A job shop completely fills an order demand. Production is conducted in "lot". Lot-sizes that are too large may cause unnecessarily high production cost. On the other hand, too small lot-sizes may result in high costs due to frequent setup. The objective is to determine the first run size to minimize the expected total cost. This problem arises in unreliable production systems and is formulated as a stochastic dynamic program. Although algorithms have been provided in the literature for binomial and discrete uniform yields, the interrupted geometric yield did not have an excellent procedure to solve for the optimal lot-size. In this research, we consider the multiple lot sizing problem with interrupted geometric yield distribution, the customer demand is rigid, and products cost is linear. An effective algorithm is developed to search an optimal lot-size for this problem.