Re: cluster re-config when removing network interface
If the interface is used for sending/receiving cluster heartbeats, you would have to at least remove this node from the cluster and re-add it as a "new" node.
The heartbeat configuration cannot be changed while the node is running... but a new node can have a different heartbeat configuration than the old ones.
Besides, why remove it?
If the fault is detected and does not cause any further problems, it can wait until you take the node down for fixing the failed interface. (If you have hardware that allows you to hot-swap the network interface, good for you.)
If the broken interface *does* cause problems, well... one of the reasons your cluster was built was certainly to make it possible to fix a broken node without affecting the applications, right? If this does not work in practice, find out why and fix that problem too. Usually this means the software bloat has eaten away your power reserve, and the remaining nodes cannot handle any extra load. The solution is to add more power, either by adding CPUs or RAM to the existing nodes, or increasing the number of nodes.
The alternative to fixing it now *at a time of your choosing* is to have a simple node failure at some unknown future time turn into a total catastrophe because the cluster cannot perform satisfactorily with one node down.
If you don't intend to replace the broken interface, you must realize you're weakening the redundancy built into your cluster configuration. The loss of the remaining interface card(s) on that node will now be more likely to drop the entire node out of the cluster.
