The invention relates to the field of cluster storage and provides a failure detection method suitable to a large-scale storage cluster. According to the failure detection method, a failure detection module is arranged in a storage cluster system, a failure detection fault model used in the method is a failure/recovery model, and a transmission mechanism of failure node information used in the method is a spreading model permanently immune to infectious disease. By use of the method, the failure detection time is enabled to be unrelated to the scale of the storage cluster, the expansibility is excellent, and a stable low load is guaranteed under the condition that the large-scale storage cluster changes constantly.