RocketMQ关闭consumer时抛异常原因分析

红太狼 2024-01-04 06:14 229阅读 0赞

### 问题分析 ###

#### RemotingConnectException ####

##### 问题描述 #####

重启服务时抛出下面的远程连接异常，且是偶发情况，客户端封装的consumer关闭是放在shutdown钩子方法中执行的

2019-05-09 14:47:00,071 WARN [Thread-16] RocketmqRemoting:createChannel:464 {
        } createChannel: connect remote host[brokerHost:10611] failed, DefaultChannelPromise@3721dbe4(failure: java.nio.channels.ClosedChannelException)
    java.nio.channels.ClosedChannelException: null
    2019-05-09 14:47:00,072 ERROR [Thread-16] RocketmqClient:: {
        } unregister client exception from broker: brokerHost:10611
    org.apache.rocketmq.remoting.exception.RemotingConnectException: connect to <brokerHost:10611> failed
            at org.apache.rocketmq.remoting.netty.NettyRemotingClient.invokeSync(NettyRemotingClient.java:357) ~[rocketmq-remoting-4.0.0-incubating.jar!/:4.0.0-incubating]
            at org.apache.rocketmq.client.impl.MQClientAPIImpl.unregisterClient(MQClientAPIImpl.java:864) ~[rocketmq-client-4.0.0-incubating.jar!/:4.0.0-incubating]
            at org.apache.rocketmq.client.impl.factory.MQClientInstance.unregisterClient(MQClientInstance.java:843) ~[rocketmq-client-4.0.0-incubating.jar!/:4.0.0-incubating]
            at org.apache.rocketmq.client.impl.factory.MQClientInstance.unregisterClientWithLock(MQClientInstance.java:817) ~[rocketmq-client-4.0.0-incubating.jar!/:4.0.0-incubating]
            at org.apache.rocketmq.client.impl.factory.MQClientInstance.unregisterConsumer(MQClientInstance.java:810) ~[rocketmq-client-4.0.0-incubating.jar!/:4.0.0-incubating]
            at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.shutdown(DefaultMQPushConsumerImpl.java:525) ~[rocketmq-client-4.0.0-incubating.jar!/:4.0.0-incubating]
            at org.apache.rocketmq.client.consumer.DefaultMQPushConsumer.shutdown(DefaultMQPushConsumer.java:464) ~[rocketmq-client-4.0.0-incubating.jar!/:4.0.0-incubating]
            at com......rocketmq.client.AbstractRocketMQConsumer.destroy(AbstractRocketMQConsumer.java:249) ~[universe-rocketmq-client-1.1.0-SNAPSHOT.jar!/:?]
            at com......rocketmq.client.AbstractRocketMQConsumer$1.run(AbstractRocketMQConsumer.java:177) ~[universe-rocketmq-client-1.1.0-SNAPSHOT.jar!/:?]
            at java.lang.Thread.run(Thread.java:745) [?:1.8.0_77]

1.  跟着堆栈跟踪到代码里销毁过程是调用的DefaultMQPushConsumerImpl的shutdown方法
2.  问题出在了mQClientFactory的unregisterConsumer方法内
3.  mQclientFactory工厂调用unregisterClientWithLock方法关闭连接
4.  注销客户端调用unregisterClient方法

if (addr != null) {
        
        try {
        
            this.mQClientAPIImpl.unregisterClient(addr, this.clientId, producerGroup, consumerGroup, 3000);
            log.info("unregister client[Producer: {} Consumer: {}] from broker[{} {} {}] success", producerGroup, consumerGroup, brokerName, entry1.getKey(), addr);
        } catch (RemotingException | InterruptedException | MQBrokerException e) {
        
            log.error("unregister client exception from broker: " + addr, e);
        }
    }

1.  注销方法调用mQClientAPIImpl客户端API注销客户端unregisterClient，我们所捕获到的异常是connection连接异常，此处的异常抛出没有该类型，继续跟进
2.  remotingClient客户端同步发送注销客户端请求，如果响应失败抛出MQBrokerException异常，继续查看同步调用逻辑
3.  查看NettyRemotingClient同步调用逻辑发现在Channel通道为null或者处于非active状态时会抛出RemotingConnectException连接异常
4.  那么继续看下获取channel的逻辑，发现，如果缓存的channel不存在的话，会直接创建一个新的channel，那么缓存中是否存在呢？可以看到文章开头日志中第一行写明了是在创建channel，说明了此处获取的channel为空（如果客户端已经shutdown的话会清空channelTables缓存），也就有可能此时的客户端已经被调用过了shutdown，下面第三个异常分析也证实该推断

