In normal usage, the barrier is very simple. All the threads wait until the number of required parties arrive. Upon arrival of the last thread, the action is executed, the threads are released, and the barrier can be reused. However, exception conditions can occur and cause the barrier to fail. When the barrier fails, the CyclicBarrier class breaks the barrier and releases all of the threads waiting on the await( ) method with a BrokenBarrierException. The barrier can be broken for a number of reasons. The waiting threads can be interrupted, a thread may break through the barrier due to a timeout condition, or an exception could be thrown by the barrier action.
In every exception condition, the barrier simply breaks, thus requiring that the individual threads resolve the matter. Furthermore, the barrier can no longer be reused until it is reinitialized. That is, part of the complex (and application-specific) algorithm to resolve the situation includes the need to reinitialize the barrier. To reinitialize the barrier, you use the reset() method. However, if there are threads already waiting on the barrier, the barrier will not initialize; in fact, it will break. Reinitialization of the barrier is complex enough that it may be safer to create a new barrier.
Finally, theCyclicBarrier class provides a few operational support methods. These methods provide informational data on the number of threads already waiting on the barrier, or whether the barrier is already broken.
