Abstarct: Traditionally, location theory has been concerned with the problem in which the facility weights and their availability are known precisely. However, it is often impossible to have an accurate estimate of all these parameters. Real-life data often involve a significant portion of uncertainty, and these parameters may change with time. Thus, location models involving uncertainty have attracted increasing research efforts in recent years. Several cases of uncertainty have been defined and studied. One case of this problem is the reliability model where servers may sometimes fail, and the clients originally allocated to these servers have to request service from functioning servers. The 2-median problem considered as a reliability-baxsed formulation, in which each server may fail with a given probability. They called this model as the backup 2-median problem and proposed an The backup 2-median problem, asks for a pair of vertices which minimize the expected sum of distances from all vertices to the set of functioning servers.Because of important of this field, we have study many models presented in this area, then we consider the backup Semi-Obnoxious 2-median problem and propose an algorithm for this problem.