Thanks for replaying.
The setup is 4 nodes. with 4 nics. all heartbeat and production communitating are going thrue one of our 226 switches. and all Isci communication is going thrue a standalone switch that have no external access.
The production network is set on 2 nic that uses Microsoft software team function.
Ok so production network is redundant against the same switch? Heartbeat is a unique nic dedicated to heartbeat. You also have ISCSI network and this run's on a single nic?
If I'm correct I would object to the single ISCSI nic but that's out of scoop here
What happens when you lose communication? Is there a failover in the cluster? The first thing I would verify is that you traffic is running on the switch ship and not on the CPU. If you are hitting the CPU of the switch my bet is that during heavy traffic you may see long latency and even package loss and if this affects the clusterheartbeat (running over the same switch) this may cause a failover in the cluster. The CPU in the switch is weak and there are many examples in this forum where a miss configuration in the switch causes the CPU to handle the traffic causing performance issues.
If this does not apply please explain the issue in more detail and from your username you might even speak Swedish and in this case PM me as it might be easier
