In this paper we describe a framework for robust system maintenance that addresses specific challenges of data-centric applications. We show that for data-centric applications, classical simultaneous roundtrip engineering approaches are not sufficient. Instead we propose an architecture that is an integrated model-based approach for software development, database access and data migration. We explain the canonical development process to exploit its features. We explain how the approach fits into the model-driven architecture vision. We report on experiences with the approach in the IMIS environmental mass database project.