Embodiments of the invention disclose a failover method and device. In one embodiment, a method executed at a first node in a cluster comprising a plurality of heterogeneous nodes is provided. The method comprises the steps of determining whether an application at a second node in the cluster has a fault or not; and in response to operation of determining that the application has the fault, facilitating data and a service associated with the application to be migrated to a third node in the cluster from the second node, wherein the migration involves at least one node heterogeneous for the second node in the cluster. The invention furthermore discloses a method executed at the third node in the cluster, a corresponding device and a computer program product.