Reinforcement Learning for Continual Control in Computer Systems