结合JUC谈Springboot优雅停机

川长思鸟来 2023-09-24 17:11 9阅读 0赞

# 废话少说 #

## 涉及到的知识点 ##

### Spring ###

*  SmartLifecycle
 *  DefaultLifecycleProcessor
 *  WebServerGracefulShutdownLifecycle
 *  WebServerStartStopLifecycle
 *  WebServerManager
 *  TomcatWebServer implements WebServer

### java ###

*  java.util.concurrent.CountDownLatch
 *  java.lang.Runtime
 *  java.lang.ApplicationShutdownHooks
 *  java.lang.Shutdown

## 简述：springboot的优雅停机是借助于ShutdownHook实现的 ##

## 关于spring hook的二三事 ##

*  什么时候设置的hook
 *  什么时候触发的hook
 *  触发hook后续的流程

## 设置hook线程、触发hook ##

### SpringApplication\#run() ###

*  org.springframework.boot.SpringApplication\#refreshContext()
    
     *  org.springframework.context.support.AbstractApplicationContext\#registerShutdownHook()

@Override
    public void registerShutdownHook() {
       if (this.shutdownHook == null) {
          // No shutdown hook registered yet.
          this.shutdownHook = new Thread(SHUTDOWN_HOOK_THREAD_NAME) {
             @Override
             public void run() {
                synchronized (startupShutdownMonitor) {
                   doClose();
                }
             }
          };
          Runtime.getRuntime().addShutdownHook(this.shutdownHook);
       }
    }
    复制代码

*  从上述代码中可以看到，spring在刷新上下文时会向Runtime中注册一个shutdownHook，根据Runtime api中注释可以看出，当虚拟机响应关闭信号后（有些信号不会响应例如 kill -9），会执行这个线程

## 触发hook后续的流程 ##

### 核心入口 ###

*  从注册hook时可以看到，当虚拟机回调时会执行 doClose()方法，也就是说这个方法是关闭容器的核心入口
 *  org.springframework.context.support.AbstractApplicationContext\#doClose()

### 模拟关闭 ###

public static void main(String[] args) {
     ConfigurableApplicationContext context = SpringApplication.run(MvcApplication.class, args);
     // 模拟 shutdown调用
     context.close();
    }
    复制代码

@Override
    public void close() {
       synchronized (this.startupShutdownMonitor) {
          // 此处调用真正的关闭方法
          doClose();
          if (this.shutdownHook != null) {
             try {
                Runtime.getRuntime().removeShutdownHook(this.shutdownHook);
             }
             catch (IllegalStateException ex) {
                // ignore - VM is already shutting down
             }
          }
       }
    }
    
    protected void doClose() {
           ....... 忽略不在本次范围的代码，有兴趣的可以去源码看看
           // Stop all Lifecycle beans, to avoid delays during individual destruction.
          if (this.lifecycleProcessor != null) {
             try {
                // 停止实现Lifecycle的bean
                this.lifecycleProcessor.onClose();
             }
             catch (Throwable ex) {
                logger.warn("Exception thrown from LifecycleProcessor on context close", ex);
             }
          }
        .....
    }
    复制代码

*  上述代码可以忽略不看，只是Springboot停机的外部代码

private void stopBeans() {
       Map<String, Lifecycle> lifecycleBeans = getLifecycleBeans();
       Map<Integer, LifecycleGroup> phases = new HashMap<>();
       lifecycleBeans.forEach((beanName, bean) -> {
          int shutdownPhase = getPhase(bean);
          LifecycleGroup group = phases.get(shutdownPhase);
          if (group == null) {
             group = new LifecycleGroup(shutdownPhase, this.timeoutPerShutdownPhase, lifecycleBeans, false);
             phases.put(shutdownPhase, group);
          }
          group.add(beanName, bean);
       });
       if (!phases.isEmpty()) {
          List<Integer> keys = new ArrayList<>(phases.keySet());
          keys.sort(Collections.reverseOrder());
          for (Integer key : keys) {
             // TODO 重点
             phases.get(key).stop();
          }
       }
    }
    复制代码

