RocketMQ中Broker的HA策略源码分析

Broker的HA策略分为两部分
①同步元数据
②同步消息数据

 

同步元数据

在Slave启动时，会启动一个定时任务用来从master同步元数据

if (role == BrokerRole.SLAVE) {
    if (null != slaveSyncFuture) {
        slaveSyncFuture.cancel(false);
    }
    this.slaveSynchronize.setMasterAddr(null);
    slaveSyncFuture = this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {
        @Override
        public void run() {
            try {
                BrokerController.this.slaveSynchronize.syncAll();
            }
            catch (Throwable e) {
                log.error("ScheduledTask SlaveSynchronize syncAll error.", e);
            }
        }
    }, 1000 * 3, 1000 * 10, TimeUnit.MILLISECONDS);
} 

这里设置了定时任务，执行slaveSynchronize的syncAll方法
可以注意在之前会通过setMasterAddr将Master的地址设为null，这是由于在后面会通过另一个定时任务registerBrokerAll来向NameServer获取Master的地址，详见：

【RocketMQ中Broker的启动源码分析（二）】

 

SlaveSynchronize的syncAll方法：

public void syncAll() {
    this.syncTopicConfig();
    this.syncConsumerOffset();
    this.syncDelayOffset();
    this.syncSubscriptionGroupConfig();
}

这个方法会依次调用四个方法，来同步相应信息：
syncTopicConfig：同步topic的配置信息
syncConsumerOffset：同步Consumer的Offset信息
syncDelayOffset：同步延迟队列信息
syncSubscriptionGroupConfig：同步订阅信息

由于这几个方法的实现是类似的，这里就只看下syncTopicConfig的实现：
syncTopicConfig方法：

private void syncTopicConfig() {
    String masterAddrBak = this.masterAddr;
    if (masterAddrBak != null && !masterAddrBak.equals(brokerController.getBrokerAddr())) {
        try {
            TopicConfigSerializeWrapper topicWrapper =
                this.brokerController.getBrokerOuterAPI().getAllTopicConfig(masterAddrBak);
            if (!this.brokerController.getTopicConfigManager().getDataVersion()
                .equals(topicWrapper.getDataVersion())) {

                this.brokerController.getTopicConfigManager().getDataVersion()
                    .assignNewOne(topicWrapper.getDataVersion());
                this.brokerController.getTopicConfigManager().getTopicConfigTable().clear();
                this.brokerController.getTopicConfigManager().getTopicConfigTable()
                    .putAll(topicWrapper.getTopicConfigTable());
                this.brokerController.getTopicConfigManager().persist();

                log.info("Update slave topic config from master, {}", masterAddrBak);
            }
        } catch (Exception e) {
            log.error("SyncTopicConfig Exception, {}", masterAddrBak, e);
        }
    }
}

这里首先获取master的地址masterAddr，由于registerBrokerAll定时任务的存在，即便这一次没有获取到masterAddr，只要节点中有master，总会在后面定时执行时从NameServer中获取到

当获取到master地址后，通过BrokerOuterAPI的getAllTopicConfig方法，向master请求
BrokerOuterAPI的getAllTopicConfig方法：

public TopicConfigSerializeWrapper getAllTopicConfig(
    final String addr) throws RemotingConnectException, RemotingSendRequestException,
    RemotingTimeoutException, InterruptedException, MQBrokerException {
    RemotingCommand request = RemotingCommand.createRequestCommand(RequestCode.GET_ALL_TOPIC_CONFIG, null);

    RemotingCommand response = this.remotingClient.invokeSync(MixAll.brokerVIPChannel(true, addr), request, 3000);
    assert response != null;
    switch (response.getCode()) {
        case ResponseCode.SUCCESS: {
            return TopicConfigSerializeWrapper.decode(response.getBody(), TopicConfigSerializeWrapper.class);
        }
        default:
            break;
    }

    throw new MQBrokerException(response.getCode(), response.getRemark());
}

首先构建GET_ALL_TOPIC_CONFIG求情指令，然后通过remotingClient的invokeSync进行同步发送，注意这里会通过MixAll的brokerVIPChannel方法，得到对应的master地址的VIP通道地址，就是端口号减2，这在我之前的博客中介绍过
有关同步发送在  【RocketMQ中Producer消息的发送源码分析】 中详细介绍过

 

请求发送给master后，来看看master是怎么处理的
master端在收到请求后会通过AdminBrokerProcessor的processRequest方法判别请求指令：

case RequestCode.GET_ALL_TOPIC_CONFIG:
    return this.getAllTopicConfig(ctx, request);

