Today's businesses have relied much on system backup and disaster recovery, of which there are more and more products that based on various virtualization platforms, as the virtualization technologies kept developing. Container based virtualization, which is a kind of high efficiency virtualization technology, has great potential to support building more flexible backup and disaster recovery system on its platform. In this paper, we introduce a consistent backup mechanism for disaster recovery that using container based virtualization. First, we proposed the concept of consistent checkpoint which contains both memory and disk image at the backup point in time. Then, for backing up disk image, we use incremental backup method and a two-step aggressive backup process, in which production system and data backup can run at the same time, to deal with it. At last, we combine our disk backup method and the memory checkpoint function of virtualization platform together to accomplish the whole consistent checkpoint's backup. A prototype of our system, in which a non-volatile local buffer is introduced for both speed and reliability purpose, is implemented. And the experimental testing result shows that our system's running overhead to production system is sensible, especially the overhead can drop to only 0.5% when backup is done. Also by using our system, backup frequency of production system has the potential to reach over fifteen times per minute. All these can prove that our consistent backup mechanism is suitable for disaster recovery purpose in container based virtualization platform.