*  stopBeans 一共做了两件事 组装 和 排序 这个不重要
 *  重要的是 经过一系列组装，将相同排序的lifecycle加入到同一个 LifecycleGroup 这个类 里面会维护多个 lifecycle成员，在执行stop的时候，多个成员for循环依次执行

// LifecycleGroup
    public void stop() {
       if (this.members.isEmpty()) {
          return;
       }
       if (logger.isDebugEnabled()) {
          logger.debug("Stopping beans in phase " + this.phase);
       }
       this.members.sort(Collections.reverseOrder());
       // 倒数器， count数量就是 lifecycle成员的数量
       CountDownLatch latch = new CountDownLatch(this.smartMemberCount);
       
       Set<String> countDownBeanNames = Collections.synchronizedSet(new LinkedHashSet<>());
       
       // 里面的类名，会在doStop时被移除
       Set<String> lifecycleBeanNames = new HashSet<>(this.lifecycleBeans.keySet());
       
       for (LifecycleGroupMember member : this.members) {
          if (lifecycleBeanNames.contains(member.name)) {
             doStop(this.lifecycleBeans, member.name, latch, countDownBeanNames);
          }
          else if (member.bean instanceof SmartLifecycle) {
             // Already removed: must have been a dependent bean from another phase
             latch.countDown();
          }
       }
       try {
           // await 等待， 也就意味着 如果在上述方法时候，一直不执行countDown ，这里就是一个兜底方案，强制放行
          latch.await(this.timeout, TimeUnit.MILLISECONDS);
          if (latch.getCount() > 0 && !countDownBeanNames.isEmpty() && logger.isInfoEnabled()) {
             logger.info("Failed to shut down " + countDownBeanNames.size() + " bean" +
                   (countDownBeanNames.size() > 1 ? "s" : "") + " with phase value " +
                   this.phase + " within timeout of " + this.timeout + "ms: " + countDownBeanNames);
          }
       }
       catch (InterruptedException ex) {
          Thread.currentThread().interrupt();
       }
    }
    
    复制代码

private void doStop(Map<String, ? extends Lifecycle> lifecycleBeans, final String beanName,
          final CountDownLatch latch, final Set<String> countDownBeanNames) {
        // 移除当前这个bean，并返回bean的实例
       Lifecycle bean = lifecycleBeans.remove(beanName);
       if (bean != null) {
           // 依赖关系 依次stop
          String[] dependentBeans = getBeanFactory().getDependentBeans(beanName);
          for (String dependentBean : dependentBeans) {
             doStop(lifecycleBeans, dependentBean, latch, countDownBeanNames);
          }
          try {
             if (bean.isRunning()) {
                if (bean instanceof SmartLifecycle) {
                   if (logger.isTraceEnabled()) {
                      logger.trace("Asking bean '" + beanName + "' of type [" +
                            bean.getClass().getName() + "] to stop");
                   }
                   countDownBeanNames.add(beanName);
                   // 核心 执行stop，执行完毕后回调函数中 进行countDown
                   ((SmartLifecycle) bean).stop(() -> {
                      latch.countDown();
                      countDownBeanNames.remove(beanName);
                      if (logger.isDebugEnabled()) {
                         logger.debug("Bean '" + beanName + "' completed its stop procedure");
                      }
                   });
                }
                else {
                   if (logger.isTraceEnabled()) {
                      logger.trace("Stopping bean '" + beanName + "' of type [" +
                            bean.getClass().getName() + "]");
                   }
                   bean.stop();
                   if (logger.isDebugEnabled()) {
                      logger.debug("Successfully stopped bean '" + beanName + "'");
                   }
                }
             }
             else if (bean instanceof SmartLifecycle) {
                // Don't wait for beans that aren't running...
                latch.countDown();
             }
          }
          catch (Throwable ex) {
             if (logger.isWarnEnabled()) {
                logger.warn("Failed to stop bean '" + beanName + "'", ex);
             }
          }
       }
    }
    复制代码

