This paper examines how exposure to disruptive students affects teacher retention using linked teacher-student administrative records from North Carolina. To address non-random classroom assignment, we instrument for classroom exposure using the school-by-grade share of disruptive students based on prior-year disciplinary infractions. A one standard deviation increase in the share of disruptive students raises the probability of a teacher leaving the school in the following year by 1.6 percentage points. We do not find differential effects by teacher characteristics. However, working in a school environment with supportive leadership and greater teacher input into decision-making mitigates the impact of student disruptions.