4. System Clock Synchronization

Time plays a fundamental role in AccuRev’s architecture and in its day-to-day operations. Some examples: each transaction is recorded in the database at a particular time; a snapshot reconstructs the state of a stream at an arbitrary time; the stat command and the AccuRev GUI use timestamps to optimize the lookup of modified files within a workspace.

AccuRev is a networked product: programs execute on one or more server machines and multiple client machines. In a perfect world, the system clocks on all these machines would always be perfectly synchronized. The data items on the (master) server machine (say, versions created by keep commands) and corresponding data items on a client machine (the files that were kept) would have timestamps that are consistent with each other.

Software systems do exist that synchronize all the machines in a network to within milliseconds. If your organization uses such a system, then you don’t need to read any further in this chapter!

Most software development organizations don’t have — and don’t need — synchronization at the millisecond level. AccuRev defines a 5-second tolerance as “good enough for software configuration management”. This chapter describes AccuRev’s own facilities for detecting system-clock discrepancies, along with other related facilities commonly available on Windows and UNIX/Linux systems.