RocketMQ源码 — 三、 Producer消息发送过程

Producer

消息发送

producer start

producer启动过程如下图

public void start(final boolean startFactory) throws MQClientException {
    switch (this.serviceState) {
        case CREATE_JUST:
            this.serviceState = ServiceState.START_FAILED;

            // check  GroupName
            this.checkConfig();
            // 改变ClientConfig.instanceName为pid
            if (!this.defaultMQProducer.getProducerGroup().equals(MixAll.CLIENT_INNER_PRODUCER_GROUP)) {
                this.defaultMQProducer.changeInstanceNameToPID();
            }

            // 初始化mQClientFactory为MQClientInstance，并将该实例加入factoryTable
            this.mQClientFactory = MQClientManager.getInstance().getAndCreateMQClientInstance(this.defaultMQProducer, rpcHook);

            // 将producer注册到MQClientInstance.producerTbale
            boolean registerOK = mQClientFactory.registerProducer(this.defaultMQProducer.getProducerGroup(), this);
            if (!registerOK) {
                this.serviceState = ServiceState.CREATE_JUST;
                throw new MQClientException("The producer group[" + this.defaultMQProducer.getProducerGroup()
                        + "] has been created before, specify another name please." + FAQUrl.suggestTodo(FAQUrl.GROUP_NAME_DUPLICATE_URL),
                        null);
            }

            // 保存topic对应的routeInfo
            this.topicPublishInfoTable.put(this.defaultMQProducer.getCreateTopicKey(), new TopicPublishInfo());

            if (startFactory) {
                // 启动MQClientInstance
                mQClientFactory.start();
            }

            log.info("the producer [{}] start OK. sendMessageWithVIPChannel={}", this.defaultMQProducer.getProducerGroup(),
                    this.defaultMQProducer.isSendMessageWithVIPChannel());
            this.serviceState = ServiceState.RUNNING;
            break;
        case RUNNING:
        case START_FAILED:
        case SHUTDOWN_ALREADY:
            throw new MQClientException("The producer service state not OK, maybe started once, "//
                    + this.serviceState//
                    + FAQUrl.suggestTodo(FAQUrl.CLIENT_SERVICE_NOT_OK),
                    null);
        default:
            break;
    }

    // 启动的时候向所有的broker发送heartbeat
    this.mQClientFactory.sendHeartbeatToAllBrokerWithLock();
}

消息发送流程

消息发送过程：先由producer封装通过netty发送到broker，然后由broker进行保存，过程如下

============================= producer发送消息 =============================

主要逻辑在DefaultMQProducerImpl.sendDefaultImpl

1. 获取topicRouteInfo
   
   DefaultMQProducerImpl.tryToFindTopicPublishInfo
   
   ->
   
   MQClientInstance.updateTopicRouteInfoFromNameServer

private final ConcurrentHashMap<String/* group */, MQProducerInner> producerTable：放置所有的producer
updateTopicRouteInfoFromNameServer
- 获取topicRouteInfo
- 然后遍历producerTable，更新所有producer的topicRouteInfo
- 遍历consumerTable，更新所有的consumer的topicSubscribeInfo
- 加入topicRouteTable

->

MQClientAPIImpl.getTopicRouteInfoFromNameServer

通过netty向nameServer发出请求，请求topicRouteInfo

2. 选取一个messageQueue

从messageQueueList取一个MessageQueue

3. netty发送消息

DefaultMQProducer.sendKernelImpl，

============================= broker接收消息 =============================

因为使用netty作为网络通信工具，broker也是先使用netty接收到信息，然后调用注册的processor处理

1. parseRequestHeader
   
   利用反射构造requestHeader
2. 构造SendMessageContext
3. 执行beforeHook
4. 发送消息
   • 构造MessageExtBrokerInner
   • 保存消息DefaultMessageStore.putMessage
     • CommitLog.putMessage
       • 从mapedFileQueue中取出一个mapedFile,appendMessag使用directBuffer的方式写入commitLog
       • 同步或者异步刷盘
       • 同步双写
5. 执行afterHook

