A time-based adaptive checkpointing is an efficient coordinated strategy because a process needn?t send extra message to directly coordinate with others. However, when a failure occurs in some special period, the time-based adaptive checkpoint schemes may be not consistent. In this paper, the issues of time-based adaptive checkpoint strategy that will result in system inconsistency are first discussed and then a new two-phase time-based strategy is proposed. The performance of the proposed strategy is analyzed. The two-phase strategy is consistent and it has better performance than other time-based algorithm because it doesn?t need blocking the processes and doesn?t need logging all messages.