一. 为什么不推荐使用类Executors
的静态方法创建线程池
阿里的Java规范不推荐使用类Executors
的静态方法创建线程池。至于为什么不推荐,通过看源码,如下:
public static ExecutorService newFixedThreadPool(int nThreads) {
return new ThreadPoolExecutor(nThreads, nThreads,
0L, TimeUnit.MILLISECONDS,
new LinkedBlockingQueue());
}
public static ExecutorService newSingleThreadExecutor() {
return new FinalizableDelegatedExecutorService
(new ThreadPoolExecutor(1, 1,
0L, TimeUnit.MILLISECONDS,
new LinkedBlockingQueue()));
}
public static ExecutorService newCachedThreadPool() {
return new ThreadPoolExecutor(0, Integer.MAX_VALUE,
60L, TimeUnit.SECONDS,
new SynchronousQueue());
}
public static ScheduledExecutorService newScheduledThreadPool(int corePoolSize) {
return new ScheduledThreadPoolExecutor(corePoolSize);
}
public ScheduledThreadPoolExecutor(int corePoolSize) {
super(corePoolSize, Integer.MAX_VALUE, 0, NANOSECONDS,
new DelayedWorkQueue());
}
可以得知原因如下:
-
newFixedThreadPool, newSingleThreadExecutor
两种方法没有限制LinkedBlockingQueue
队列的大小,当任务越来越多时,队列的大小会越来越大,可能会造成OOM。 -
newCachedThreadPool, newScheduledThreadPool
两种方法的线程最大数量为Integer.MAX_VALUE
,当任务越来越多时,也可能会造成OOM。
以上四种静态方法实际上都是通过new ThreadPoolExecutor()
实现的,所以我们可以根据实际需要创建符合自己要求的线程池。
二. 构造方法详解
public ThreadPoolExecutor(int corePoolSize,
int maximumPoolSize,
long keepAliveTime,
TimeUnit unit,
BlockingQueue workQueue,
ThreadFactory threadFactory,
RejectedExecutionHandler handler) {
...
}
下面解释一下构造参数解释:
-
corePoolSize
:核心池大小 -
maximumPoolSize
:线程池最大值 -
keepAliveTime
:非核心池的线程执行完毕后存活时间,可以通过pool.allowCoreThreadTimeOut(true)
方法设置核心池中线程也适用于存活时间限制。 -
unit
:存活时间的单位 -
workQueue
:任务队列,队列有很多种,比如ArrayBlockingQueue、LinkedBlockingDeque、SynchronousQueue
等三种队列,前面两种就是数据结构不同,前者数组,后者链表。SynchronousQueue
是一种阻塞队列,每个插入操作必须等待另一个线程的移除操作,反之亦然。任务不会在队列中停留,所以其内部没有容量,size()方法永远为0。仅当有足够多的消费者,并且总是有一个消费者准备好获取交付的工作时,才适合使用同步队列。 -
threadFactory
:线程工厂,创建线程的工厂类 -
handler
:拒绝策略,拒绝策略有四种:-
AbortPolicy
: 丢弃新任务,并抛出异常,默认策略 -
DiscardPolicy
: 丢弃新任务,不抛出异常 -
CallerRunsPolicy
: 交由调用execute方法的线程去执行新任务 -
DiscardOldestPolicy
: 接收新任务,但会丢弃队列头部的任务
-
三. 线程池执行任务的流程
- 先创建线程去执行任务,当
线程数量=corePoolSize
,后续的新任务都丢进队列。 - 当队列也满了,如果
maximumPoolSize>corePoolSize
,继续创建线程执行新任务,否则执行拒绝策略。 - 当
线程数量=maximumPoolSize
,后续的任务执行拒绝策略。
四. 举例验证线程池流程
下面是一个默认拒绝策略的例子:
public class ThreadPoolTest {
public static void main(String[] args) {
LinkedBlockingDeque deque = new LinkedBlockingDeque<>(3);
ThreadPoolExecutor pool = new ThreadPoolExecutor(3, 4, 0, TimeUnit.MILLISECONDS, deque);
for (int i = 0; i < 8; i++) {
System.out.println("PoolSize: "+ pool.getLargestPoolSize());
System.out.println("dequeSize: "+ deque.size());
int b = i;
pool.submit(() -> {
System.out.println(b+"is run...");
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
}
});
}
pool.shutdown();
}
}
打印结果:
PoolSize: 0
dequeSize: 0
Exception in thread "main" java.util.concurrent.RejectedExecutionException: Task java.util.concurrent.FutureTask@7ba4f24f rejected from java.util.concurrent.ThreadPoolExecutor@3b9a45b3[Running, pool size = 4, active threads = 4, queued tasks = 3, completed tasks = 0]
at java.util.concurrent.ThreadPoolExecutor$AbortPolicy.rejectedExecution(ThreadPoolExecutor.java:2047)
at java.util.concurrent.ThreadPoolExecutor.reject(ThreadPoolExecutor.java:823)
at java.util.concurrent.ThreadPoolExecutor.execute(ThreadPoolExecutor.java:1369)
at java.util.concurrent.AbstractExecutorService.submit(AbstractExecutorService.java:112)
at cn.tl.thread.ThreadPoolTest.main(ThreadPoolTest.java:29)
PoolSize: 1
dequeSize: 0
PoolSize: 2
0is run...
1is run...
dequeSize: 0
PoolSize: 3
dequeSize: 0
2is run...
PoolSize: 3
dequeSize: 1
PoolSize: 3
dequeSize: 2
PoolSize: 3
dequeSize: 3
PoolSize: 4
dequeSize: 3
6is run...
3is run...
5is run...
4is run...
很明显,当执行第8次循环时,线程数量=maximumPoolSize,线程池执行了默认拒绝策略AbortPolicy: 丢弃新任务,并抛出异常。所以设置线程数和队列容量的时候需根据实际需求指定。
再看一个同步队列和CallerRunsPolicy策略的示例:
public class SynchronousQueueTest {
public static void main(String[] args) {
ThreadPoolExecutor pool = new ThreadPoolExecutor(2, 2, 0, TimeUnit.SECONDS, new SynchronousQueue<>(), Executors.defaultThreadFactory(), new ThreadPoolExecutor.CallerRunsPolicy());
for (int i = 0; i < 5; i++) {
int b = i;
pool.execute(() -> {
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println(Thread.currentThread().getName() + " " + b);
});
}
pool.shutdown();
}
}
打印结果:
main 2
pool-1-thread-1 0
pool-1-thread-2 1
main 3
pool-1-thread-2 4
该示例中,线程池只能同时执行2个任务,多余的任务暂时交给主线程执行,当线程池有空闲的线程,就继续执行新任务。
五. 线程池命名
线程池默认命名不适合查看日志,可以根据需求设置自己的线程名,引入第三方的jar包,如下:
ThreadFactory namedThreadFactory = new ThreadFactoryBuilder()
.setNameFormat("demo-pool-%d").build();
ThreadPoolExecutor pool = new ThreadPoolExecutor(3, 4, 0, TimeUnit.MILLISECONDS, new LinkedBlockingDeque<>(3), namedThreadFactory);
com.google.guava
guava
23.0
六. 合理配置线程池大小
- I/O密集型,I/O是耗时操作,应配置尽可能多的线程。一般,线程数 = CPU核数的2倍。
- CPU密集型,CPU使用率较高(复杂计算等等),应配置尽可能小的线程,避免线程频繁切换。一般,线程数 = CPU核数+1。
以上。