During the testing of Hyper-V live migration on a freshly built failover cluster (2008 R2), all seemed to work well apart from one Linux (CentOS) VM. It had been running fine on a single Hyper-V server with the integration components installed, however when I tried to do a live migration it failed, not giving much clue why.
Moving the VM to the other node and starting it using the Hyper-V manager, generated the following error…
‘SV04′ failed to start.
Microsoft Emulated IDE controller (Instance GUID): Failed to power on with error ‘a device attached to the system is not functioning’
Failed to open attachment: C:\ClusterStorage\Volume1\Hyper-V\Virtual Hard Disks\SV04.vhd. Error – A device attached to the system is not functioning. (0x8007001F)
After some investigation, I found that the VM would only start on whichever cluster node was the owner of the CSV (Custer Shared Volume). I later discovered that disabling the AV (Sophos in this case) on the node that wasn’t the CSV owner resolved the issue – Odd, as I had added an exclusion for scanning on c:\ClusterStorage\*, seems that the exclusion didn’t quite work with it being CSV due to the way it links into the filesystem… Adding an exclusion for *.vhd, then renabling the AV fixed the issue.
