Hi there.
I have a very strange phenomena at one of my customers and can't really understand why this happened.
This is a a Windows 2008 R2 two node cluster (actually patched)
We have a Public network, a cluster network and a live migration network. Storage is Fibrechannel.
Live Migration network is a crossover 10GBit connection, the rest are non-teamed 1GBit switch connection.
All three networks are cluster use enabled. The cluster network was the network with the lowest metric.
Due to some network problems we decided to change the network for the clustertraffic to use the 10GBit connection. So I've lowered the metric for the livemigration network.
Today we tested the config. As we disabled the switchport for the "old" cluster network we've run immediately into a splitbrain situation! All the other networks where still up and running. So the nodes had three routes in total to see each other.
But anyway, we've disabled a non-primary-heartbeat network so nothing should have happened.....
I've digged through the clusterlogs and see the entries that the route from the "old cluster network" vanished and that both nodes decided that the other one is dead!
Cluster validation test is fine!
When I use the network monitor to see where the heartbeat packets are actually running I can see them on the "old cluster network" and not on the LiveMigration network which they should using...........
So according to all settings I've checked Heartbeats should be running through the Livemigration Network and all the other networks should be redundancy. But the heartbeats are still running via the cluster network and redundancy is not working.
Does anybody have an idea?
Regards
Olaf