The log data management system is one of the key infrastructures for cloud computing. Missing of important log data leads to inaccurate and one-sided data analysis and decision making. However, the stronger the log data capturing capability is, the higher the runtime overhead is. In order to capture necessary log data and reduce the runtime overhead as much as possible, this paper first put forward the log data capturing grain level concept was put forward firstly in this paper, and a grain-level self-configuring log data capturing platform was designed then for cloud computing. This platform is consisted of a log data capturing tool, a knowledge base storing grain-level based log capturing rules and facts, a rule-based inference engine for adding and removing specific log data capturing modules, and graphical interfaces for managing the knowledge base and querying log data sets. Finally, our preliminary case study demonstrates the efficiency of our platform.