@Async异步任务与线程池

7,752 阅读17分钟

写在前面:本篇文章是关于使用@Async进行异步任务,并且关于线程池做了一个初步的梳理和总结,包括遇到过的一些坑

在工作中用到的一些线程池

以下代码已做脱敏处理

1.newCachedThreadPool

    private void startTask(List<String> usersList){
        ExecutorService executor = Executors.newCachedThreadPool();
        executor.submit(()->{
		//do someting
        });
    }

2.newScheduledThreadPool


@Configuration
public class ScheduleConfig implements SchedulingConfigurer {

    @Override
    public void configureTasks(ScheduledTaskRegistrar taskRegistrar) {
        //当然了,这里设置的线程池是corePoolSize也是很关键了,自己根据业务需求设定
        taskRegistrar.setScheduler(Executors.newScheduledThreadPool(10));
    }

}

如果在idea中安装了阿里规范插件,就会发现上面两种创建线程池的方式都会报红。原因是:

线程池不允许使用Executors去创建,而是通过ThreadPoolExecutor的方式,这样的处理方式让写的同学更加明确线程池的运行规则,规避资源耗尽的风险。 说明:Executors返回的线程池对象的弊端如下:

  1. FixedThreadPool和SingleThreadPool:

    允许的请求队列长度为Integer.MAX_VALUE,可能会堆积大量的请求,从而导致OOM。

  2. CachedThreadPool:

    允许的创建线程数量为Integer.MAX_VALUE,可能会创建大量的线程,从而导致OOM。

其实这里CachedThreadPool和newScheduledThreadPool是一样的,都是因为最大线程数被设置成了Integer.MAX_VALUE。


    public ScheduledThreadPoolExecutor(int corePoolSize) {
        super(corePoolSize, Integer.MAX_VALUE, 0, NANOSECONDS,
              new DelayedWorkQueue());
    }
    public static ExecutorService newCachedThreadPool() {
        return new ThreadPoolExecutor(0, Integer.MAX_VALUE,
                                      60L, TimeUnit.SECONDS,
                                      new SynchronousQueue<Runnable>());
    }

在源码中可以看的出newCachedThreadPool使用的是synchronousqueue队列,也可以看作是一个长度为1的BlockingQueue所以,再加上最大允许线程数为Integer.MAX_VALUE,就导致可能会创建大量线程导致OOM。

同理ScheduledThreadPoolExecutor使用的是DelayedWorkQueue,初始化大小为16。当队列满后就会创建新线程,就导致可能会创建大量线程导致OOM。

我们不妨实际来测试一下,以newCachedThreadPool为例,jvm参数-Xms64m -Xmx192m -Xss1024K -XX:MetaspaceSize=64m -XX:MaxMetaspaceSize=128m。


    @PostMapping("/newCachedThreadPoolExample")
    @ResponseBody
    public void  newCachedThreadPoolExample(){
        ExecutorService executorService = Executors.newCachedThreadPool();
        while (true){
            executorService.submit(()->{
                log.info("submit:"+LocalDateTime.now());
                try {
                    Thread.sleep(1000);
                }catch (InterruptedException e){
                    e.printStackTrace();
                }
            });
        }

    }

刚启动时的情况:

x1.png

请求接口后就开始爆炸

x2.png 然后就开始卡着不动了

x3.png

比较尴尬的是一直没有出现报错OOM的情况,就直接卡死了。

x4.png 总结

以上的线程池虽然可以在外部限制的情况下避免OOM等情况,但是还是建议尽量根据自己的业务情况自定义线程池。

使用@Async快速创建一个异步任务

1. application.yml

这里是线程池相关配置,先不详细说,同理可以在代码里面配置config。

线程池缓冲队列的选择

以上发生的问题大多数都和线程池的缓冲队列有关,选择一个符合自己业务特点的缓冲队列也十分重要。

x5'.png

spring:
  task:
    execution:
      pool:
        # 最大线程数
        max-size: 16
        # 核心线程数
        core-size: 16
        # 存活时间
        keep-alive: 10s
        # 队列大小
        queue-capacity: 100
        # 是否允许核心线程超时
        allow-core-thread-timeout: true
      # 线程名称前缀
      thread-name-prefix: async-task-

2.ThreadpoolApplication

这里需要在 Application上添加 @EnableAsync注解,开启异步任务。如果是选择在代码里面写config,则需要在config文件上添加@EnableAsync注解。

@EnableAsync
@SpringBootApplication
public class ThreadpoolApplication {

    public static void main(String[] args) {
        SpringApplication.run(ThreadpoolApplication.class, args);
    }

}

3.AsyncTask

编写一个异步任务处理类,在需要开启异步的方法上面添加@Async

@Component
@Slf4j
public class AsyncTask {
    @Async
    public void  asyncRun() throws InterruptedException {
        Thread.sleep(10);
        log.info(Thread.currentThread().getName()+":处理完成");
    }
}

4.AsyncService

编写一个调用异步方法的service

@Service
@Slf4j
public class AsyncService {
    @Autowired
    private AsyncTask asyncTask;

    public void  asyncSimpleExample() {
        try {
            log.info("service start");
            asyncTask.asyncRun();
            log.info("service end");
        }catch (InterruptedException e){
            e.printStackTrace();
        }
    }


}

5.AsyncController

编写一个Controller去调用AsyncService


/**
 * @author kurtl
 */
@Controller
@RequestMapping("/")
public class AsyncController {
    @Autowired
    private AsyncService asyncService;
    @PostMapping("/asyncSimpleExample")
    @ResponseBody
    public void  asyncSimpleExample(){
        asyncService.asyncSimpleExample();
    }
}

最后请求这个接口

x6.png

可以看到,先输出了asyncSimpleExample里面打印的service start与service end,表示service方法先执行完毕了,而异步方法则在调用后进行了一个sleep,service没有同步等待sleep完成,而是直接返回,表示这个是异步任务。至此我们已经通过@Async成功创建的异步任务。

关于@Async和@EnableAsync的原理

个人觉得源码中很重要的一部分就是源码中的注释,阅读注释也可以帮你快速了解源码的作用等,所有我会把重要的注释稍微翻译一下

1.@Async源码




@Target({ElementType.TYPE, ElementType.METHOD})
@Retention(RetentionPolicy.RUNTIME)
@Documented
public @interface Async {

	/**
	 * A qualifier value for the specified asynchronous operation(s).
	 * <p>May be used to determine the target executor to be used when executing
	 * the asynchronous operation(s), matching the qualifier value (or the bean
	 * name) of a specific {@link java.util.concurrent.Executor Executor} or
	 * {@link org.springframework.core.task.TaskExecutor TaskExecutor}
	 * bean definition.
	 * <p>When specified on a class-level {@code @Async} annotation, indicates that the
	 * given executor should be used for all methods within the class. Method-level use
	 * of {@code Async#value} always overrides any value set at the class level.
	 * @since 3.1.2
	 */

    /**
     * 在这些注释中有三个非常重要的部分
     * 1.使用@Async的方法只能返回Void 或者 Future类型
     * 2.表明了@Async是通过org.springframework.core.task.TaskExecutor
     * 或者java.util.concurrent.Executor来创建线程池
     * 3.写了@Async的作用范围 在类上使用@Async会覆盖方法上的@Async
     */

	String value() default "";

