Guest Cluster error in Hyper-V Cluster

Question

Hello everybody,
in my environment I do have an issue with failover clusters (Exchange, Fileserver) while performing a live migration of one virtual clusternode. The clustergroup is going offline.
The environment is the following:
2x Hyper-V Clusters: Hyper-V-Cluster1 and Hyper-V-Cluster2 (Windows Server 2012 R2) with 5 Nodes per Cluster
1x Scaleout Fileserver (Windows Server 2012 R2) with 2 Nodes
1x Exchange Cluster (Windows Server 2012 R2) with EX01 VM running on Hyper-V-Cluster1 and EX02 VM running on Hyper-V-Cluster2
1x Fileserver Failover Cluster (Windows Server 2012 R2) with FS01 VM running on Hyper-V-Cluster1 and FS02 VM running on Hyper-V-Cluster2
The physical networks on the Hyper-V Nodes are redundant with 2x 10Gb/s uplinks to 2x physical switches for VMs in a LBFO Team:
New-NetLbfoTeam
-Name 10Gbit_TEAM -TeamMembers 10Gbit_01,10Gbit_02
-TeamingMode SwitchIndependent -LoadBalancingAlgorithm HyperVPort
The SMB 3 traffic runs on 2x 10Gb/s NIC without NIC-Teaming (SMB-Multichannel).
SMB is used for livemigrations.
The VMs for clustering were installed according to the technet guideline:
http://technet.microsoft.com/en-us/library/dn265980.aspx
Because my Hyper-V Uplinks are allready redundant, I am using one NIC inside the VM.
As I understand, there is no advantage of using two NICs inside the VM as long they are connected to the same vSwitch.
Now, when I want to perform a hardware maintenance, I have to livemigrate the EX01 VM from Hyper-V-Cluster1-Node-1 to Hyper-V-Cluster1-Node-2.
EX02 VM still runs untouched on Hyper-V-Cluster2-Node-1.
At the end of the livemigration I see error 1135 (source: FailoverClustering) on EX01 VM, which says that EX02 VM was removed from Failover Cluster and I have to check my network.
The clustergroup of exchange is offline after that event and I have to bring it online again manually.
Any ideas what can cause this behavior?
Thanks.
Greetings,
torsten

Answer 1

Hello again,
I found the cause and the solution :-)
In the article here: http://technet.microsoft.com/en-us/library/dn440540.aspx
is the description of my cluster failure:
########## relevant part from article #######################
Protect against short-term network interruptions
Failover cluster nodes use the network to send heartbeat packets to other nodes of the cluster. If a node does not receive a response from another node for a specified period of time, the cluster removes the node from cluster membership. By default, a guest
cluster node is considered down if it does not respond within 5 seconds. Other nodes that are members of the cluster will take over any clustered roles that were running on the removed node.
Typically, during the live migration of a virtual machine there is a fast final transition when the virtual machine is stopped on the source node and is running on the destination node. However, if something causes the final transition to take longer than
the configured heartbeat threshold settings, the guest cluster considers the node to be down even though the live migration eventually succeeds. If the live migration final transition is completed within the TCP time-out interval (typically around 20 seconds),
clients that are connected through the network to the virtual machine seamlessly reconnect.
To make the cluster heartbeat time-out more consistent with the TCP time-out interval, you can change the
SameSubnetThreshold and CrossSubnetThreshold cluster properties from the default of 5 seconds to 20 seconds. By default, the cluster sends a heartbeat every 1 second. The threshold specifies how many heartbeats to miss in succession
before the cluster considers the cluster node to be down.
After changing both parameters in failover cluster as described the error is gone.
Greetings,
torsten