issue

broker怎么会使用SendMessageProcessor来处理producer发来的消息？

// 在broker初始化的时候会注册所有的processor，registerProcessor
SendMessageProcessor sendProcessor = new SendMessageProcessor(this);
sendProcessor.registerSendMessageHook(sendMessageHookList)
sendProcessor.registerConsumeMessageHook(consumeMessageHookList);

this.remotingServer.registerProcessor(RequestCode.SEND_MESSAGE, sendProcessor, this.sendMessageExecutor);
this.remotingServer.registerProcessor(RequestCode.SEND_MESSAGE_V2, sendProcessor, this.sendMessageExecutor);
this.remotingServer.registerProcessor(RequestCode.CONSUMER_SEND_MSG_BACK, sendProcessor, this.sendMessageExecutor);
this.fastRemotingServer.registerProcessor(RequestCode.SEND_MESSAGE, sendProcessor, this.sendMessageExecutor);
this.fastRemotingServer.registerProcessor(RequestCode.SEND_MESSAGE_V2, sendProcessor, this.sendMessageExecutor);
this.fastRemotingServer.registerProcessor(RequestCode.CONSUMER_SEND_MSG_BACK, sendProcessor, this.sendMessageExecutor);

broker怎么接收消息并写入文件

1. 找到需要写入的mapedFile（最后一个，或者新建一个）
2. 使用mapedFile.appendMessage

怎么获取需要写入的mapedFile

见时序图：RocketMQ.asta（Broker收到sendRequest）

mapedFile.appendMessage的过程

见时序图：RocketMQ.asta（Broker收到sendRequest）

消息发送过程中涉及到的类

DefaultMQProducerImpl

topicPublishInfoTable

里面存放topic对应的messageQueue等信息

// topic对应的消息是否有序
private boolean orderTopic = false;
// 有没有routerInfo
private boolean haveTopicRouterInfo = false;
// topic对应的messageQueue
private List<MessageQueue> messageQueueList = new ArrayList<MessageQueue>();
// 消息发往哪一个queue
private volatile ThreadLocalIndex sendWhichQueue = new ThreadLocalIndex(0);
private TopicRouteData topicRouteData;

private final ConcurrentHashMap<String/* topic */, TopicPublishInfo> topicPublishInfoTable = new ConcurrentHashMap<String, TopicPublishInfo>();

put:

1. 在producer start的时候会根据当前producer 的topic新建一个TopicPushlishInfo放进去
2. 在发送信息之前会获取topic对应的topicPublishInfo，这个时候会去nameServer 查询最新的信息并更新table中所有的记录

defaultMQProducer

private final DefaultMQProducer defaultMQProducer;

发送消息的类，在DefaultMQProducer的构造函数里new DefaultMQProducerImpl传入DefaultMQProducer自身，所以这里DefaultMQProducerImpl.defaultMQProducer默认就是DefaultMQProducer

mQClientFactory

private MQClientInstance mQClientFactory;

MQClientManager是单例，包含两个属性

// 用来生成MQClientInstance的id，每个递增
private AtomicInteger factoryIndexGenerator = new AtomicInteger();
// 包含clientId到MQClientInstance的映射，管理client
private ConcurrentHashMap<String/* clientId */, MQClientInstance> factoryTable =
            new ConcurrentHashMap<String, MQClientInstance>();

通过getInstance获取实例之后调用getAndCreateMQClientInstance创建MQClientInstance。

在DefaultMQProducerImpl.start的时候调用mQClientFactory.start启动

MQClientInstance

字段

// 在MQClientManager中new MQCLientInstance的时候，传入DefaultMQProducer，因为DefaultMQProducer继承了ClientConfig
private final ClientConfig clientConfig;
// new 的时候传入，由MQClientManager生成
private final int instanceIndex;
// new的时候传入，由clientConfig.buildMQClientId生成，形式为：ip@instanceName
private final String clientId;

