基于SpringBoot多线程@Async的使用体验

目录

• 多线程@Async的使用体验
• 场景
• 1.线程池配置
• 2.子父线程之间共享一个Request的配置方案
• 3.阻塞主线程，等待所有子线程执行完毕后继续执行主线程
• 1.CountDownLatch
• 2.Future
• 4.多线程共用一个事务
• 异步调用@Async问题
• 1.使用背景
• 2.异步处理方式
• 3.@Async不返回数据
• 4.@Async返回数据
• 5.异常处理

多线程@Async的使用体验

场景

导入：可以将大批量的数据insert操作采用多线程的方式并行执行

第三方服务的接口调用：由于存在个别第三方服务调用比较耗时的场景，此时就可以与自身服务的逻辑并行执行

简而言之：接口中部份业务逻辑可以通过并行的方式来优化接口性能

1.线程池配置

@Configuration
@EnableAsync
public class TaskPoolConfig {

    @Bean("taskExecutor") // bean 的名称，默认为首字母小写的方法名
    public Executor taskExecutor() {
        ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();
        //核心线程数（CPU核心数+1）
        executor.setCorePoolSize(10);
        //最大线程数（2*CPU核心数+1）
        executor.setMaxPoolSize(20);
        //缓冲队列数
        executor.setQueueCapacity(200);
        //允许线程空闲时间（单位：默认为秒）
        executor.setKeepAliveSeconds(60);
        //线程池名前缀
        executor.setThreadNamePrefix("sub-thread-");
        // 增加 TaskDecorator 属性的配置
        executor.setTaskDecorator(new ContextDecorator());
        // 线程池对拒绝任务的处理策略:CallerRunsPolicy：不在新线程中执行任务，而是由调用者所在的线程来执行
        executor.setRejectedExecutionHandler(new ThreadPoolExecutor.CallerRunsPolicy());
        executor.initialize();
        return executor;
    }
}

2.子父线程之间共享一个Request的配置方案

1.实现TaskDecorator接口

/**
 * 子线程装饰器
 *
 * @author Da Shuai
 * @date 2021-06-10 18:28:17
 */
public class SubThreadTaskDecorator implements TaskDecorator {

    @Override
    public Runnable decorate(Runnable runnable) {
        RequestAttributes context = RequestContextHolder.currentRequestAttributes();
        return () -> {
            try {
                RequestContextHolder.setRequestAttributes(context);
                runnable.run();
            } finally {
                RequestContextHolder.resetRequestAttributes();
            }
        };
    }
}

2.之前的线程池配置加如下代码使其生效

// 增加 TaskDecorator 属性的配置
executor.setTaskDecorator(new ContextDecorator());

3.阻塞主线程，等待所有子线程执行完毕后继续执行主线程

1.CountDownLatch

思路：

• 实例化CountDownLatch对象，同时传入x（线程数量：这个数量必须等于子线程数量）进行构造
• 每个子线程执行完毕后会调用countDown()方法
• 子线程逻辑后方调用await()方法

这样线程计数器为0之前，主线程就一直处于pending状态

主线程逻辑

new CountDownLatch(X)
latch.await()

@Override
@Transactional
public void importExcel(File file) {
    CountDownLatch latch = new CountDownLatch(3);
    for (int i = 0; i < 3; i++) {
        VoteDO voteDO = new VoteDO();
        voteDO.setTitle(i + "");
        asyncManager.asyncSaveVote(voteDO);
    }
    //System.out.println(1/0);
    try {
        latch.await();
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
}

子线程逻辑

latch.countDown()

@Override
@Async
public void asyncSaveVote(VoteDO voteDO, CountDownLatch latch) {
    log.info("当前线程为 {}，休眠10s开始", Thread.currentThread().getName());
    try {
        Thread.sleep(10000L);
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
    log.info("当前线程为 {}，休眠10s结束", Thread.currentThread().getName());
    log.info("当前线程为 {}，保存开始", Thread.currentThread().getName());
    voteDO.setDesc(Thread.currentThread().getName());
    voteDao.insert(voteDO);
    latch.countDown();
    log.info("当前线程为 {}，保存结束", Thread.currentThread().getName());
}