}

2.@EnableAsync源码




/**
 * Enables Spring's asynchronous method execution capability, similar to functionality
 * found in Spring's {@code <task:*>} XML namespace.
 *
 * <p>To be used together with @{@link Configuration Configuration} classes as follows,
 * enabling annotation-driven async processing for an entire Spring application context:
 *
 * <pre class="code">
 * &#064;Configuration
 * &#064;EnableAsync
 * public class AppConfig {
 *
 * }</pre>
 *  这里表示需要联合@Configuration注解一起使用,所以@EnableAsync应该
 *  添加在线程池Config或者SpringBootApplication 上
 * {@code MyAsyncBean} is a user-defined type with one or more methods annotated with
 * either Spring's {@code @Async} annotation, the EJB 3.1 {@code @javax.ejb.Asynchronous}
 * annotation, or any custom annotation specified via the {@link #annotation} attribute.
 * The aspect is added transparently for any registered bean, for instance via this
 * configuration:
 *
 * <pre class="code">
 * &#064;Configuration
 * public class AnotherAppConfig {
 *
 *     &#064;Bean
 *     public MyAsyncBean asyncBean() {
 *         return new MyAsyncBean();
 *     }
 * }</pre>
 *
 * <p>By default, Spring will be searching for an associated thread pool definition:
 * either a unique {@link org.springframework.core.task.TaskExecutor} bean in the context,
 * or an {@link java.util.concurrent.Executor} bean named "taskExecutor" otherwise. If
 * neither of the two is resolvable, a {@link org.springframework.core.task.SimpleAsyncTaskExecutor}
 * 默认情况下spring会先搜索TaskExecutor类型的bean或者名字为 
 * taskExecutor的Executor类型的bean,都不存在使用 
 * SimpleAsyncTaskExecutor执行器但是这个SimpleAsyncTaskExecutor实际 
 * 上是有很大的坑的,建议是自定义一个线程池,这个后面会说
 * will be used to process async method invocations. Besides, annotated methods having 
 *
 * @author Chris Beams
 * @author Juergen Hoeller
 * @author Stephane Nicoll
 * @author Sam Brannen
 * @since 3.1
 * @see Async
 * @see AsyncConfigurer
 * @see AsyncConfigurationSelector
 */
@Target(ElementType.TYPE)
@Retention(RetentionPolicy.RUNTIME)
@Documented
@Import(AsyncConfigurationSelector.class)
public @interface EnableAsync {

	/**
	 * Indicate the 'async' annotation type to be detected at either class
	 * or method level.
	 * <p>By default, both Spring's @{@link Async} annotation and the EJB 3.1
	 * {@code @javax.ejb.Asynchronous} annotation will be detected.
	 * <p>This attribute exists so that developers can provide their own
	 * custom annotation type to indicate that a method (or all methods of
	 * a given class) should be invoked asynchronously.
	 */
	Class<? extends Annotation> annotation() default Annotation.class;

	/**
	 * Indicate whether subclass-based (CGLIB) proxies are to be created as opposed
	 * to standard Java interface-based proxies.
	 * <p><strong>Applicable only if the {@link #mode} is set to {@link AdviceMode#PROXY}</strong>.
	 * <p>The default is {@code false}.
	 * <p>Note that setting this attribute to {@code true} will affect <em>all</em>
	 * Spring-managed beans requiring proxying, not just those marked with {@code @Async}.
	 * For example, other beans marked with Spring's {@code @Transactional} annotation
	 * will be upgraded to subclass proxying at the same time. This approach has no
	 * negative impact in practice unless one is explicitly expecting one type of proxy
	 * vs. another &mdash; for example, in tests.
	 *
	 * 这个字段用来表示,是否要创建基于CGLIB的代理,实际上在高版本		 			 
         * 的spring 上(大概3.x)是自动选择使用jdk动态代理还是CGLIB.
	 * 设置为true时,其它spring管理的bean也会升级到CGLIB代理
	*/
	boolean proxyTargetClass() default false;

	/**
	 * Indicate how async advice should be applied.
	 * <p><b>The default is {@link AdviceMode#PROXY}.</b>
	 * Please note that proxy mode allows for interception of calls through the proxy
	 * only. Local calls within the same class cannot get intercepted that way; an
	 * {@link Async} annotation on such a method within a local call will be ignored
	 * since Spring's interceptor does not even kick in for such a runtime scenario.
	 * For a more advanced mode of interception, consider switching this to
	 * {@link AdviceMode#ASPECTJ}.
	 * 这个字段用来标识异步通知的模式,默认PROXY,当这个字段为	 	 
         * PROXY的时候,在同一个类中,非异步方法调用异步方法,会导致异	 
         * 步不生效,相反如果,想实现同一个类非异步方法调用异步方法就应 
         * 该设置为ASPECTJ
	 */
	AdviceMode mode() default AdviceMode.PROXY;

	/**
	 * Indicate the order in which the {@link AsyncAnnotationBeanPostProcessor}
	 * should be applied.
	 * <p>The default is {@link Ordered#LOWEST_PRECEDENCE} in order to run
	 * after all other post-processors, so that it can add an advisor to
	 * existing proxies rather than double-proxy.
	 * 标明异步注解bean处理器应该遵循的执行顺序,默认最低的优先级 
         *(Integer.MAX_VALUE,值越小优先级越高)
	 */
	int order() default Ordered.LOWEST_PRECEDENCE;

}


在上面的源码中,其实最核心的代码只有一句,@Import(AsyncConfigurationSelector.class),引入了相关的配置。




/**
 * Selects which implementation of {@link AbstractAsyncConfiguration} should
 * be used based on the value of {@link EnableAsync#mode} on the importing
 * {@code @Configuration} class.
 *
 * @author Chris Beams
 * @author Juergen Hoeller
 * @since 3.1
 * @see EnableAsync
 * @see ProxyAsyncConfiguration
 */
public class AsyncConfigurationSelector extends AdviceModeImportSelector<EnableAsync> {

	private static final String ASYNC_EXECUTION_ASPECT_CONFIGURATION_CLASS_NAME =
			"org.springframework.scheduling.aspectj.AspectJAsyncConfiguration";


	/**
	 * Returns {@link ProxyAsyncConfiguration} or {@code AspectJAsyncConfiguration}
	 * for {@code PROXY} and {@code ASPECTJ} values of {@link EnableAsync#mode()},
	 * respectively.
	 */
	/**
	 * 这整个方法其实就是一个选择器和ImportSelector接口的selectImports()方法很像,基于不同的代理模式,加载不同的配置类
	 */
	@Override
	@Nullable
	public String[] selectImports(AdviceMode adviceMode) {

		switch (adviceMode) {
			case PROXY:
				return new String[] {ProxyAsyncConfiguration.class.getName()};
			case ASPECTJ:
				return new String[] {ASYNC_EXECUTION_ASPECT_CONFIGURATION_CLASS_NAME};
			default:
				return null;
		}
	}

}

接下来我们看看默认的ProxyAsyncConfiguration.class




/**
 * {@code @Configuration} class that registers the Spring infrastructure beans necessary
 * to enable proxy-based asynchronous method execution.
 *
 * @author Chris Beams
 * @author Stephane Nicoll
 * @author Juergen Hoeller
 * @since 3.1
 * @see EnableAsync
 * @see AsyncConfigurationSelector
 */
@Configuration
@Role(BeanDefinition.ROLE_INFRASTRUCTURE)
//继承了AbstractAsyncConfiguration类
public class ProxyAsyncConfiguration extends AbstractAsyncConfiguration {