@Override
    public void shutdown() {
        
    ...
      //客户端shutdown时会清空缓存
            this.channelTables.clear();
    ...
    }
    private Channel createChannel(final String addr) throws InterruptedException {
        
        ChannelWrapper cw = this.channelTables.get(addr);
        ...
        if (cw != null) {
        
            ChannelFuture channelFuture = cw.getChannelFuture();
            if (channelFuture.awaitUninterruptibly(this.nettyClientConfig.getConnectTimeoutMillis())) {
        
                if (cw.isOK()) {
        
                    log.info("createChannel: connect remote host[{}] success, {}", addr, channelFuture.toString());
                    return cw.getChannel();
                } else {
        
    				// 连接异常，打印异常原因及堆栈
                    log.warn("createChannel: connect remote host[" + addr + "] failed, " + channelFuture.toString(), channelFuture.cause());
                }
            } else {
        
                log.warn("createChannel: connect remote host[{}] timeout {}ms, {}", addr, this.nettyClientConfig.getConnectTimeoutMillis(),
                    channelFuture.toString());
            }
        }
    	//返回null，后面判断如果为空或者非active状态则抛出连接异常堆栈
        return null;
    }

我们可以看到连接失败，异常原因是ClosedChannelException，发生异常打印日志后返回null，也就是后面判断channel如果为null或者非active状态便抛出连接异常的原因

##### 为什么会连接失败呢？ #####

看异常的堆栈已经将case输出：ClosedChannelException：null，没错与前面的分析一致，已经有其他地方调用了shutdown方法并关闭了channel通道

#### 同类问题导致的异常堆栈 ####

##### RemotingSendRequestException #####

问题与上个异常原因几乎一致，只是发生的位置与时间点发生了一些小变化，上个异常发生在注册时失败，当前异常是发生在发送请求时产生的失败，失败原因也是一致的ClosedChannelException。此处不再赘述

org.apache.rocketmq.remoting.exception.RemotingSendRequestException: send request to <172.24.1.32:10611> failed
    	at org.apache.rocketmq.remoting.netty.NettyRemotingAbstract.invokeSyncImpl(NettyRemotingAbstract.java:294)
    	at org.apache.rocketmq.remoting.netty.NettyRemotingClient.invokeSync(NettyRemotingClient.java:338)
    	at org.apache.rocketmq.client.impl.MQClientAPIImpl.unregisterClient(MQClientAPIImpl.java:864)
    	at org.apache.rocketmq.client.impl.factory.MQClientInstance.unregisterClient(MQClientInstance.java:843)
    	at org.apache.rocketmq.client.impl.factory.MQClientInstance.unregisterClientWithLock(MQClientInstance.java:817)
    	at org.apache.rocketmq.client.impl.factory.MQClientInstance.unregisterConsumer(MQClientInstance.java:810)
    	at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.shutdown(DefaultMQPushConsumerImpl.java:525)
    	at org.apache.rocketmq.client.consumer.DefaultMQPushConsumer.shutdown(DefaultMQPushConsumer.java:464)
    	at com......rocketmq.client.AbstractRocketMQConsumer.destroy(AbstractRocketMQConsumer.java:249)
    	at com......rocketmq.client.AbstractRocketMQConsumer$1.run(AbstractRocketMQConsumer.java:177)
    	at java.lang.Thread.run(Thread.java:745)
    Caused by: java.nio.channels.ClosedChannelException

#### MQClientException ####

##### 问题描述 #####

同样是重启是产生的异常堆栈，但是与上面有些不同的是此次的异常堆栈是有spring bean的销毁方法调用过程中产生的，问题已经很明显指明状态不对导致的，之前的分析也是该抽象类产生，但是发生在shutdown钩子方法内，这样问题就较为明显了，钩子方法与销毁方法中都对同一个consumer执行了shutdown，导致该异常的出现