日志

2021-06-11 16:31:08.653 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : ===============请求内容===============
2021-06-11 16:31:08.653 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : 请求地址:http://localhost:8080/api/import
2021-06-11 16:31:08.653 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : 请求方式:POST
2021-06-11 16:31:08.655 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : 请求类方法:com.zhdj.controller.ImportController.importExcel
2021-06-11 16:31:08.655 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : 请求类方法参数:[org.springframework.web.multipart.support.StandardMultipartHttpServletRequest$StandardMultipartFile@42c3f403]
2021-06-11 16:31:08.655 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : ===============请求内容===============
2021-06-11 16:31:08.676 INFO 27516 --- [nio-8080-exec-1] com.zaxxer.hikari.HikariDataSource : HikariPool-1 - Starting...
2021-06-11 16:31:08.894 INFO 27516 --- [nio-8080-exec-1] com.zaxxer.hikari.HikariDataSource : HikariPool-1 - Start completed.
2021-06-11 16:31:08.921 INFO 27516 --- [ sub-thread-3] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-3，休眠10s开始
2021-06-11 16:31:08.921 INFO 27516 --- [ sub-thread-1] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-1，休眠10s开始
2021-06-11 16:31:08.921 INFO 27516 --- [ sub-thread-2] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-2，休眠10s开始
2021-06-11 16:31:18.921 INFO 27516 --- [ sub-thread-2] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-2，休眠10s结束
2021-06-11 16:31:18.921 INFO 27516 --- [ sub-thread-3] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-3，休眠10s结束
2021-06-11 16:31:18.921 INFO 27516 --- [ sub-thread-2] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-2，保存开始
2021-06-11 16:31:18.921 INFO 27516 --- [ sub-thread-1] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-1，休眠10s结束
2021-06-11 16:31:18.921 INFO 27516 --- [ sub-thread-3] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-3，保存开始
2021-06-11 16:31:18.921 INFO 27516 --- [ sub-thread-1] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-1，保存开始
2021-06-11 16:31:19.080 DEBUG 27516 --- [ sub-thread-3] com.zhdj.dao.VoteDao.insert : ==> Preparing: INSERT INTO vote ( title, `desc`, gmt_create, gmt_modified ) VALUES ( ?, ?, ?, ? )
2021-06-11 16:31:19.080 DEBUG 27516 --- [ sub-thread-1] com.zhdj.dao.VoteDao.insert : ==> Preparing: INSERT INTO vote ( title, `desc`, gmt_create, gmt_modified ) VALUES ( ?, ?, ?, ? )
2021-06-11 16:31:19.080 DEBUG 27516 --- [ sub-thread-2] com.zhdj.dao.VoteDao.insert : ==> Preparing: INSERT INTO vote ( title, `desc`, gmt_create, gmt_modified ) VALUES ( ?, ?, ?, ? )
2021-06-11 16:31:19.156 DEBUG 27516 --- [ sub-thread-1] com.zhdj.dao.VoteDao.insert : ==> Parameters: 0(String), sub-thread-1(String), 2021-06-11T16:31:19.032(LocalDateTime), 2021-06-11T16:31:19.037(LocalDateTime)
2021-06-11 16:31:19.156 DEBUG 27516 --- [ sub-thread-3] com.zhdj.dao.VoteDao.insert : ==> Parameters: 2(String), sub-thread-3(String), 2021-06-11T16:31:19.032(LocalDateTime), 2021-06-11T16:31:19.037(LocalDateTime)
2021-06-11 16:31:19.156 DEBUG 27516 --- [ sub-thread-2] com.zhdj.dao.VoteDao.insert : ==> Parameters: 1(String), sub-thread-2(String), 2021-06-11T16:31:19.032(LocalDateTime), 2021-06-11T16:31:19.037(LocalDateTime)
2021-06-11 16:31:19.172 DEBUG 27516 --- [ sub-thread-3] com.zhdj.dao.VoteDao.insert : <== Updates: 1
2021-06-11 16:31:19.178 DEBUG 27516 --- [ sub-thread-2] com.zhdj.dao.VoteDao.insert : <== Updates: 1
2021-06-11 16:31:19.187 DEBUG 27516 --- [ sub-thread-1] com.zhdj.dao.VoteDao.insert : <== Updates: 1
2021-06-11 16:31:19.224 INFO 27516 --- [ sub-thread-3] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-3，保存结束
2021-06-11 16:31:19.224 INFO 27516 --- [ sub-thread-1] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-1，保存结束
2021-06-11 16:31:19.224 INFO 27516 --- [ sub-thread-2] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-2，保存结束
2021-06-11 16:31:19.226 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : --------------返回内容----------------
2021-06-11 16:31:19.328 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : Response内容:null
2021-06-11 16:31:19.328 INFO 27516 --- [nio-8080-exec-1] com.zhdj.config.LogAspect : --------------返回内容----------------

