java - 当计数为 0 时，线程在 CountDownLatch await() 上被阻塞

Question

在分析其中一个生产环境的日志时，我在倒计时等待()中看到一个处于“W​​AITING”状态的线程

...sun.misc.Unsafe.park(Native Method)
...java.util.concurrent.locks.LockSupport.park(Unknown Source)
...java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(Unknown Source)
...java.util.concurrent.locks.AbstractQueuedSynchronizer.doAcquireSharedInterruptibly(Unknown Source)
...java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireSharedInterruptibly(Unknown Source)
...java.util.concurrent.CountDownLatch.await(Unknown Source)

闩锁被初始化为 1 并且另一个线程确实在闩锁的同一实例上调用了 countDown() 方法，但主线程仍然在闩锁上被阻塞。这导致 jvm 无限期挂起。

即使闩锁计数达到零时也被阻止听起来不合理，我正在寻找有关进一步解决此问题的建议。

有任何想法吗？

注 - 使用的 jvm 版本如下

java版本“1.5.0_15”
Java(TM) 2 运行时环境，标准版(构建 1.5.0_15-b04)
Java HotSpot(TM) 客户端 VM(构建 1.5.0_15-b04，混合模式，共享)

更新 - 以下是我在上面谈论的线程的代码片段

private class MyRunnable implements Runnable, Thread.UncaughtExceptionHandler {

        private AtomicBoolean shouldStop = new AtomicBoolean(false);
        private CountDownLatch stopLatch = new CountDownLatch(1);
        private Thread currentThread;

    public void run() {

        Thread.currentThread().setName("My Thread");
            Thread.currentThread().setUncaughtExceptionHandler(this);
            currentThread = Thread.currentThread();
            if (currentThread.isInterrupted()) {
                logger.debug("The pool thread had its interrupted stattus set. Clearing...");
                Thread.interrupted();
                logger.debug("The pool thread had its interrupted stattus set. Clearing...DONE");
            }
            try {
                doBusinessLogic(shouldStop);
            } catch (Exception e) {
                logger.error("An exception was encountered in the thread", e);
            } finally {
                if (currentThread.isInterrupted()) {
                    logger.debug("Clearing interupted status for the thread and returning to pool...");
                    Thread.interrupted();
                }
                stopLatch.countDown();
                logger.debug("Stopped task after counting down on the latch");
            }
        }

        public void stopThread() {
            shouldStop.set(true);
            logger.debug("Stop flag was set to true.. waiting for thread method to return...");
            try {
                stopLatch.await();
                logger.debug("Stop flag was set to true... task has finished. Returning.");
            } catch (InterruptedException e) {
                logger.error("Interrupted while awaiting thread stop event...", e);
            }
        }

        public void uncaughtException(Thread t, Throwable e) {
            logger.error("An uncaught exception occurred in the task thread ", e);
        }


        private void doBusinessLogic(AtomicBoolean shouldStop) {
    long sleepPeriod = 11;
    while (!shouldStop.get()) {
        try {
            Thread.sleep(sleepPeriod);
        } catch (InterruptedException e) {
            logger.debug("Thread was interrupted.Clearing interrupted status and proceeding", e);
            if (Thread.currentThread().isInterrupted())
                Thread.interrupted();
        }
        if (shouldStop.get()) {
            logger.debug("Stop flag was set. Returning.");
            return;
        }
        try {
            logger.debug("Performing business logic...");
            //.....
            logger.debug("Performing business logic...DONE");
        } catch (Throwable e) {
            logger.error("An exception occurred", e);
        }
        if (shouldStop.get()) {
            logger.debug("Stop flag was set. Returning.");
            return;
        }
    }

}


}

这是我在日志中看到的

DEBUG [main Thread] - Stop flag was set to true.. waiting for thread method to return...
DEBUG [My Thread]   - Stop flag was set. Returning.
DEBUG [My Thread]   - Stopped task after counting down on the latch

永远不会打印latch.await() 之后的logger 语句，并且线程转储还表明主线程在latch 上被阻塞。

Answer 1

我本来想把它写成评论，但我的声誉还不允许评论，所以我会在这里提供我的小经验。

我遇到了同样的问题。我的 countDownLatch 是从同步方法访问的:删除同步解决了这个问题(当然我不得不调整方法的代码以应对没有同步的情况)。我不知道如何解释这种行为。