The current paper shows an end-to-end approach how to process XML files in the Hadoop ecosystem. The work demonstrates a way how to handle problems faced during the analysis of a large amounts of XML files. The paper presents a completed Extract, Load and Transform (ELT) cycle, which is based on the open source software stack Apache Hadoop, which became a standard for processing of a huge amounts of data. This work shows that applying open source solutions to a particular set of problems could not be enough. In fact, most of big data processing open source tools were implemented only to address a limited number of the use cases. This work explains and shows, why exactly specific use cases may require significant extension with a self-developed multiple software components. The use case described in the paper deals with huge amounts of semi-structured XML files, which supposed to be persisted and processed daily.