	@Bean(name = TaskManagementConfigUtils.ASYNC_ANNOTATION_PROCESSOR_BEAN_NAME)
	@Role(BeanDefinition.ROLE_INFRASTRUCTURE)
	public AsyncAnnotationBeanPostProcessor asyncAdvisor() {
		Assert.notNull(this.enableAsync, "@EnableAsync annotation metadata was not injected");
		//初始化AsyncAnnotationBeanPostProcessor类型的bean
		AsyncAnnotationBeanPostProcessor bpp = new AsyncAnnotationBeanPostProcessor();
		//设置执行器和异常处理器
		bpp.configure(this.executor, this.exceptionHandler);
		Class<? extends Annotation> customAsyncAnnotation = this.enableAsync.getClass("annotation");
		//设置annotation
		if (customAsyncAnnotation != AnnotationUtils.getDefaultValue(EnableAsync.class, "annotation")) {
			bpp.setAsyncAnnotationType(customAsyncAnnotation);
		}
		//设置注解属性
		bpp.setProxyTargetClass(this.enableAsync.getBoolean("proxyTargetClass"));
		bpp.setOrder(this.enableAsync.<Integer>getNumber("order"));
		return bpp;
	}

}



这一个类继承了AbstractAsyncConfiguration类,其实也就做了一件事初始化AsyncAnnotationBeanPostProcessor,@Async注解的就是通过AsyncAnnotationBeanPostProcessor这个后置处理器生成一个代理对象来实现异步的,我们先看继承的config。




/**
 * Abstract base {@code Configuration} class providing common structure for enabling
 * Spring's asynchronous method execution capability.
 *
 * @author Chris Beams
 * @author Juergen Hoeller
 * @author Stephane Nicoll
 * @since 3.1
 * @see EnableAsync
 */
@Configuration
public abstract class AbstractAsyncConfiguration implements ImportAware {

	@Nullable
	protected AnnotationAttributes enableAsync; //;//enableAsync的注解属性

	@Nullable
	protected Supplier<Executor> executor; //线程执行器

	@Nullable
	protected Supplier<AsyncUncaughtExceptionHandler> exceptionHandler; //异常处理器 和上面的代码对应


	@Override
	//设置注解属性
	public void setImportMetadata(AnnotationMetadata importMetadata) {
		this.enableAsync = AnnotationAttributes.fromMap(
				importMetadata.getAnnotationAttributes(EnableAsync.class.getName(), false));
		if (this.enableAsync == null) {
			throw new IllegalArgumentException(
					"@EnableAsync is not present on importing class " + importMetadata.getClassName());
		}
	}

	/**
	 * Collect any {@link AsyncConfigurer} beans through autowiring.
	 */
	@Autowired(required = false)
	//设置执行器和异常处理器
	void setConfigurers(Collection<AsyncConfigurer> configurers) {
		if (CollectionUtils.isEmpty(configurers)) {
			return;
		}
		if (configurers.size() > 1) {
			throw new IllegalStateException("Only one AsyncConfigurer may exist");
		}
		AsyncConfigurer configurer = configurers.iterator().next();
		this.executor = configurer::getAsyncExecutor;
		this.exceptionHandler = configurer::getAsyncUncaughtExceptionHandler;
	}

}


整个代码的结构其实非常明确,我们回到上一个类,看他设置的bean AsyncAnnotationBeanPostProcessor。这个bean很复杂,所以干脆先生成类图。弄清楚baen的生命周期。AsyncAnnotationBeanPostProcessor是一个后置处理器,所以我们先找父类AbstractAdvisingBeanPostProcessor中。

x9.png




/**
 * Base class for {@link BeanPostProcessor} implementations that apply a
 * Spring AOP {@link Advisor} to specific beans.
 *
 * @author Juergen Hoeller
 * @since 3.2
 */
@SuppressWarnings("serial")
public abstract class AbstractAdvisingBeanPostProcessor extends ProxyProcessorSupport implements BeanPostProcessor {

	@Nullable
	protected Advisor advisor;

	protected boolean beforeExistingAdvisors = false;

	private final Map<Class<?>, Boolean> eligibleBeans = new ConcurrentHashMap<>(256);



	public void setBeforeExistingAdvisors(boolean beforeExistingAdvisors) {
		this.beforeExistingAdvisors = beforeExistingAdvisors;
	}


	@Override
	public Object postProcessBeforeInitialization(Object bean, String beanName) {
		return bean;
	}

	@Override
	public Object postProcessAfterInitialization(Object bean, String beanName) {
		 // 没有通知,或者是AopInfrastructureBean,那么不进行代理
		if (this.advisor == null || bean instanceof AopInfrastructureBean) {
			// Ignore AOP infrastructure such as scoped proxies.
			return bean;
		}
		// 添加advisor
		if (bean instanceof Advised) {
			Advised advised = (Advised) bean;
			if (!advised.isFrozen() && isEligible(AopUtils.getTargetClass(bean))) {
				// Add our local Advisor to the existing proxy's Advisor chain...
				// 这里通过beforeExistingAdvisors决定是将通知添加到所有通知之前还是添加到所有通知之后
				// 默认false 在@Async中被设置为true
				if (this.beforeExistingAdvisors) {
					advised.addAdvisor(0, this.advisor);
				}
				else {
					advised.addAdvisor(this.advisor);
				}
				return bean;
			}
		}
		//构造ProxyFactory代理工厂
		if (isEligible(bean, beanName)) {
			ProxyFactory proxyFactory = prepareProxyFactory(bean, beanName);
			//添加代理的接口
			if (!proxyFactory.isProxyTargetClass()) {
				evaluateProxyInterfaces(bean.getClass(), proxyFactory);
			}
			//设置切面
			proxyFactory.addAdvisor(this.advisor);
			customizeProxyFactory(proxyFactory);
			//返回代理类
			return proxyFactory.getProxy(getProxyClassLoader());
		}

		// No proxy needed.
		return bean;
	}

	//isEligible用于判断这个类或者这个类中的某个方法是否含有注解
	protected boolean isEligible(Object bean, String beanName) {
		return isEligible(bean.getClass());
	}


}


在上面代码中可以看出来,proxyFactory.addAdvisor(this.advisor);这里持有一个AsyncAnnotationAdvisor类的对象advisor:buildAdvice()方法生成通知,buildPointcut生成切点。定位到这个类的buildPointcut方法中,看看他的切点匹配规则。




@SuppressWarnings("serial")
public class AsyncAnnotationAdvisor extends AbstractPointcutAdvisor implements BeanFactoryAware {

	private Advice advice;

	private Pointcut pointcut;


	/**
	 * Create a new {@code AsyncAnnotationAdvisor} for bean-style configuration.
	 */
	public AsyncAnnotationAdvisor() {
		this((Supplier<Executor>) null, (Supplier<AsyncUncaughtExceptionHandler>) null);
	}


	@SuppressWarnings("unchecked")
	public AsyncAnnotationAdvisor(
			@Nullable Executor executor, @Nullable AsyncUncaughtExceptionHandler exceptionHandler) {

		this(SingletonSupplier.ofNullable(executor), SingletonSupplier.ofNullable(exceptionHandler));
	}