// 每个group对应的MQProducerInner
// 在producer启动的时候注册到这儿
// 在每次发送message的时候从nameServer获取topicRouteData并更新每个producer对应的信息
private final ConcurrentHashMap<String/* group */, MQProducerInner> producerTable = new ConcurrentHashMap<String, MQProducerInner>();
// 每个group对应的MQConsumerInner
// consumer start的时候注册到这儿
// 在每次发送message的时候从nameServer获取topicRouteData并更新每个consumer对应的信息
private final ConcurrentHashMap<String/* group */, MQConsumerInner> consumerTable = new ConcurrentHashMap<String, MQConsumerInner>();
// 每个group对应的adminExtInner，在NameServer 启动的时候会注册DefaultMQAdminExt
private final ConcurrentHashMap<String/* group */, MQAdminExtInner> adminExtTable = new ConcurrentHashMap<String, MQAdminExtInner>();

// topic对应的TopicRouteTable
// 在每次发送message的时候从nameServer获取topicRouteData并更新对应的信息
private final ConcurrentHashMap<String/* Topic */, TopicRouteData> topicRouteTable = new ConcurrentHashMap<String, TopicRouteData>();

// 所有broker的地址
// 在每次发送message的时候从nameServer获取topicRouteData并更新对应的信息
private final ConcurrentHashMap<String/* Broker Name */, HashMap<Long/* brokerId */, String/* address */>> brokerAddrTable =
            new ConcurrentHashMap<String, HashMap<Long, String>>();

// 定时任务线程池，包括：fetchNameServerAddr，updateTopicRouteInfoFromNameServer等
private final ScheduledExecutorService scheduledExecutorService = Executors.newSingleThreadScheduledExecutor(new ThreadFactory() {
    @Override
    public Thread newThread(Runnable r) {
        return new Thread(r, "MQClientFactoryScheduledThread");
    }
});

// service
private final PullMessageService pullMessageService;
private final RebalanceService rebalanceService;

private final NettyClientConfig nettyClientConfig;
// 用来client远程通信，使用netty
private final MQClientAPIImpl mQClientAPIImpl;

关键方法

public void start() throws MQClientException {
    synchronized (this) {
        switch (this.serviceState) {
            case CREATE_JUST:
                this.serviceState = ServiceState.START_FAILED;
                // If not specified,looking address from name server
                if (null == this.clientConfig.getNamesrvAddr()) {
                    this.clientConfig.setNamesrvAddr(this.mQClientAPIImpl.fetchNameServerAddr());
                }
                // Start request-response channel
                this.mQClientAPIImpl.start();
                // Start various schedule tasks
                this.startScheduledTask();
                // Start pull service
                this.pullMessageService.start();
                // Start rebalance service
                this.rebalanceService.start();
                // Start push service
                this.defaultMQProducer.getDefaultMQProducerImpl().start(false);
                log.info("the client factory [{}] start OK", this.clientId);
                this.serviceState = ServiceState.RUNNING;
                break;
            case RUNNING:
                break;
            case SHUTDOWN_ALREADY:
                break;
            case START_FAILED:
                throw new MQClientException("The Factory object[" + this.getClientId() + "] has been created before, and failed.", null);
            default:
                break;
        }
    }
}

private void startScheduledTask() {
    if (null == this.clientConfig.getNamesrvAddr()) {
        this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {

