put on the cluster, I can test all day long, bringing one server offline and back on, then the other and back on and all works perfect, the cluster works as expected. Under load,
IIS on the NLB hosts even stops answering requests, both  of them.  During the outage I tried accessing the hosts by the name of the machine instead of the NLB name (this usually works also, host headers are not being used) and iis would not respond.  It's been so long now I can't remember if I ping tested the VIP or the individual IP's during the outage, but I can remote to the servers so I know at least the machine ip is active.  By Reset, I mean i had to stop and restart the NLB control hosts using the NLB manager. After doing so, It would start working again, but not for long, maybe 15 minutes.
seaytechAuthor Commented:
So sorry this has been left dormant for so long.  We found the issue was actually with the web application and not with the NLB at all.  thank you for all your assistance.  I will give points to jbvernej for the logical mac breakdown that made me think a little deeper into using hyperv VM's for IIS apps.
thank you.
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.