	@SuppressWarnings("unchecked")
	public AsyncAnnotationAdvisor(
			@Nullable Supplier<Executor> executor, @Nullable Supplier<AsyncUncaughtExceptionHandler> exceptionHandler) {

		Set<Class<? extends Annotation>> asyncAnnotationTypes = new LinkedHashSet<>(2);
		asyncAnnotationTypes.add(Async.class);
		try {
			asyncAnnotationTypes.add((Class<? extends Annotation>)
					ClassUtils.forName("javax.ejb.Asynchronous", AsyncAnnotationAdvisor.class.getClassLoader()));
		}
		catch (ClassNotFoundException ex) {
			// If EJB 3.1 API not present, simply ignore.
		}
		this.advice = buildAdvice(executor, exceptionHandler);
		this.pointcut = buildPointcut(asyncAnnotationTypes);
	}



	public void setAsyncAnnotationType(Class<? extends Annotation> asyncAnnotationType) {
		Assert.notNull(asyncAnnotationType, "'asyncAnnotationType' must not be null");
		Set<Class<? extends Annotation>> asyncAnnotationTypes = new HashSet<>();
		asyncAnnotationTypes.add(asyncAnnotationType);
		this.pointcut = buildPointcut(asyncAnnotationTypes);
	}

	/**
	 * Set the {@code BeanFactory} to be used when looking up executors by qualifier.
	 */
	@Override
	public void setBeanFactory(BeanFactory beanFactory) {
		if (this.advice instanceof BeanFactoryAware) {
			((BeanFactoryAware) this.advice).setBeanFactory(beanFactory);
		}
	}


	@Override
	public Advice getAdvice() {
		return this.advice;
	}

	@Override
	public Pointcut getPointcut() {
		return this.pointcut;
	}

	//构建通知,一个简单的拦截器
	protected Advice buildAdvice(
			@Nullable Supplier<Executor> executor, @Nullable Supplier<AsyncUncaughtExceptionHandler> exceptionHandler) {

		AnnotationAsyncExecutionInterceptor interceptor = new AnnotationAsyncExecutionInterceptor(null);
		interceptor.configure(executor, exceptionHandler);
		return interceptor;
	}


	protected Pointcut buildPointcut(Set<Class<? extends Annotation>> asyncAnnotationTypes) {
		ComposablePointcut result = null;
		for (Class<? extends Annotation> asyncAnnotationType : asyncAnnotationTypes) {
			// 根据cpc和mpc 匹配器进行匹配
			//检查类上是否有@Async注解
			Pointcut cpc = new AnnotationMatchingPointcut(asyncAnnotationType, true);
			//检查方法是是否有@Async注解。
			Pointcut mpc = new AnnotationMatchingPointcut(null, asyncAnnotationType, true);
			if (result == null) {
				result = new ComposablePointcut(cpc);
			}
			else {
				result.union(cpc);
			}
			result = result.union(mpc);
		}
		return (result != null ? result : Pointcut.TRUE);
	}

}


再找到它的通知逻辑buildAdvice,就是一个拦截器,生成AnnotationAsyncExecutionInterceptor对象,对于Interceptor,关注它的核心方法invoke就行了。它的父类AsyncExecutionInterceptor重写了AsyncExecutionInterceptor接口的invoke方法。代码如下





public class AsyncExecutionInterceptor extends AsyncExecutionAspectSupport implements MethodInterceptor, Ordered {


	public AsyncExecutionInterceptor(@Nullable Executor defaultExecutor) {
		super(defaultExecutor);
	}

	public AsyncExecutionInterceptor(@Nullable Executor defaultExecutor, AsyncUncaughtExceptionHandler exceptionHandler) {
		super(defaultExecutor, exceptionHandler);
	}



	@Override
	@Nullable
	//
	public Object invoke(final MethodInvocation invocation) throws Throwable {
		Class<?> targetClass = (invocation.getThis() != null ? AopUtils.getTargetClass(invocation.getThis()) : null);
		Method specificMethod = ClassUtils.getMostSpecificMethod(invocation.getMethod(), targetClass);
		final Method userDeclaredMethod = BridgeMethodResolver.findBridgedMethod(specificMethod);
		// 获取到一个线程池
		AsyncTaskExecutor executor = determineAsyncExecutor(userDeclaredMethod);
		if (executor == null) {
			throw new IllegalStateException(
					"No executor specified and no default executor set on AsyncExecutionInterceptor either");
		}
		 // 然后将这个方法封装成一个 Callable对象传入到线程池中执行
		Callable<Object> task = () -> {
			try {
				Object result = invocation.proceed();
				if (result instanceof Future) {
					//阻塞等待执行完毕得到结果
					return ((Future<?>) result).get();
				}
			}
			catch (ExecutionException ex) {
				handleError(ex.getCause(), userDeclaredMethod, invocation.getArguments());
			}
			catch (Throwable ex) {
				handleError(ex, userDeclaredMethod, invocation.getArguments());
			}
			return null;
		};

		return doSubmit(task, executor, invocation.getMethod().getReturnType());
	}


	@Override
	@Nullable
	protected String getExecutorQualifier(Method method) {
		return null;
	}


	@Override
	@Nullable
	protected Executor getDefaultExecutor(@Nullable BeanFactory beanFactory) {
		Executor defaultExecutor = super.getDefaultExecutor(beanFactory);
		return (defaultExecutor != null ? defaultExecutor : new SimpleAsyncTaskExecutor());
	}

	@Override
	public int getOrder() {
		return Ordered.HIGHEST_PRECEDENCE;
	}

}


可以看到,invoke首先是包装了一个Callable的对象,然后传入doSubmit,所以代码的核心就在doSubmit这个方法中。


	@Nullable
	protected Object doSubmit(Callable<Object> task, AsyncTaskExecutor executor, Class<?> returnType) {
	//先判断是否存在CompletableFuture这个类,优先使用CompletableFuture执行任务
		if (CompletableFuture.class.isAssignableFrom(returnType)) {
			return CompletableFuture.supplyAsync(() -> {
				try {
					return task.call();
				}
				catch (Throwable ex) {
					throw new CompletionException(ex);
				}
			}, executor);
		}
		else if (ListenableFuture.class.isAssignableFrom(returnType)) {
			return ((AsyncListenableTaskExecutor) executor).submitListenable(task);
		}
		else if (Future.class.isAssignableFrom(returnType)) {
			return executor.submit(task);
		}
		else {
			executor.submit(task);
			return null;
		}
	}

这里主要是判断不同的返回值,最终都走进了submit方法,而submit根据线程池的不同,其实现也有区别,下面是SimpleAsyncTaskExecutor的实现方式。


	/**
	 * Template method for the actual execution of a task.
	 * <p>The default implementation creates a new Thread and starts it.
	 * @param task the Runnable to execute
	 * @see #setThreadFactory
	 * @see #createThread
	 * @see java.lang.Thread#start()
	 */
	protected void doExecute(Runnable task) {
		Thread thread = (this.threadFactory != null ? this.threadFactory.newThread(task) : createThread(task));
		thread.start();
	}


@Async的默认线程池

1.使用@Async一定要定义线程池

在上面的源码中写的很清楚,默认情况下spring会先搜索TaskExecutor类型的bean或者名字为taskExecutor的Executor类型的bean,都不存在使 SimpleAsyncTaskExecutor执行器。但是这个SimpleAsyncTaskExecutor不是真的线程池,这个类不重用线程,每次调用都会创建一个新的线程。很有可能导致OOM。





