More on routine loss of external network connectivity on Hyper-V hosts (not guests)

Tristan Watkins — Thu, 22 Oct 2009 23:13:19 +0000

Further to my post from a few months ago on the this topic (dating back to the RC build), I’ve seen this same problem a few more times on the RTM build of Windows Server 2008 R2. My suggested approach still fixes the problem and it doesn’t recur, but I’ve still not been able to pin down a cause and I can find no documentation on host machine MAC assignments anywhere. Search engines results are entirely focused on MAC pool duplication, which should be a completely distinct issue since the MAC address in question falls outside of Hyper-V’s MAC pool.

To help pin this down, I will add that this Hyper-V host machine has been deployed from a WDS image, but I initially discounted that as an issue because the Hyper-V role is not part of the image; it is added during the deployment as specified in an unattended installation file and the Hyper-v networks are created by script post-deployment. Since the Hyper-V host’s MAC address is distinct from the physical adapter’s address, I’ve been assuming that there must be an algorithm that generates the host’s virtual MAC address uniquely, but given that this problem seems to manifest itself soon after installation I’m leaving this possibility on the table.

I’ve noticed a fair amount of traffic for this topic. If you’re having the same problem and this suggestion fixes it, I’d be grateful to know if the Hyper-V system was deployed from an image, as that might help to narrow the troubleshooting effort.

Routine loss of connectivity on a Hyper-V host’s external connection

Tristan Watkins — Tue, 25 Aug 2009 19:00:47 +0000

We’ve recently been piloting a laptop developer build on Windows Server 2008 R2 Release Candidate (Build 7100) with the Hyper-V role. One of the first receipients of this build complained of connectivity problems in Office Communicator once every minute or two. For as-yet undiagnosed reasons we have lengthy sign-in times for Communicator, so this loss of connectivity rendered it completely unusable. This same problem was visible in Outlook, although less disruptive since we use Cached Exchange Mode. Both Exchange and the OCS server are hosted but we also noticed the problem with interrupted file transfers so it clearly wasn’t just an internet connectivity issue. It looked like something to do with the NIC, the cable or a network device.

The network trouble was accompanied by a series of System log event errors from MVSMP:

Port ‘SWITCHPORT-SM-F277C685-E5F8-490D-8CD1-913B854FABD2-0-1′ was prevented from using MAC address ’00-15-C5-7E-EB-39′ because it is pinned to port ‘SWITCHPORT-SM-F277C6′.

The only coverage I could find on similar topics suggested it might be a MAC conflict or a problem with the physical switch at the end of the wire.

The first thing I checked was the host’s external connection’s MAC address and whether it resided in the Hyper-V external network MAC pool (something I’d never inspected before). It turned out it wasn’t. Each laptop with this image had its own randomly-generated MAC pool, as the Hyper-V role was manually added post-deployment. In short, a MAC conflict on a Hyper-V MAC pool range seemed unlikely, especially since the host’s adapter wan’t even on that range.

Having exhausted the quick troubleshooting and investigative options, I set about fixing it, leaving the root cause for another day. These were the steps I took to resolve the issue in our environment, which assumes use of internet connection sharing (ICS) from a Hyper-V external network connection to a Hyper-V internal network connection:

Disable the ICS connection in any running guest VMs
Disable ICS in the host
Delete the external network in Hyper-V Manager
Uninstall the host’s physical NIC in Device Manager
Scan for new devices
Add a new External Network using the newly reinstalled device, called Hyper-V External Network in Virtual Network Manager
Rename the Local Area Connection in the host to Virtual Network Switch
Rename the new Local Area Connection for the host’s external network connection to Host External Network Connection
Share ICS from the Host External Network Connection to the Host ICS Network Connection
Re-enable the ICS adapter in the running guest VMs
Confirm that the System event log errors are resolved and that the connection appears to be stable

This has only happened once in a number of deployments of this build, so I suspect it’s a fluke. If it arises again it will be worth tracking MAC address allocation throughout this process, as the host external connection will change from the old conflicting MAC to the physical MAC when the adapter is reinstalled to a new MAC when the NIC is given to Hyper-V again. One assumes that if a new MAC is not generated the problem will persist, based on that error message.

Update (15/09/09): this has now happened twice more but I haven’t been able to pin down a cause still. In each case the fix has been straight-forward, as above and the problem has not returned on those machines.

Update (22/10/09): Further updates.

Tristan Watkins on IT Infrastructure » connectivity Archive

More on routine loss of external network connectivity on Hyper-V hosts (not guests)

Routine loss of connectivity on a Hyper-V host’s external connection