A method for quickly detecting a fault includes: detecting, by a Kernel Black Box KBox set, a fault occurred in an operation system; and generating, by the KBox set, fault information based on the detected fault; and transmitting, by the KBox set, system fault notification information including the fault information to an application high availability HA subsystem via a management unit of an infrastructure layer, to trigger a service fault processing of the application HA subsystem. Thus, the fault or unhealthiness of an OS is detected rapidly and a service application layer is timely notified to process the fault, thus reducing service loss.