In this paper we present the failure management system of IRL (Interoperable Replications Logic), a FT-CORBA platform that provides transparent client/server interactions and server failovers to application clients by using a set of replicated CORBA objects (IRL components). The IRL failure management system includes two basic mechanisms, failure detection and recovery. Each of these mechanisms follows a service stack appoach (i) to reduce as much as possible the impact of ther services in terms of network traffic and (ii) to use the same services both for application objects and for IRL components.Index Terms - High Availability,Fault Tolerance, Object Replication,Fault Tolerant CORBA.