After installing Failover Cluster Manager on a Server 2012R2, when a VM is started using a virtual switch connection to a Cisco 3750, the physical switch gets overwhelmed with high CPU usage and won't stop until the switch is power cycled. When I did the same thing in Hyper-V before I enabled the FCM feature everything worked okay. But high availability is the goal. What is going on with the interaction of FCM and the physical infrastructure.

Hi RonArmbruster,

Does the specific vm or any vm start will effect the switch CPU load? Are you using NIC teaming? Can you clarify about FCM feature, please try to recreate the external  virtual switch then monitor the issue again, the quick method is you can install the network monitor on your host then capture the vm data when it start, it will clearly indicate what data effect the switch.

Microsoft Network Monitor 3.4 (archive)

https://www.microsoft.com/en-us/download/details.aspx?id=4865

Setting up Port Mirroring to Capture Mirrored Traffic on a Hyper-V Virtual Machine

http://blogs.technet.com/b/networking/archive/2015/01/06/setting-up-port-mirroring-to-capture-mirrored-traffic-on-a-hyper-v-virtual-machine.aspx

Im glad to be of help to you!

Hi Ron,

Capture a network data is a good advice.

Do you have any of MAC addresses used by your new cluster reserved in DHCP? A few years ago I have seen an interesting behavior:

Server has let's say MAC Address "A". In a DHCP server for this MAC address a reservation were created. But the reserved address was statically assigned to another server. So nothing tried to use this reservation and it were 'sleeping'. When we start a new server (with the MAC address "A") that is what happened:

- Server asked DHCP for an address

- DHCP offered the reserved IP Address

- Client agreed and asked for a lease

- DHCP checked if this address is already in use (conflict detection), found a conflict and declined the lease request.

Here starts an endless cycle. As the result: gigabytes of dhcp traffic were generated in a less than 10 minutes, very high switch load and switch DoS...

  

Thx for your interest in getting this resolved.

We still don't know what caused the unstable switching loops but we feel there is an errant incompatibility within this network. There was a misuse of VLANs and many services were on the default VLAN along with the new services we were installing that could have been in conflict. We moved the server nodes running the Hyper-V and Failover Cluster Mgr to their own server VLAN to isolate any traffic that was misread by some other device.

The Hyper-V and high availability work as designed.