org.apache.rocketmq.client.exception.MQClientException: The consumer service state not OK, SHUTDOWN_ALREADY
    See http://rocketmq.apache.org/docs/faq/ for further details.
      at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.makeSureStateOK(DefaultMQPushConsumerImpl.java:426)
      at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.persistConsumerOffset(DefaultMQPushConsumerImpl.java:932)
      at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.shutdown(DefaultMQPushConsumerImpl.java:524)
      at org.apache.rocketmq.client.consumer.DefaultMQPushConsumer.shutdown(DefaultMQPushConsumer.java:464)
      at com......rocketmq.client.AbstractRocketMQConsumer.destroy(AbstractRocketMQConsumer.java:249)
      at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
      at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
      at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
      at java.lang.reflect.Method.invoke(Method.java:498)
      at org.springframework.beans.factory.support.DisposableBeanAdapter.invokeCustomDestroyMethod(DisposableBeanAdapter.java:364)
      at org.springframework.beans.factory.support.DisposableBeanAdapter.destroy(DisposableBeanAdapter.java:287)
      at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroyBean(DefaultSingletonBeanRegistry.java:578)
      at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroySingleton(DefaultSingletonBeanRegistry.java:554)
      at org.springframework.beans.factory.support.DefaultListableBeanFactory.destroySingleton(DefaultListableBeanFactory.java:961)
      at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroySingletons(DefaultSingletonBeanRegistry.java:523)
      at org.springframework.beans.factory.support.DefaultListableBeanFactory.destroySingletons(DefaultListableBeanFactory.java:968)
      at org.springframework.context.support.AbstractApplicationContext.destroyBeans(AbstractApplicationContext.java:1033)
      at org.springframework.context.support.AbstractApplicationContext.doClose(AbstractApplicationContext.java:1009)
      at org.springframework.context.support.AbstractApplicationContext$2.run(AbstractApplicationContext.java:928)
    org.apache.rocketmq.client.exception.MQClientException: The consumer service state not OK, SHUTDOWN_ALREADY
    See http://rocketmq.apache.org/docs/faq/ for further details.
      at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.makeSureStateOK(DefaultMQPushConsumerImpl.java:426)
      at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.persistConsumerOffset(DefaultMQPushConsumerImpl.java:932)
      at org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl.shutdown(DefaultMQPushConsumerImpl.java:524)
      at org.apache.rocketmq.client.consumer.DefaultMQPushConsumer.shutdown(DefaultMQPushConsumer.java:464)
      at com......rocketmq.client.AbstractRocketMQConsumer.destroy(AbstractRocketMQConsumer.java:249)
      at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
      at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
      at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
      at java.lang.reflect.Method.invoke(Method.java:498)
      at org.springframework.beans.factory.support.DisposableBeanAdapter.invokeCustomDestroyMethod(DisposableBeanAdapter.java:364)
      at org.springframework.beans.factory.support.DisposableBeanAdapter.destroy(DisposableBeanAdapter.java:287)
      at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroyBean(DefaultSingletonBeanRegistry.java:578)
      at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroySingleton(DefaultSingletonBeanRegistry.java:554)
      at org.springframework.beans.factory.support.DefaultListableBeanFactory.destroySingleton(DefaultListableBeanFactory.java:961)
      at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroySingletons(DefaultSingletonBeanRegistry.java:523)
      at org.springframework.beans.factory.support.DefaultListableBeanFactory.destroySingletons(DefaultListableBeanFactory.java:968)
      at org.springframework.context.support.AbstractApplicationContext.destroyBeans(AbstractApplicationContext.java:1033)
      at org.springframework.context.support.AbstractApplicationContext.doClose(AbstractApplicationContext.java:1009)
      at org.springframework.context.support.AbstractApplicationContext$2.run(AbstractApplicationContext.java:928)

##### 为什么是偶现而不是必现呢？ #####

原因就是由于shutdown中对状态判断是线程不安全的

1.  如果并发调用时读取状态是SHUTDOWN\_ALREADY，则正常返回
2.  如果并发调用时读取状态不是SHUTDOWN\_ALREADY而是RUNNING状态，则就会再次shutdown也就出现了我们所遇到的种种重启时产生的异常，所以只有在并发场景下才能出现该类异常

public void shutdown() {
        
        switch (this.serviceState) {
        
            case CREATE_JUST:
                break;
            case RUNNING:
                this.consumeMessageService.shutdown();
                this.persistConsumerOffset();
                this.mQClientFactory.unregisterConsumer(this.defaultMQPushConsumer.getConsumerGroup());
                this.mQClientFactory.shutdown();
                log.info("the consumer [{}] shutdown OK", this.defaultMQPushConsumer.getConsumerGroup());
                this.rebalanceImpl.destroy();
                this.serviceState = ServiceState.SHUTDOWN_ALREADY;
                break;
            case SHUTDOWN_ALREADY:
                break;
            default:
                break;
        }
    }

### 问题总结 ###

问题原因已经证实，是由于客户端bean的init方法添加shutdown钩子方法，在shutdown服务时调用销毁consumer，同时又在使用客户端的destroy方法(spring回调监听销毁bean时回调bean的自定义销毁方法)中再次调用销毁consumer导致重启时会偶发出现以上列举的异常  
如何解决呢？

1.  最直接的方法便是不要重复并发调用同一个consumer的shutdown方法
2.  当然如果对DefaultMQPushConsumerImpl的shutdown方法能加一个锁保证线程安全，个人认为这种方式对于使用者更加友好些。当然如果存在频繁shutdown的场景会影响性能，不过这是一个权衡取舍的问题