we are planning to run two-node clusters (for database and file-services) with cluster nodes in different sites with a fast, but sometimes unreliable network connection. Data should be replicated with DRBD between the cluster nodes/sites (no shared storage).
I’m specially afraid of split brain here. In this scenario fencing depends on the (unreliable) network connection between the two sites – in my opinion fencing is unreliable and not possible in this scenario.
What would be the best way to run such a cluster?