2.Future

思路：

1.子线程逻辑返回Future对象

2.主线程逻辑循环判断每个子线程返回的Future对象isDone()是否为true

主线程逻辑

循环判断future.isDone()是否为true

@Override
@Transactional
public void importExcel(File file) {
 List<Future> futureList = new ArrayList<>();
 for (int i = 0; i < 3; i++) {
     VoteDO voteDO = new VoteDO();
     voteDO.setTitle(i + "");
     Future future = asyncManager.asyncSaveVote(voteDO);
     futureList.add(future);
 }
 //检查所有子线程是否均执行完毕
 while (true) {
     boolean isAllDone = true;
     for (Future future : futureList) {
         if (null == future || !future.isDone()) {
             isAllDone = false;
         }
     }
     if (isAllDone) {
         log.info("所有子线程执行完毕");
         break;
     }
 }
}

子线程逻辑

返回Future对象

@Override
public Future asyncSaveVote(VoteDO voteDO) {
    log.info("当前线程为 {}，休眠10s开始", Thread.currentThread().getName());
    try {
        Thread.sleep(10000L);
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
    log.info("当前线程为 {}，休眠10s结束", Thread.currentThread().getName());
    log.info("当前线程为 {}，保存开始", Thread.currentThread().getName());
    voteDO.setDesc(Thread.currentThread().getName());
    voteDao.insert(voteDO);
    log.info("当前线程为 {}，保存结束", Thread.currentThread().getName());
    //返回需要用AsyncResult类
    return new AsyncResult<>(true);
}

