This thesis presents a discrete time model of a multi-stage production system with three major types of uncertainty : the yield at each production stage, demand for finished product, and vendor unstability. Our goal is to analyze the impact of above three uncertainties on the production level and inventory level at each production stage in order to provide insights for production management. We propose three linear production control models in this thesis. The first model is proposed to analyze the difference of used information level when he decides the production level at each production stage. The second model is a backward linear production control model. This conceptual model takes into account two uncertainties : the vendor unstability and the yield rate at each process. The last model is an integrated model that integrates the forward linear control model (this is similar to the pull type production control) and the backward linear production control model. In this model, we can analyze the impact of above three major types of uncertainty on the production system. Above three new models lead the production system to steady state, in which the means and the variances of the production and the inventory levels can be expressed in closed forms.