*  上述两段代码，其实真正核心的就是一个CountDownLatch的运用
 *  LifecycleGroup的member作为countDown的count，stop成功一个释放一个count，直到全部释放成功
 *  latch.await(this.timeout, TimeUnit.MILLISECONDS)
    
     *  如果countDown内部的count一直没被消费，则一直阻塞在这里
     *  作为一个兜底，如果超过timeout时间还没有stop完毕，则不再阻塞线程，这里的timeout就是咱们在yaml文件中配置的

### SmartLifecycle的回调 ###

default void stop(Runnable callback) {
       stop();
       callback.run();
    }
    复制代码

((SmartLifecycle) bean).stop(() -> {
       latch.countDown();
       countDownBeanNames.remove(beanName);
       if (logger.isDebugEnabled()) {
          logger.debug("Bean '" + beanName + "' completed its stop procedure");
       }
    });
    复制代码

具体看下SmartLifecycle这个方法，我们发现，是一个callback函数，只有当stop完成后，再会执行我们设置的函数，也就是latch.countDown()

### 什么情况下stop迟迟不结束 ###

*  org.springframework.boot.web.reactive.context.WebServerGracefulShutdownLifecycle\#stop(java.lang.Runnable)
    
     *  org.springframework.boot.web.reactive.context.WebServerManager\#shutDownGracefully
        
         *  org.springframework.boot.web.embedded.tomcat.TomcatWebServer\#shutDownGracefully
            
             *  org.springframework.boot.web.embedded.tomcat.GracefulShutdown\#shutDownGracefully

void shutDownGracefully(GracefulShutdownCallback callback) {
       logger.info("Commencing graceful shutdown. Waiting for active requests to complete");
       new Thread(() -> doShutdown(callback), "tomcat-shutdown").start();
    }
    
    private void doShutdown(GracefulShutdownCallback callback) {
       List<Connector> connectors = getConnectors();
       connectors.forEach(this::close);
       try {
          for (Container host : this.tomcat.getEngine().findChildren()) {
             for (Container context : host.findChildren()) {
                while (isActive(context)) {
                   if (this.aborted) {
                      logger.info("Graceful shutdown aborted with one or more requests still active");
                      callback.shutdownComplete(GracefulShutdownResult.REQUESTS_ACTIVE);
                      return;
                   }
                   Thread.sleep(50);
                }
             }
          }
    
       }
       catch (InterruptedException ex) {
          Thread.currentThread().interrupt();
       }
       logger.info("Graceful shutdown complete");
       callback.shutdownComplete(GracefulShutdownResult.IDLE);
    }
    复制代码

*  代码可能有点多，既然坚持到这里了，还是把调用栈详细写出来
 *  shutDownGracefully （callback）
    
     *  我们看到这里启动了一个新的线程，并且执行，全部交给异步执行（不要忘了入参是个 callback）
     *  内部再调用doShutDown(callback）
 *  doShutdown（callback） 关键
    
     *  关闭所有Connector，熟悉tomcat的都知道，Connector是管理socket连接的，关闭了Connector也就代表不再接受新的请求了。
     *  isActive(context) == true就一直执行，进入内部源码看下就会清楚，里面是tomcat正在处理的任务，只要有一个就返回true，这个方法也就是说明了，优雅关闭的核心，当有请求没有处理完，就允许他继续处理

总结

*  定义 countDownLatch 阻塞hook的线程, count数量就是 实现lifecycle的子类
 *  循环每一个lifecycle进行stop,stop完成后会进行countDownLatch.countDown()
 *  最外层countDownLatch.await,设置超时时间,如果超时不再阻塞主进程,正常走完hook流程,结束进程