@SuppressWarnings("serial")
public class SimpleAsyncTaskExecutor extends CustomizableThreadCreator
		implements AsyncListenableTaskExecutor, Serializable {

	/**
	 * Permit any number of concurrent invocations: that is, don't throttle concurrency.
	 * @see ConcurrencyThrottleSupport#UNBOUNDED_CONCURRENCY
	 */
	public static final int UNBOUNDED_CONCURRENCY = ConcurrencyThrottleSupport.UNBOUNDED_CONCURRENCY;

	/**
	 * Switch concurrency 'off': that is, don't allow any concurrent invocations.
	 * @see ConcurrencyThrottleSupport#NO_CONCURRENCY
	 */
	public static final int NO_CONCURRENCY = ConcurrencyThrottleSupport.NO_CONCURRENCY;


	/** Internal concurrency throttle used by this executor. */
	private final ConcurrencyThrottleAdapter concurrencyThrottle = new ConcurrencyThrottleAdapter();

	@Nullable
	private ThreadFactory threadFactory;

	@Nullable
	private TaskDecorator taskDecorator;


	/**
	 * Create a new SimpleAsyncTaskExecutor with default thread name prefix.
	 */
	public SimpleAsyncTaskExecutor() {
		super();
	}

	/**
	 * Create a new SimpleAsyncTaskExecutor with the given thread name prefix.
	 * @param threadNamePrefix the prefix to use for the names of newly created threads
	 */
	public SimpleAsyncTaskExecutor(String threadNamePrefix) {
		super(threadNamePrefix);
	}

	/**
	 * Create a new SimpleAsyncTaskExecutor with the given external thread factory.
	 * @param threadFactory the factory to use for creating new Threads
	 */
	public SimpleAsyncTaskExecutor(ThreadFactory threadFactory) {
		this.threadFactory = threadFactory;
	}


	/**
	 * Specify an external factory to use for creating new Threads,
	 * instead of relying on the local properties of this executor.
	 * <p>You may specify an inner ThreadFactory bean or also a ThreadFactory reference
	 * obtained from JNDI (on a Java EE 6 server) or some other lookup mechanism.
	 * @see #setThreadNamePrefix
	 * @see #setThreadPriority
	 */
	public void setThreadFactory(@Nullable ThreadFactory threadFactory) {
		this.threadFactory = threadFactory;
	}

	/**
	 * Return the external factory to use for creating new Threads, if any.
	 */
	@Nullable
	public final ThreadFactory getThreadFactory() {
		return this.threadFactory;
	}


	public final void setTaskDecorator(TaskDecorator taskDecorator) {
		this.taskDecorator = taskDecorator;
	}


	//这里可以设置最大线程数,通过限流去限制线程数
	public void setConcurrencyLimit(int concurrencyLimit) {
		this.concurrencyThrottle.setConcurrencyLimit(concurrencyLimit);
	}

	/**
	 * Return the maximum number of parallel accesses allowed.
	 */
	public final int getConcurrencyLimit() {
		return this.concurrencyThrottle.getConcurrencyLimit();
	}

	/**
	 * Return whether this throttle is currently active.
	 * @return {@code true} if the concurrency limit for this instance is active
	 * @see #getConcurrencyLimit()
	 * @see #setConcurrencyLimit
	 */
	public final boolean isThrottleActive() {
		return this.concurrencyThrottle.isThrottleActive();
	}


	/**
	 * Executes the given task, within a concurrency throttle
	 * if configured (through the superclass's settings).
	 * @see #doExecute(Runnable)
	 */
	@Override
	public void execute(Runnable task) {
		execute(task, TIMEOUT_INDEFINITE);
	}

	/**
	 * Executes the given task, within a concurrency throttle
	 * if configured (through the superclass's settings).
	 * <p>Executes urgent tasks (with 'immediate' timeout) directly,
	 * bypassing the concurrency throttle (if active). All other
	 * tasks are subject to throttling.
	 * @see #TIMEOUT_IMMEDIATE
	 * @see #doExecute(Runnable)
	 */
	//
	@Override
	public void execute(Runnable task, long startTimeout) {
		Assert.notNull(task, "Runnable must not be null");
		Runnable taskToUse = (this.taskDecorator != null ? this.taskDecorator.decorate(task) : task);
		if (isThrottleActive() && startTimeout > TIMEOUT_IMMEDIATE) {
			this.concurrencyThrottle.beforeAccess();
			doExecute(new ConcurrencyThrottlingRunnable(taskToUse));
		}
		else {
			doExecute(taskToUse);
		}
	}

	@Override
	public Future<?> submit(Runnable task) {
		FutureTask<Object> future = new FutureTask<>(task, null);
		execute(future, TIMEOUT_INDEFINITE);
		return future;
	}

	@Override
	public <T> Future<T> submit(Callable<T> task) {
		FutureTask<T> future = new FutureTask<>(task);
		execute(future, TIMEOUT_INDEFINITE);
		return future;
	}

	@Override
	public ListenableFuture<?> submitListenable(Runnable task) {
		ListenableFutureTask<Object> future = new ListenableFutureTask<>(task, null);
		execute(future, TIMEOUT_INDEFINITE);
		return future;
	}

	@Override
	public <T> ListenableFuture<T> submitListenable(Callable<T> task) {
		ListenableFutureTask<T> future = new ListenableFutureTask<>(task);
		execute(future, TIMEOUT_INDEFINITE);
		return future;
	}

	/**
	 * Template method for the actual execution of a task.
	 * <p>The default implementation creates a new Thread and starts it.
	 * @param task the Runnable to execute
	 * @see #setThreadFactory
	 * @see #createThread
	 * @see java.lang.Thread#start()
	 */
	//判断是否有工厂,没有的话调用父类创建线程
	protected void doExecute(Runnable task) {
		Thread thread = (this.threadFactory != null ? this.threadFactory.newThread(task) : createThread(task));
		thread.start();
	}


	/**
	 * Subclass of the general ConcurrencyThrottleSupport class,
	 * making {@code beforeAccess()} and {@code afterAccess()}
	 * visible to the surrounding class.
	 */
	private static class ConcurrencyThrottleAdapter extends ConcurrencyThrottleSupport {

		@Override
		protected void beforeAccess() {
			super.beforeAccess();
		}

		@Override
		protected void afterAccess() {
			super.afterAccess();
		}
	}


	/**
	 * This Runnable calls {@code afterAccess()} after the
	 * target Runnable has finished its execution.
	 */
	private class ConcurrencyThrottlingRunnable implements Runnable {

		private final Runnable target;

		public ConcurrencyThrottlingRunnable(Runnable target) {
			this.target = target;
		}

		@Override
		public void run() {
			try {
				this.target.run();
			}
			finally {
				concurrencyThrottle.afterAccess();
			}
		}
	}

}

最主要的就是这段代码


	/**
	 * Template method for the actual execution of a task.
	 * <p>The default implementation creates a new Thread and starts it.
	 * @param task the Runnable to execute
	 * @see #setThreadFactory
	 * @see #createThread
	 * @see java.lang.Thread#start()
	 */
	//判断是否有工厂,没有的话调用父类创建线程
	protected void doExecute(Runnable task) {
		Thread thread = (this.threadFactory != null ? this.threadFactory.newThread(task) : createThread(task));
		thread.start();
	}