日志

2021-06-11 16:42:28.974 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : ===============请求内容===============
2021-06-11 16:42:28.974 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : 请求地址:http://localhost:8080/api/import
2021-06-11 16:42:28.974 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : 请求方式:POST
2021-06-11 16:42:28.975 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : 请求类方法:com.zhdj.controller.ImportController.importExcel
2021-06-11 16:42:28.975 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : 请求类方法参数:[org.springframework.web.multipart.support.StandardMultipartHttpServletRequest$StandardMultipartFile@7e23bacc]
2021-06-11 16:42:28.975 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : ===============请求内容===============
2021-06-11 16:42:28.979 INFO 20492 --- [ sub-thread-5] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-5，休眠10s开始
2021-06-11 16:42:28.979 INFO 20492 --- [ sub-thread-4] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-4，休眠10s开始
2021-06-11 16:42:28.979 INFO 20492 --- [ sub-thread-6] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-6，休眠10s开始
2021-06-11 16:42:38.980 INFO 20492 --- [ sub-thread-6] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-6，休眠10s结束
2021-06-11 16:42:38.980 INFO 20492 --- [ sub-thread-4] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-4，休眠10s结束
2021-06-11 16:42:38.980 INFO 20492 --- [ sub-thread-5] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-5，休眠10s结束
2021-06-11 16:42:38.980 INFO 20492 --- [ sub-thread-6] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-6，保存开始
2021-06-11 16:42:38.980 INFO 20492 --- [ sub-thread-5] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-5，保存开始
2021-06-11 16:42:38.980 INFO 20492 --- [ sub-thread-4] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-4，保存开始
2021-06-11 16:42:38.981 DEBUG 20492 --- [ sub-thread-4] com.zhdj.dao.VoteDao.insert : ==> Preparing: INSERT INTO vote ( title, `desc`, gmt_create, gmt_modified ) VALUES ( ?, ?, ?, ? )
2021-06-11 16:42:38.981 DEBUG 20492 --- [ sub-thread-5] com.zhdj.dao.VoteDao.insert : ==> Preparing: INSERT INTO vote ( title, `desc`, gmt_create, gmt_modified ) VALUES ( ?, ?, ?, ? )
2021-06-11 16:42:38.981 DEBUG 20492 --- [ sub-thread-6] com.zhdj.dao.VoteDao.insert : ==> Preparing: INSERT INTO vote ( title, `desc`, gmt_create, gmt_modified ) VALUES ( ?, ?, ?, ? )
2021-06-11 16:42:38.982 DEBUG 20492 --- [ sub-thread-5] com.zhdj.dao.VoteDao.insert : ==> Parameters: 1(String), sub-thread-5(String), 2021-06-11T16:42:38.980(LocalDateTime), 2021-06-11T16:42:38.981(LocalDateTime)
2021-06-11 16:42:38.982 DEBUG 20492 --- [ sub-thread-4] com.zhdj.dao.VoteDao.insert : ==> Parameters: 0(String), sub-thread-4(String), 2021-06-11T16:42:38.980(LocalDateTime), 2021-06-11T16:42:38.981(LocalDateTime)
2021-06-11 16:42:38.982 DEBUG 20492 --- [ sub-thread-6] com.zhdj.dao.VoteDao.insert : ==> Parameters: 2(String), sub-thread-6(String), 2021-06-11T16:42:38.980(LocalDateTime), 2021-06-11T16:42:38.981(LocalDateTime)
2021-06-11 16:42:38.988 DEBUG 20492 --- [ sub-thread-5] com.zhdj.dao.VoteDao.insert : <== Updates: 1
2021-06-11 16:42:38.989 INFO 20492 --- [ sub-thread-5] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-5，保存结束
2021-06-11 16:42:38.993 DEBUG 20492 --- [ sub-thread-6] com.zhdj.dao.VoteDao.insert : <== Updates: 1
2021-06-11 16:42:38.993 INFO 20492 --- [ sub-thread-6] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-6，保存结束
2021-06-11 16:42:39.004 DEBUG 20492 --- [ sub-thread-4] com.zhdj.dao.VoteDao.insert : <== Updates: 1
2021-06-11 16:42:39.005 INFO 20492 --- [ sub-thread-4] com.zhdj.AsyncManagerImpl : 当前线程为 sub-thread-4，保存结束
2021-06-11 16:42:39.005 INFO 20492 --- [nio-8080-exec-2] com.zhdj.service.impl.VoteServiceImpl : 所有子线程执行完毕
2021-06-11 16:42:39.005 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : --------------返回内容----------------
2021-06-11 16:42:39.005 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : Response内容:null
2021-06-11 16:42:39.005 INFO 20492 --- [nio-8080-exec-2] com.zhdj.config.LogAspect : --------------返回内容----------------

4.多线程共用一个事务

暂时无解决方案，这是弊端

异步调用@Async问题

1.使用背景

在项目中，当访问其他人的接口较慢或者做耗时任务时，不想程序一直卡在耗时任务上，想程序能够并行执行，我们可以使用多线程来并行的处理任务，也可以使用spring提供的异步处理方式@Async。

2.异步处理方式

调用之后，不返回任何数据。

调用之后，返回数据，通过Future来获取返回数据

3.@Async不返回数据

使用@EnableAsync启用异步注解

@Configuration
@EnableAsync
@Slf4j
public class AsyncConfig{
}

在异步处理的方法dealNoReturnTask上添加注解@Async

@Component
@Slf4j
public class AsyncTask {

    @Async
    public void dealNoReturnTask(){
        log.info("Thread {} deal No Return Task start", Thread.currentThread().getName());
        try {
            Thread.sleep(3000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        log.info("Thread {} deal No Return Task end at {}", Thread.currentThread().getName(), System.currentTimeMillis());
    }
}

Test测试类：

@SpringBootTest(classes = SpringbootApplication.class)
@RunWith(SpringJUnit4ClassRunner.class)
@Slf4j
public class AsyncTest {
    @Autowired
    private AsyncTask asyncTask;

    @Test
    public void testDealNoReturnTask(){
        asyncTask.dealNoReturnTask();
        try {
            log.info("begin to deal other Task!");
            Thread.sleep(10000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
    }

