The action when a node in a cluster fails, and another node takes over. This usually involves a slight service disruption, as the presently active connections to the failed node are severed, and the caches on the node that is being failed over to are cold, and have to go through cache warming.

See also High Availability.