解决ForkJoinPool的Zipkin跟踪问题

zipkin对于线程池已经提供了很完善的方式处理，例如：

CurrentTraceContext currentTraceContext = MDCCurrentTraceContext.create();
currentTraceContext.wrap(()->{System.out.println("hi");});
ExecutorService executorService =  currentTraceContext.executorService(Executors.newFixedThreadPool(1));
executorService.execute(()-> System.out.println("hi"));

但上述实现仅对于Thread与ExecutorService有效，而对java 8的并行流则无能为力。例如：

String[] hi = new String[]{"hi"};
Arrays.stream(hi).parallel().forEach(System.out::println);

这里主要是由于java 8的并行流使用的是ForkJoinPool，系统启动时，会默认初始化一个commonPool。因而加入zipkin的关键也在于如何修改commonPool的默认属性。

为此，感谢文章给指明了前进的方向 。解决方案如下：

static {
	System.setProperty("java.util.concurrent.ForkJoinPool.common.threadFactory",
			ZipkinForkJoinWorkerThreadFactory.class.getName());
}

public class ZipkinForkJoinWorkerThread extends ForkJoinWorkerThread {
	private final TraceContext invocationContext = ZipkinContext.getCurrentTraceContext().get();

	public ZipkinForkJoinWorkerThread(ForkJoinPool pool) {
		super(pool);
	}

	@Override
	public void run() {
		try (CurrentTraceContext.Scope scope = ZipkinContext.getCurrentTraceContext().maybeScope(invocationContext)) {
			super.run();
		}
	}
}

public class ZipkinForkJoinWorkerThreadFactory implements ForkJoinPool.ForkJoinWorkerThreadFactory {
	@Override
	public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
		return new ZipkinForkJoinWorkerThread(pool);
	}
}