Machine Learning Operations (MLOps) streamline the lifecycle of machine learning (ML) models in production. In recent years, the topic has attracted the interest of practitioners, and consequently, a considerable number of tools and gray literature on architecting MLOps environments have emerged. However, this has created a new problem for organizations: selecting the most appropriate tools and design options to implement their MLOps environments. To alleviate this problem, this paper proposes a reference architecture and 32 requirements for MLOps by systematically reviewing 59 articles in the industrial gray literature. Furthermore, we used a survey and conducted semi-structured interviews with six MLOps experts to validate, refine, and extend our findings. This reference architecture, derived from the current state of practice, will enable organizations to make informed design and technology choices when embarking on their MLOps journey, while providing a technology-independent baseline for further MLOps research.