日志打印结果为：

begin to deal other Task!
AsyncExecutorThread-1 deal No Return Task start
AsyncExecutorThread-1 deal No Return Task end at 1499751227034

从日志中我们可以看出，方法dealNoReturnTask()是异步执行完成的。

dealNoReturnTask()设置sleep 3s是为了模拟耗时任务

testDealNoReturnTask()设置sleep 10s是为了确认异步是否执行完成

4.@Async返回数据

异步调用返回数据，Future表示在未来某个点获取执行结果，返回数据类型可以自定义

    @Async
    public Future<String> dealHaveReturnTask() {
        try {
            Thread.sleep(3000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        JSONObject jsonObject = new JSONObject();
        jsonObject.put("thread", Thread.currentThread().getName());
        jsonObject.put("time", System.currentTimeMillis());
        return new AsyncResult<String>(jsonObject.toJSONString());
    }

测试类用isCancelled判断异步任务是否取消，isDone判断任务是否执行结束

    @Test
    public void testDealHaveReturnTask() throws Exception {

        Future<String> future = asyncTask.dealHaveReturnTask();
        log.info("begin to deal other Task!");
        while (true) {
            if(future.isCancelled()){
                log.info("deal async task is Cancelled");
                break;
            }
            if (future.isDone() ) {
                log.info("deal async task is Done");
                log.info("return result is " + future.get());
                break;
            }
            log.info("wait async task to end ...");
            Thread.sleep(1000);
        }
    }

日志打印如下，我们可以看出任务一直在等待异步任务执行完毕，用future.get()来获取异步任务的返回结果

begin to deal other Task!
wait async task to end ...
wait async task to end ...
wait async task to end ...
wait async task to end ...
deal async task is Done
return result is {"thread":"AsyncExecutorThread-1","time":1499752617330}

5.异常处理

我们可以实现AsyncConfigurer接口，也可以继承AsyncConfigurerSupport类来实现

在方法getAsyncExecutor()中创建线程池的时候，必须使用 executor.initialize()，

不然在调用时会报线程池未初始化的异常。

如果使用threadPoolTaskExecutor()来定义bean，则不需要初始化

@Configuration
@EnableAsync
@Slf4j
public class AsyncConfig implements AsyncConfigurer {

//    @Bean
//    public ThreadPoolTaskExecutor threadPoolTaskExecutor(){
//        ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();
//        executor.setCorePoolSize(10);
//        executor.setMaxPoolSize(100);
//        executor.setQueueCapacity(100);
//        return executor;
//    }

    @Override
    public Executor getAsyncExecutor() {
        ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();
        executor.setCorePoolSize(10);
        executor.setMaxPoolSize(100);
        executor.setQueueCapacity(100);
        executor.setThreadNamePrefix("AsyncExecutorThread-");
        executor.initialize(); //如果不初始化，导致找到不到执行器
        return executor;
    }
    @Override
    public AsyncUncaughtExceptionHandler getAsyncUncaughtExceptionHandler() {
        return new AsyncExceptionHandler();
    }
}

异步异常处理类：

@Slf4j
public class AsyncExceptionHandler implements AsyncUncaughtExceptionHandler {
    @Override
    public void handleUncaughtException(Throwable ex, Method method, Object... params) {
        log.info("Async method: {} has uncaught exception,params:{}", method.getName(), JSON.toJSONString(params));

        if (ex instanceof AsyncException) {
            AsyncException asyncException = (AsyncException) ex;
            log.info("asyncException:{}",asyncException.getErrorMessage());
        }

        log.info("Exception :");
        ex.printStackTrace();
    }
}

异步处理异常类：

@Data
@AllArgsConstructor
public class AsyncException extends Exception {
    private int code;
    private String errorMessage;
}

在无返回值的异步调用中，异步处理抛出异常，AsyncExceptionHandler的handleUncaughtException()会捕获指定异常，原有任务还会继续运行，直到结束。

在有返回值的异步调用中，异步处理抛出异常，会直接抛出异常，异步任务结束，原有处理结束执行。

以上为个人经验，希望能给大家一个参考，也希望大家多多支持我们。