            @Override
            public void run() {
                try {
                    // 获取nameServer的地址
                    MQClientInstance.this.mQClientAPIImpl.fetchNameServerAddr();
                } catch (Exception e) {
                    log.error("ScheduledTask fetchNameServerAddr exception", e);
                }
            }
        }, 1000 * 10, 1000 * 60 * 2, TimeUnit.MILLISECONDS);
    }

    this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {

        @Override
        public void run() {
            try {
                // 更新topicRouteData
                MQClientInstance.this.updateTopicRouteInfoFromNameServer();
            } catch (Exception e) {
                log.error("ScheduledTask updateTopicRouteInfoFromNameServer exception", e);
            }
        }
    }, 10, this.clientConfig.getPollNameServerInteval(), TimeUnit.MILLISECONDS);

    this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {

        @Override
        public void run() {
            try {
                // 清理掉线的broker
                MQClientInstance.this.cleanOfflineBroker();
                // 给broker发送心跳
                MQClientInstance.this.sendHeartbeatToAllBrokerWithLock();
            } catch (Exception e) {
                log.error("ScheduledTask sendHeartbeatToAllBroker exception", e);
            }
        }
    }, 1000, this.clientConfig.getHeartbeatBrokerInterval(), TimeUnit.MILLISECONDS);

    this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {

        @Override
        public void run() {
            try {
                // 保存consumerOffset
                MQClientInstance.this.persistAllConsumerOffset();
            } catch (Exception e) {
                log.error("ScheduledTask persistAllConsumerOffset exception", e);
            }
        }
    }, 1000 * 10, this.clientConfig.getPersistConsumerOffsetInterval(), TimeUnit.MILLISECONDS);

    this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {

        @Override
        public void run() {
            try {
                // 根据processQueueTable的大小决定是否需要增加或者减少threadPool的大小
                // 目前尚未实现具体的增加或者减少的逻辑
                MQClientInstance.this.adjustThreadPool();
            } catch (Exception e) {
                log.error("ScheduledTask adjustThreadPool exception", e);
            }
        }
    }, 1, 1, TimeUnit.MINUTES);
}

DefaultMessageStore

字段

private final MessageStoreConfig messageStoreConfig;
// CommitLog
private final CommitLog commitLog;

// topic和queueId唯一确定一个consumeQueue
// put：
// 1. 在getMessage的时候会调用findConsumeQueue，如果consumeQueueTable不存在对应的（topic，queueId），则新建一个加入table
// 2. 在启动的时候load
private final ConcurrentHashMap<String/* topic */, ConcurrentHashMap<Integer/* queueId */, ConsumeQueue>> consumeQueueTable;

/** service */
// 消息刷盘，依次循环consumeQueueTable，将每个consumeQueue中的mapedFileQueue commit
private final FlushConsumeQueueService flushConsumeQueueService;
// 删除过时的commitlog文件
private final CleanCommitLogService cleanCommitLogService;
// 删除consumeQueue文件
private final CleanConsumeQueueService cleanConsumeQueueService;
// 在新建DefaultMessageStore的时候新建new
// 在messageStore start的时候start
// 在run方法中循环取出requestQueue（PriorityBlokingQueue，take的时候是阻塞的）里面的request处理——即新建mapedFile
private final AllocateMapedFileService allocateMapedFileService;
// TODO 暂时还不能理解indexservice
private final IndexService indexService;
// TODO
private final ReputMessageService reputMessageService;
// HA 高可用，同步双写，异步复制
private final HAService haService;

private final ScheduleMessageService scheduleMessageService;

private final StoreStatsService storeStatsService;

关键方法

public void start() throws Exception {
    this.flushConsumeQueueService.start();
    this.commitLog.start();
    this.storeStatsService.start();

    if (this.scheduleMessageService != null && SLAVE != messageStoreConfig.getBrokerRole()) {
        this.scheduleMessageService.start();
    }

    if (this.getMessageStoreConfig().isDuplicationEnable()) {
        this.reputMessageService.setReputFromOffset(this.commitLog.getConfirmOffset());
    } else {
        this.reputMessageService.setReputFromOffset(this.commitLog.getMaxOffset());
    }
    this.reputMessageService.start();

    this.haService.start();