这里并不是用线程池,而是直接创建新的线程,所以会大量创建线程导致OOM。其实这个类是可以通过setConcurrencyLimit设置最大线程数,通过synchronized和wati and notify去进行限流,这里不展开讲。所以结论是在使用@Async一定要设置线程池。

@Async异步失效

以下代码已做脱敏处理

在看公司代码的时候,发现这样一段代码

    public UserVO saveUser(HttpServletRequest request,
                                       String source) {
        String token = RequestUtils.getToken(request);
        String uid = checkUserLoginReturnUid(token);
        log.info("注册登录, token : {}, uid : {}", token, uid);
        //获取用户信息
        User User = getLoginUser(uid);
        if(User == null){
            User = new User();
            //获取用户信息
            Map<String,String> userMap = redisTemplateMain.getUserMapByToken(token);
            //保存用户
            saveUser(User, userMap, source);
            sendUserSystem(Integer.valueOf(userMap.get("id")));
        }
        //用户信息放进缓存
        setAuth2Redis(User);
        return setUser2Redis(User);
    }


    //通知用户系统,我们这边成功注册了一个用户
    @Async
    public void sendUserSystem(Integer userId){
        Map<String,Object> map = new HashMap<>();
        map.put("mainUid", userId);
        map.put("source", "");
        String json = HttpUtil.post(property.userRegisterSendSystem, map);
        log.info("sendUserSystem  userId : {}, json : {}", userId, json);
    }

在之前我们看源码的时候已经知道了,由于@Async的AdviceMode默认为PROXY,所以当调用方和被调用方是在同一个类中,无法产生切面,@Async没有被Spring容器管理。 所以这个方法跑了这么久一直是同步。

我们可以写一个方法去测试一下。


    public void  asyncInvalid() {
        try {
            log.info("service start");
            asyncInvalidExample();
            log.info("service end");
        }catch (InterruptedException e){
            e.printStackTrace();
        }
    }


    @Async
    public void  asyncInvalidExample() throws InterruptedException{
        Thread.sleep(10);
        log.info(Thread.currentThread().getName()+":处理完成");
    }


调用结果很明显,没有进行异步操作,而是同步。

x7.1.png

线程池拒绝导致线程丢失

既然线程池都已一个缓冲队列来保存未被消费的任务,那么就一定存在队列被塞满,导致线程丢失的情况。我们写一段代码模拟一下。

配置文件

spring:
  task:
    execution:
      pool:
        # 最大线程数
        max-size: 16
        # 核心线程数
        core-size: 16
        # 存活时间
        keep-alive: 10s
        # 队列大小
        queue-capacity: 100
        # 是否允许核心线程超时
        allow-core-thread-timeout: true
      # 线程名称前缀
      thread-name-prefix: async-task-

异步方法


    @Async
    public void  asyncRefuseRun() throws InterruptedException {
        Thread.sleep(5000000);
    }

调用方法



    public void  asyncRefuseRun() {
        for (int t = 0;t<2000;t++){
            log.info(""+t);
            try {
                asyncTask.asyncRefuseRun();
            }catch (InterruptedException e){
                e.printStackTrace();
            }
        }
    }

这里我循环了2000个线程,理论上来说当线程到达maxPoolSize + queueCapacity时会进行拒绝,也就是16+100。

x8.png

到了116的时候果然抛出了异常java.util.concurrent.RejectedExecutionException。证明线程执行了它的拒绝策略。

要理解线程池的拒绝策略,先来看看它的接口。


public interface RejectedExecutionHandler {
    void rejectedExecution(Runnable r, ThreadPoolExecutor executor);
}

当线程池出现拒绝的情况,就会来调用你设置的拒绝策略,将当前提交的任务以及线程池实例本身传递给你处理。这里建议根据自己的业务场景,去实现拒绝策略。

当然如果JDK内置的实现可以满足当前业务,可以直接用jdk实现的。

AbortPolicy(中止策略)

这个中止策略就是我们刚刚演示的,触发拒绝策略后,直接中止任务,抛出异常,这个也是ThreadPoolExecutor默认实现。

   /**
     * A handler for rejected tasks that throws a
     * {@code RejectedExecutionException}.
     */
    public static class AbortPolicy implements RejectedExecutionHandler {
        /**
         * Creates an {@code AbortPolicy}.
         */
        public AbortPolicy() { }

        /**
         * Always throws RejectedExecutionException.
         *
         * @param r the runnable task requested to be executed
         * @param e the executor attempting to execute this task
         * @throws RejectedExecutionException always
         */
        public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
            throw new RejectedExecutionException("Task " + r.toString() +
                                                 " rejected from " +
                                                 e.toString());
        }
    }


DiscardPolicy(丢弃策略)

    /**
     * A handler for rejected tasks that silently discards the
     * rejected task.
     */
    public static class DiscardPolicy implements RejectedExecutionHandler {
        /**
         * Creates a {@code DiscardPolicy}.
         */
        public DiscardPolicy() { }

        /**
         * Does nothing, which has the effect of discarding task r.
         *
         * @param r the runnable task requested to be executed
         * @param e the executor attempting to execute this task
         */
        public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
        }
    }

很明显,啥也不干,就是一个空实现。

DiscardOldestPolicy(弃老策略)

    /**
     * A handler for rejected tasks that discards the oldest unhandled
     * request and then retries {@code execute}, unless the executor
     * is shut down, in which case the task is discarded.
     */
    public static class DiscardOldestPolicy implements RejectedExecutionHandler {
        /**
         * Creates a {@code DiscardOldestPolicy} for the given executor.
         */
        public DiscardOldestPolicy() { }

        /**
         * Obtains and ignores the next task that the executor
         * would otherwise execute, if one is immediately available,
         * and then retries execution of task r, unless the executor
         * is shut down, in which case task r is instead discarded.
         *
         * @param r the runnable task requested to be executed
         * @param e the executor attempting to execute this task
         */
        public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
            if (!e.isShutdown()) {
                e.getQueue().poll();
                e.execute(r);
            }
        }
    }

如果线程池未关闭,就弹出队列头部的元素,然后尝试执行。实际上还是会丢弃任务,如果头部元素执行失败,就丢弃了。区别是优先丢弃的是老的元素。

CallerRunsPolicy(调用者运行策略)

    /**
     * A handler for rejected tasks that runs the rejected task
     * directly in the calling thread of the {@code execute} method,
     * unless the executor has been shut down, in which case the task
     * is discarded.
     */
    public static class CallerRunsPolicy implements RejectedExecutionHandler {
        /**
         * Creates a {@code CallerRunsPolicy}.
         */
        public CallerRunsPolicy() { }

        /**
         * Executes task r in the caller's thread, unless the executor
         * has been shut down, in which case the task is discarded.
         *
         * @param r the runnable task requested to be executed
         * @param e the executor attempting to execute this task
         */
        public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
            if (!e.isShutdown()) {
                r.run();
            }
        }
    }

