我需要一些帮助才能更好地理解ThreadPoolExecutor的工作原理。
我有1000个任务需要处理，这个数字在我的程序开始时就知道了。
我设置了一个ThreadPoolExecutor，并希望改变它在运行中使用的线程数，这样如果服务器不是那么活跃（例如在晚上），我们就可以增加它可以使用的线程数。 / p>
增加线程数工作正常，我遇到的问题是当我尝试使用setCorePoolSize减少线程数时。
据我所知，这个值只有在线程空闲时才会被更改。由于不断处理任务，因此该线程永远不会空闲，因此永远不会关闭。
我应该使用什么来减少线程数？
这是我的演示代码：
package com.test.executortest;
import java.util.ArrayList;
import java.util.List;
import java.util.concurrent.LinkedBlockingQueue;
import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
/**
*
* @author peter.marcoen
*/
public class ExecutorTest {
final static Logger log = LogManager.getLogger(ExecutorTest.class);
public static void main(String[] args) throws InterruptedException {
ThreadPoolExecutor x = new ThreadPoolExecutor(1, 100, 30, TimeUnit.DAYS, new LinkedBlockingQueue<Runnable>());
for (int i = 1; i <= 10; i++) {
RunnableTest r = new RunnableTest();
r.num = String.valueOf(i);
x.execute(r);
}
x.shutdown();
int i = 0;
while (!x.isTerminated()) {
i++;
log.debug("Active count: {}, Core pool size: {}, Maximum pool size: {}, Pool size: {}", x.getActiveCount(), x.getCorePoolSize(), x.getMaximumPoolSize(), x.getPoolSize());
if (i == 2) {
log.info("!!!! Setting core pool size to 2 !!!!");
x.setCorePoolSize(2);
} else if (i == 10) {
log.info("!!!! Setting core pool size to 1 !!!!");
x.setCorePoolSize(1);
}
Thread.sleep(1000);
}
}
}
RunnableTest只是睡了5秒钟：
package com.test.executortest;
import java.util.logging.Level;
import java.util.logging.Logger;
import org.apache.logging.log4j.LogManager;
/**
*
* @author peter.marcoen
*/
public class RunnableTest implements Runnable {
public String num;
final static org.apache.logging.log4j.Logger log = LogManager.getLogger(RunnableTest.class);
@Override
public void run() {
log.debug("Started parsing #{}", num);
try {
Thread.sleep(5000);
} catch (InterruptedException ex) {
Logger.getLogger(RunnableTest.class.getName()).log(Level.SEVERE, null, ex);
}
log.debug("Finished parsing #{}", num);
}
}
这是输出。如您所见，即使将coreThreads设置为1，也总会处理2个任务：
11:43:45.696 - Active count: 1, Core pool size: 1, Maximum pool size: 100, Pool size: 1
11:43:45.696 - Started parsing #1
11:43:46.710 - Active count: 1, Core pool size: 1, Maximum pool size: 100, Pool size: 1
11:43:46.710 - !!!! Setting core pool size to 2 !!!!
11:43:46.710 - Started parsing #2
11:43:47.724 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:48.738 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:49.738 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:50.705 - Finished parsing #1
11:43:50.705 - Started parsing #3
11:43:50.752 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:51.719 - Finished parsing #2
11:43:51.719 - Started parsing #4
11:43:51.766 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:52.780 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:53.794 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:54.795 - Active count: 2, Core pool size: 2, Maximum pool size: 100, Pool size: 2
11:43:54.795 - !!!! Setting core pool size to 1 !!!!
11:43:55.715 - Finished parsing #3
11:43:55.715 - Started parsing #5
11:43:55.809 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:43:56.729 - Finished parsing #4
11:43:56.729 - Started parsing #6
11:43:56.822 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:43:57.836 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:43:58.851 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:43:59.865 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:00.723 - Finished parsing #5
11:44:00.723 - Started parsing #7
11:44:00.879 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:01.739 - Finished parsing #6
11:44:01.739 - Started parsing #8
11:44:01.880 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:02.894 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:03.908 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:04.922 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:05.733 - Finished parsing #7
11:44:05.733 - Started parsing #9
11:44:05.936 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:06.749 - Finished parsing #8
11:44:06.749 - Started parsing #10
11:44:06.936 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:07.937 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:08.937 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:09.951 - Active count: 2, Core pool size: 1, Maximum pool size: 100, Pool size: 2
11:44:10.747 - Finished parsing #9
11:44:10.965 - Active count: 1, Core pool size: 1, Maximum pool size: 100, Pool size: 1
11:44:11.762 - Finished parsing #10
答案 0 :(得分：1)
在过去，我创建了一个辅助ThreadPoolExecutor，而不是动态地更改核心ThreadPoolExecutor的大小（正如您所发现的那样有点不稳定）。从核心执行程序中窃取任务 - 这使我能够终止辅助执行程序而不会影响核心执行程序。理想情况下，我们只能让两个执行程序共享一个工作队列，但这似乎引入了奇怪的并发错误，并且比它的价值更麻烦 - 更好地保持队列分离并使用afterExecute保持辅助执行人的游泳池充满。
public class AuxiliaryExecutor extends ThreadPoolExecutor {
private final AtomicBoolean shutdown = new AtomicBoolean(false);
private final BlockingQueue<Runnable> coreQueue;
private void pollAndExecute() {
try {
this.execute(coreQueue.poll());
} catch(NullPointerException e) {
this.execute(new Runnable() {
public void run() {
try {
this.execute(coreQueue.take());
} catch(InterruptedException e) {
return;
}
}
});
}
}
public AuxiliaryExecutor(ThreadPoolExecutor coreExecutor, int poolSize) {
super(poolSize, poolSize, 30L, TimeUnit.SECONDS, new ArrayBlockingQueue<>(poolSize));
this.coreQueue = coreExecutor.getQueue();
for(int i = 0; i < poolSize; i++) {
pollAndExecute();
}
}
@Override
protected void afterExecute(Runnable r, Throwable t) {
if(!shutdown.get()) {
pollAndExecute();
}
}
@Override
public void shutdown() {
shutdown.set(true);
}
}
如果要更改辅助执行程序的大小，则可以关闭辅助执行程序并将其替换为新的已调整大小的执行程序。