    this.createTempFile();
    this.addScheduleTask();
    this.shutdown = false;
}

CommitLog

消息写入内存，保存文件的地方

字段

// 消息文件队列，包含所有保存在磁盘上的文件
private final MapedFileQueue mapedFileQueue;
//
private final DefaultMessageStore defaultMessageStore;
// 消息刷盘
private final FlushCommitLogService flushCommitLogService;
// 添加消息的回调，在doAppend方法中追加消息到内存
private final AppendMessageCallback appendMessageCallback;
// 记录topic对应每个队列的offset
private HashMap<String/* topic-queueid */, Long/* offset */> topicQueueTable = new HashMap<String, Long>(1024);

MapedFileQueue

管理mapedFile，新建、获取、删除mapedFile将消息写入文件

字段

// 每次删除文件个数，作为是否删除的一个参数
private static final int DeleteFilesBatchMax = 10;
// 文件的存储路径
private final String storePath;
// 每个文件的大小
private final int mapedFileSize;
// 所有文件列表
private final List<MapedFile> mapedFiles = new ArrayList<MapedFile>();
// 读锁
private final ReadWriteLock readWriteLock = new ReentrantReadWriteLock();
// 新建mapedFile的地方，将新建mapedFile的 request添加到requesttable和requestQueue中
// 在MessageStore启动的时候会启动AllocateMapedFileService这个线程，执行requestQueue里面的request，新建mapedFile
private final AllocateMapedFileService allocateMapedFileService;

MapedFile

和文件一对一关系

字段

// 保存消息的文件名
private final String fileName;
// 该文件的全局offset，也就是文件名的前缀
private final long fileFromOffset;
// 文件大小
private final int fileSize;
// 文件对象
private final File file;
// 文件映射为的内存
private final MappedByteBuffer mappedByteBuffer;
// 文件写的位置
private final AtomicInteger wrotePostion = new AtomicInteger(0);
// 刷盘之后的位置
private final AtomicInteger committedPosition = new AtomicInteger(0);
// nio阻塞
private FileChannel fileChannel;

private volatile long storeTimestamp = 0;
private boolean firstCreateInQueue = false;

方法

public int commit(final int flushLeastPages) {
    // 判断是否需要flush
    if (this.isAbleToFlush(flushLeastPages)) {
        //判断文件是否被占用，也就是说每次commit的时候不一定成功
        if (this.hold()) {
            // 如果没有被占用则将内存刷到磁盘上
            int value = this.wrotePostion.get();
            // 将内存强制写入磁盘
            this.mappedByteBuffer.force();
            this.committedPosition.set(value);
            // 释放占用
            this.release();
        } else {
            // 尝试占用失败，重置committedPosition
            log.warn("in commit, hold failed, commit offset = " + this.committedPosition.get());
            this.committedPosition.set(this.wrotePostion.get());
        }
    }

    return this.getCommittedPosition();
}

public AppendMessageResult appendMessage(final Object msg, final AppendMessageCallback cb) {
    assert msg != null;
    assert cb != null;

    // file当前position
    int currentPos = this.wrotePostion.get();

    if (currentPos < this.fileSize) {
        // 获取DirectByteBuffer的一个分片，重置了mark，position，limit是剩下的大小，和原来的buffer共享同一块内存
        ByteBuffer byteBuffer = this.mappedByteBuffer.slice();
        byteBuffer.position(currentPos);
        AppendMessageResult result =
                cb.doAppend(this.getFileFromOffset(), byteBuffer, this.fileSize - currentPos, msg);
        this.wrotePostion.addAndGet(result.getWroteBytes());
        this.storeTimestamp = result.getStoreTimestamp();
        return result;
    }

    log.error("MapedFile.appendMessage return null, wrotePostion: " + currentPos + " fileSize: "
            + this.fileSize);
    return new AppendMessageResult(AppendMessageStatus.UNKNOWN_ERROR);
}