当触发拒绝策略时,判断线程池有没有关闭,没有关闭就由提交任务的当前线程处理。但是当大量提交后就会阻塞线程,导致性能降低。

hutool中的线程池拒绝策略实现

hutool作为我们经常使用的一个工具类,也有线程池工具,我们不如来看看它是如何实现的。


	/**
	 * 构建ThreadPoolExecutor
	 *
	 * @param builder {@link ExecutorBuilder}
	 * @return {@link ThreadPoolExecutor}
	 */
	private static ThreadPoolExecutor build(ExecutorBuilder builder) {
		final int corePoolSize = builder.corePoolSize;
		final int maxPoolSize = builder.maxPoolSize;
		final long keepAliveTime = builder.keepAliveTime;
		final BlockingQueue<Runnable> workQueue;
		if (null != builder.workQueue) {
			workQueue = builder.workQueue;
		} else {
			// corePoolSize为0则要使用SynchronousQueue避免无限阻塞
			workQueue = (corePoolSize <= 0) ? new SynchronousQueue<>() : new LinkedBlockingQueue<>(DEFAULT_QUEUE_CAPACITY);
		}
		final ThreadFactory threadFactory = (null != builder.threadFactory) ? builder.threadFactory : Executors.defaultThreadFactory();
		RejectedExecutionHandler handler = ObjectUtil.defaultIfNull(builder.handler, new ThreadPoolExecutor.AbortPolicy());

		final ThreadPoolExecutor threadPoolExecutor = new ThreadPoolExecutor(//
				corePoolSize, //
				maxPoolSize, //
				keepAliveTime, TimeUnit.NANOSECONDS, //
				workQueue, //
				threadFactory, //
				handler//
		);
		if (null != builder.allowCoreThreadTimeOut) {
			threadPoolExecutor.allowCoreThreadTimeOut(builder.allowCoreThreadTimeOut);
		}
		return threadPoolExecutor;
	}

可以很清晰的看到,会判断是否传入线程池拒绝策略,如果没有就用默认的AbortPolicy。

dubbo中的拒绝策略

public class AbortPolicyWithReport extends ThreadPoolExecutor.AbortPolicy {

    protected static final Logger logger = LoggerFactory.getLogger(AbortPolicyWithReport.class);

    private final String threadName;

    private final URL url;

    private static volatile long lastPrintTime = 0;

    private static Semaphore guard = new Semaphore(1);

    public AbortPolicyWithReport(String threadName, URL url) {
        this.threadName = threadName;
        this.url = url;
    }

    @Override
    public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
        String msg = String.format("Thread pool is EXHAUSTED!" +
                        " Thread Name: %s, Pool Size: %d (active: %d, core: %d, max: %d, largest: %d), Task: %d (completed: %d)," +
                        " Executor status:(isShutdown:%s, isTerminated:%s, isTerminating:%s), in %s://%s:%d!",
                threadName, e.getPoolSize(), e.getActiveCount(), e.getCorePoolSize(), e.getMaximumPoolSize(), e.getLargestPoolSize(),
                e.getTaskCount(), e.getCompletedTaskCount(), e.isShutdown(), e.isTerminated(), e.isTerminating(),
                url.getProtocol(), url.getIp(), url.getPort());
        logger.warn(msg);
        dumpJStack();
        throw new RejectedExecutionException(msg);
    }

    private void dumpJStack() {
       //省略实现
    }
}

dubbo的策略实现主要就是想让开发人员知道拒绝任务的情况以及原因。它先输出了线程池的详细设置参数,以及线程池当前的状态,还有当前拒绝任务的信息。然后又输出了当前线程堆栈详情在dumpJStack中实现,最后抛出RejectedExecutionException。

Netty中的线程池拒绝策略

    private static final class NewThreadRunsPolicy implements RejectedExecutionHandler {
        NewThreadRunsPolicy() {
            super();
        }

        public void rejectedExecution(Runnable r, ThreadPoolExecutor executor) {
            try {
                final Thread t = new Thread(r, "Temporary task executor");
                t.start();
            } catch (Throwable e) {
                throw new RejectedExecutionException(
                        "Failed to start a new thread", e);
            }
        }
    }

Netty的线程池拒绝策略很像CallerRunsPolicy(调用者运行策略),都是不会直接丢弃任务而是继续处理任务,不同的地方是CallerRunsPolicy(调用者运行策略)是在调用线程继续处理,而Netty是new了一个新线程去处理。

activeMq中的线程池拒绝策略


 new RejectedExecutionHandler() {
                @Override
                public void rejectedExecution(final Runnable r, final ThreadPoolExecutor executor) {
                    try {
                        executor.getQueue().offer(r, 60, TimeUnit.SECONDS);
                    } catch (InterruptedException e) {
                        throw new RejectedExecutionException("Interrupted waiting for BrokerService.worker");
                    }

                    throw new RejectedExecutionException("Timed Out while attempting to enqueue Task.");
                }
            });

activeMq中的策略属于最大努力执行任务型,当触发拒绝策略时,在尝试一分钟的时间重新将任务塞进任务队列,当一分钟超时还没成功时,就抛出异常。

监控线程池

在开发中,我们线程池的运行状态,线程状态,对我们来说都非常重要。所以我们应该把线程池监控起来。 我们可以通过扩展beforeExecute、afterExecute和terminated这三个方法去在执行前或执行后增加一些新的操作。用来记录线程池的情况。

方法含义
shutdown()线程池延迟关闭时(等待线程池里的任务都执行完毕),统计已执行任务、正在执行任务、未执行任务数量
shutdownNow()任务执行之前,记录任务开始时间,startTimes这个HashMap以任务的hashCode为key,开始时间为值
beforeExecute(Thread t, Runnable r)线程池延迟关闭时(等待线程池里的任务都执行完毕),统计已执行任务、正在执行任务、未执行任务数量
afterExecute(Runnable r, Throwable t)任务执行之后,计算任务结束时间。统计任务耗时、初始线程数、核心线程数、正在执行的任务数量、已完成任务数量、任务总数、队列里缓存的任务数量、池中存在的最大线程数、最大允许的线程数、线程空闲时间、线程池是否关闭、线程池是否终止信息

package com.example.threadpool;

import lombok.extern.slf4j.Slf4j;

import java.util.Date;
import java.util.List;
import java.util.Objects;
import java.util.concurrent.*;
import java.util.concurrent.atomic.AtomicInteger;

/**
 * @author kurtl
 */
@Slf4j
public class ThreadPoolMonitor extends ThreadPoolExecutor {


    /**
     * 保存任务开始执行的时间,当任务结束时,用任务结束时间减去开始时间计算任务执行时间
     */
    private final ConcurrentHashMap<String, Date> startTimes;

    /**
     * 线程池名称,一般以业务名称命名,方便区分
     */
    private final String poolName;

    /**
     * 调用父类的构造方法,并初始化HashMap和线程池名称
     *
     * @param corePoolSize    线程池核心线程数
     * @param maximumPoolSize 线程池最大线程数
     * @param keepAliveTime   线程的最大空闲时间
     * @param unit            空闲时间的单位
     * @param workQueue       保存被提交任务的队列
     * @param poolName        线程池名称
     */
    public ThreadPoolMonitor(int corePoolSize, int maximumPoolSize, long keepAliveTime,
                             TimeUnit unit, BlockingQueue<Runnable> workQueue, String poolName) {
        this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
                Executors.defaultThreadFactory(), poolName);
    }


