With the development of information technology application innovation and the integrative process of information system, the government management information system of the China Meteorological Administration is accelerating the implementation of nationalization independent and controllable upgrade iteration. Whether the historical data migration can be smoothly realized becomes a pivotal factor in measuring the success or failure of upgrading. This paper proposes a heterogeneous data migration method based on ElasticSearch, a distributed full-text retrieval framework. Using ElasticSearch’s characteristics of dynamic mapping, distributed scalability and unstructured data processing capability, it can crack the limitations of using traditional methods to migrate massive heterogeneous data. This method has been applied to the national industrialized electronic document system of the China Meteorological Administration and has achieved good practical results. It meets the physical migration needs of historical data to the localization independent and controllable ecology and supports the more flexible secondary use of historical data in the new system.