Dam is a dynamical system receiving stochastic inflows while releasing controlled outflows. Its optimal control requires considering a state constraint on the stored water volume and should use an objective function considering both human activities and downstream environmental management. In addition, the system dynamics are not always known exactly. We briefly explain a robust modeling framework of dam discharges based on a dynamic programming approach and constrained viscosity solutions with a demonstrative computational example. We focus on a single problem rather than reviewing diverse examples because it has key features that stochastic control in real applications encounters.