使用ThreadPoolExecutor时，用了两个pool，一大一小；

思路是这样的，任务过来优先使用小池来处理，小池处理不了丢进大池里处理；大池的核心线程数设置为可以回收，以减少资源的占用。

package org.test.thread;import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import java.util.concurrent.ExecutorService;
import java.util.concurrent.LinkedBlockingDeque;
import java.util.concurrent.RejectedExecutionHandler;
import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;public class MyThreadPool {/*** 线程池(大池)*/private static final ThreadPoolExecutor bigPool = new ThreadPoolExecutor(50, 100, 2, TimeUnit.MINUTES,new LinkedBlockingDeque<>(10000),new ThreadPoolExecutor.CallerRunsPolicy());/*** 线程池(小池)*/private static final ExecutorService smallPool = new ThreadPoolExecutor(25, 50, 5, TimeUnit.MINUTES,new LinkedBlockingDeque<>(100),new RunsInBigPoolPolicy());private static Logger logger = LoggerFactory.getLogger(MyThreadPool.class);static {//大池的核心线程数可以被回收bigPool.allowCoreThreadTimeOut(true);}/*** 异步执行** @param runnable*/public static void execute(Runnable runnable) {if (runnable == null) {return;}if (smallPool != null && !smallPool.isShutdown()) {smallPool.execute(runnable);} else if (bigPool != null && !bigPool.isShutdown()) {bigPool.execute(runnable);} else {runnable.run(); //主线程直接调用logger.error("线程池异常，主线程直接执行了任务...");}}/*** <br>任务运行在大池里的策略</br>* 由于CallerRunsPolicy会造成主线程阻塞，所以把小池处理不了的任务丢到大池里执行;*/public static class RunsInBigPoolPolicy implements RejectedExecutionHandler {public RunsInBigPoolPolicy() {}@Overridepublic void rejectedExecution(Runnable runnable, ThreadPoolExecutor executor) {if (runnable == null) {return;}if (bigPool != null && !bigPool.isShutdown()) {bigPool.execute(runnable);logger.warn("a task is switched into big pool...");} else {runnable.run(); //主线程直接调用logger.error("线程池异常，主线程直接执行了任务...");}}}
}