I have 2 issues with Hyper-V clustering. I have a 3 node Hyper-V Cluster
setup with HP Blade and NetApp hardware. I have 17 Hyper-V Guests running
2008 and 2003 Server Ent and Std editions.
Everything is working fine and has been for about 3 months, but recently I
added another guest (Windows 2003 x64 Ent) and for some reason when it fails
over to the other node in Hyper-V Manager I see it as running on the one node
I moved it to, but on the other node that is was running on instead of it
being removed from Hyper-V Manager it stays there with a Saved state. If I
try to delete the old one that is in Saved State it just says an error
All my other 16 guests when they are moved between nodes are removed from
Hyper-V Manager on the node it is no longer running on as I would expect.
I tried deleteing the XML files under Program
data\Microsoft\Windows\Hyper-V\Virtual Machines and then rebooted the node
which removed it from Hyper-V Manager, but then once I failed the guest over
and back again the problem appears again.
Why is it doing it with the one guest on the cluster?
My next step I am thinking of taking is to remove the guest from the cluster
and delete the XML Files in Program data\Microsoft\Windows\Hyper-V\Virtual
Machines on the nodes and then re-add the guest back to the cluster.
Also a note to anyone reading this BE CAREFUL USING SNAPSHOTS IN A CLUSTER
environment. I created a snapshot and deleted the snapshot later after the
update I was doing worked then shutdown the Server thinking it would merge
the snapshot data. Wrong!!! If you create a snapshot in a cluster and then
delete it, it will never get a chance to merge so in 2 - 3 weeks in my case
all of a sudden my VM stops running and the drive used for snapshots is full
even though I deleted the snapshot and shutdown which I would have thought
should let the merge occur. But since the cluster fires the VM back up
instantly the merge actually never occurs. You have to go into the cluster
properties and tell it not to fail over the VM and then shutdown the VM let
the merge occur and then change the cluster properties back to fail over the