Sustainable edge computing demands intelligent scheduling of containerized workloads to exploit intermittently available renewable energy at geographically distributed sites. This work introduces a Contextual Multi-Armed Bandit (CMAB) framework for green-aware container orchestration, leveraging real-time context, such as energy availability, and resource utilization, via linear CMAB algorithms. A Python-based simulator models realistic solar and wind dynamics across regions. Compared to optimal, random, and naïve baselines, our CMAB scheduler improves green-energy utilization by up to 30% and cuts brown-energy use by 20%, while maintaining application performance guarantees. These findings underscore the potential of learning-based methods for advancing energy-efficient and sustainable edge infrastructures.