Standardized and sufficient log is a necessary part of good code quality, and it plays an important role in failure diagnosis as well. Code quality management, however, is restricted by the high complexity of large-scale software. Currently, it's difficult and inefficient to reproduce and diagnose system failure with logs. This paper surveys log-related work from three aspects including log characterization, failure diagnosis with log and log enhancement. Through detailed study on several widely-used open-source software, the paper reveals some log-related observations, along with the problems which have not been well handled by existing tools. Finally, it proposes several possible log-related work, and analyzes potential challenges.