    /**
     * 调用父类的构造方法,并初始化HashMap和线程池名称
     *
     * @param corePoolSize    线程池核心线程数
     * @param maximumPoolSize 线程池最大线程数
     * @param keepAliveTime   线程的最大空闲时间
     * @param unit            空闲时间的单位
     * @param workQueue       保存被提交任务的队列
     * @param threadFactory   线程工厂
     * @param poolName        线程池名称
     */
    public ThreadPoolMonitor(int corePoolSize, int maximumPoolSize, long keepAliveTime,
                             TimeUnit unit, BlockingQueue<Runnable> workQueue,
                             ThreadFactory threadFactory, String poolName) {
        super(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue, threadFactory);
        this.startTimes = new ConcurrentHashMap<>();
        this.poolName = poolName;
    }

    /**
     * 线程池延迟关闭时(等待线程池里的任务都执行完毕),统计线程池情况
     */
    @Override
    public void shutdown() {
        // 统计已执行任务、正在执行任务、未执行任务数量
        log.info("{} Going to shutdown. Executed tasks: {}, Running tasks: {}, Pending tasks: {}",
                this.poolName, this.getCompletedTaskCount(), this.getActiveCount(), this.getQueue().size());
        super.shutdown();
    }

    /**
     * 线程池立即关闭时,统计线程池情况
     */
    @Override
    public List<Runnable> shutdownNow() {
        // 统计已执行任务、正在执行任务、未执行任务数量
        log.info("{} Going to immediately shutdown. Executed tasks: {}, Running tasks: {}, Pending tasks: {}",
                this.poolName, this.getCompletedTaskCount(), this.getActiveCount(), this.getQueue().size());
        return super.shutdownNow();
    }

    /**
     * 任务执行之前,记录任务开始时间
     */
    @Override
    protected void beforeExecute(Thread t, Runnable r) {
        startTimes.put(String.valueOf(r.hashCode()), new Date());
    }

    /**
     * 任务执行之后,计算任务结束时间
     */
    @Override
    protected void afterExecute(Runnable r, Throwable t) {
        Date startDate = startTimes.remove(String.valueOf(r.hashCode()));
        Date finishDate = new Date();
        long diff = finishDate.getTime() - startDate.getTime();
        // 统计任务耗时、初始线程数、核心线程数、正在执行的任务数量、
        // 已完成任务数量、任务总数、队列里缓存的任务数量、池中存在的最大线程数、
        // 最大允许的线程数、线程空闲时间、线程池是否关闭、线程池是否终止
        log.info("{}-pool-monitor: " +
                        "Duration: {} ms, PoolSize: {}, CorePoolSize: {}, Active: {}, " +
                        "Completed: {}, Task: {}, Queue: {}, LargestPoolSize: {}, " +
                        "MaximumPoolSize: {},  KeepAliveTime: {}, isShutdown: {}, isTerminated: {}",
                this.poolName,
                diff, this.getPoolSize(), this.getCorePoolSize(), this.getActiveCount(),
                this.getCompletedTaskCount(), this.getTaskCount(), this.getQueue().size(), this.getLargestPoolSize(),
                this.getMaximumPoolSize(), this.getKeepAliveTime(TimeUnit.MILLISECONDS), this.isShutdown(), this.isTerminated());
    }

    /**
     * 创建固定线程池,代码源于Executors.newFixedThreadPool方法,这里增加了poolName
     *
     * @param nThreads 线程数量
     * @param poolName 线程池名称
     * @return ExecutorService对象
     */
    public static ExecutorService newFixedThreadPool(int nThreads, String poolName) {
        return new ThreadPoolMonitor(nThreads, nThreads, 0L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue<>(), poolName);
    }

    /**
     * 创建缓存型线程池,代码源于Executors.newCachedThreadPool方法,这里增加了poolName
     *
     * @param poolName 线程池名称
     * @return ExecutorService对象
     */
    public static ExecutorService newCachedThreadPool(String poolName) {
        return new ThreadPoolMonitor(0, Integer.MAX_VALUE, 60L, TimeUnit.SECONDS, new SynchronousQueue<>(), poolName);
    }

    /**
     * 生成线程池所用的线程,只是改写了线程池默认的线程工厂,传入线程池名称,便于问题追踪
     */
    static class EventThreadFactory implements ThreadFactory {
        private static final AtomicInteger POOL_NUMBER = new AtomicInteger(1);
        private final ThreadGroup group;
        private final AtomicInteger threadNumber = new AtomicInteger(1);
        private final String namePrefix;

        /**
         * 初始化线程工厂
         *
         * @param poolName 线程池名称
         */
        EventThreadFactory(String poolName) {
            SecurityManager s = System.getSecurityManager();
            group = Objects.nonNull(s) ? s.getThreadGroup() : Thread.currentThread().getThreadGroup();
            namePrefix = poolName + "-pool-" + POOL_NUMBER.getAndIncrement() + "-thread-";
        }

        @Override
        public Thread newThread(Runnable r) {
            Thread t = new Thread(group, r, namePrefix + threadNumber.getAndIncrement(), 0);
            if (t.isDaemon())
                t.setDaemon(false);
            if (t.getPriority() != Thread.NORM_PRIORITY)
                t.setPriority(Thread.NORM_PRIORITY);
            return t;
        }
    }
}


线程池负载关注的核心问题是:基于当前线程池参数分配的资源够不够。对于这个问题,我们可以从事前和事中两个角度来看。事前,线程池定义了“活跃度”这个概念,来让用户在发生Reject异常之前能够感知线程池负载问题,线程池活跃度计算公式为:线程池活跃度 = activeCount/maximumPoolSize。这个公式代表当活跃线程数趋向于maximumPoolSize的时候,代表线程负载趋高。事中,也可以从两方面来看线程池的过载判定条件,一个是发生了Reject异常,一个是队列中有等待任务(支持定制阈值)。以上两种情况发生了都会触发告警,告警信息会通过大象推送给服务所关联的负责人。 ——美团技术文档

核心线程数 最大线程数 如何配置

如何合理的配置线程池参数,比较普遍的说法是。

IO密集型 = 2Ncpu(可以测试后自己控制大小,2Ncpu一般没问题)(常出现于线程中:数据库数据交互、文件上传下载、网络数据传输等等)

计算密集型 = Ncpu(常出现于线程中:复杂算法)

而这种方案没有考虑多线程池的情况,实际使用上也有偏离。

x7.png

图来自美团技术博客

所以参数的设置应该根据自己实际的应用场景定制。

多线程池的使用

一般在实际业务中,我们会定义不同的线程池来处理不同的业务。利用我们之前完成的ThreadPoolMonitor可以很快的定义不同的线程。

ThreadPoolConfig


@EnableAsync
@Configuration
public class ThreadPoolConfig {

    @Bean
    public ThreadPoolExecutor test01(){
        return new ThreadPoolMonitor(16,32,60, TimeUnit.SECONDS, new LinkedBlockingQueue<>(100),"test01");
    }

    @Bean
    public ThreadPoolExecutor test02(){
        return new ThreadPoolMonitor(8,16,60, TimeUnit.SECONDS, new LinkedBlockingQueue<>(100),"test02");
    }
}


TODO

1.动态线程池 2.基于任务的线程池监控

作者水平有限,若有错误遗漏,请指出。

参考文章

1.Java线程池实现原理及其在美团业务中的实践

2.Java并发(六)线程池监控

3.一次Java线程池误用(newFixedThreadPool)引发的线上血案和总结