执行getAllTopicConfig方法：

private RemotingCommand getAllTopicConfig(ChannelHandlerContext ctx, RemotingCommand request) {
    final RemotingCommand response = RemotingCommand.createResponseCommand(GetAllTopicConfigResponseHeader.class);
    // final GetAllTopicConfigResponseHeader responseHeader =
    // (GetAllTopicConfigResponseHeader) response.readCustomHeader();

    String content = this.brokerController.getTopicConfigManager().encode();
    if (content != null && content.length() > 0) {
        try {
            response.setBody(content.getBytes(MixAll.DEFAULT_CHARSET));
        } catch (UnsupportedEncodingException e) {
            log.error("", e);

            response.setCode(ResponseCode.SYSTEM_ERROR);
            response.setRemark("UnsupportedEncodingException " + e);
            return response;
        }
    } else {
        log.error("No topic in this broker, client: {}", ctx.channel().remoteAddress());
        response.setCode(ResponseCode.SYSTEM_ERROR);
        response.setRemark("No topic in this broker");
        return response;
    }

    response.setCode(ResponseCode.SUCCESS);
    response.setRemark(null);

    return response;
}

这里会将TopicConfigManager中保存的topicConfigTable：

private final ConcurrentMap<String, TopicConfig> topicConfigTable =
        new ConcurrentHashMap<String, TopicConfig>(1024);

将这个map通过encode方法转换成json字符串，再通过Netty发送给slave

回到slave中，在同步发送的情况下，会等待会送响应，收到响应后：

switch (response.getCode()) {
    case ResponseCode.SUCCESS: {
        return TopicConfigSerializeWrapper.decode(response.getBody(), TopicConfigSerializeWrapper.class);
    }
    default:
        break;
}

通过decode解码，将json字符串转换为map封装在 TopicConfigSerializeWrapper中

回到syncTopicConfig方法中：
得到TopicConfigSerializeWrapper实例后

if (!this.brokerController.getTopicConfigManager().getDataVersion()
    .equals(topicWrapper.getDataVersion())) {

    this.brokerController.getTopicConfigManager().getDataVersion()
        .assignNewOne(topicWrapper.getDataVersion());
    this.brokerController.getTopicConfigManager().getTopicConfigTable().clear();
    this.brokerController.getTopicConfigManager().getTopicConfigTable()
        .putAll(topicWrapper.getTopicConfigTable());
    this.brokerController.getTopicConfigManager().persist();

    log.info("Update slave topic config from master, {}", masterAddrBak);
}

判断版本是否一致，若不一致，会进行替换，这样slave的Topic配置信息就和master保持同步了

其他三种信息的同步同理

 

 同步消息数据

在master启动时，会通过JDK的NIO方式启动一个HA服务线程，用以处理slave的连接：

public void run() {
    log.info(this.getServiceName() + " service started");

    while (!this.isStopped()) {
        try {
            this.selector.select(1000);
            Set<SelectionKey> selected = this.selector.selectedKeys();

            if (selected != null) {
                for (SelectionKey k : selected) {
                    if ((k.readyOps() & SelectionKey.OP_ACCEPT) != 0) {
                        SocketChannel sc = ((ServerSocketChannel) k.channel()).accept();

                        if (sc != null) {
                            HAService.log.info("HAService receive new connection, "
                                + sc.socket().getRemoteSocketAddress());

                            try {
                                HAConnection conn = new HAConnection(HAService.this, sc);
                                conn.start();
                                HAService.this.addConnection(conn);
                            } catch (Exception e) {
                                log.error("new HAConnection exception", e);
                                sc.close();
                            }
                        }
                    } else {
                        log.warn("Unexpected ops in select " + k.readyOps());
                    }
                }

                selected.clear();
            }
        } catch (Exception e) {
            log.error(this.getServiceName() + " service has exception.", e);
        }
    }

    log.info(this.getServiceName() + " service end");
}

这里就是非常典型的JDK NIO的使用，在侦听到连接取得SocketChannel后，将其封装为HAConnection

public HAConnection(final HAService haService, final SocketChannel socketChannel) throws IOException {
    this.haService = haService;
    this.socketChannel = socketChannel;
    this.clientAddr = this.socketChannel.socket().getRemoteSocketAddress().toString();
    this.socketChannel.configureBlocking(false);
    this.socketChannel.socket().setSoLinger(false, -1);
    this.socketChannel.socket().setTcpNoDelay(true);
    this.socketChannel.socket().setReceiveBufferSize(1024 * 64);
    this.socketChannel.socket().setSendBufferSize(1024 * 64);
    this.writeSocketService = new WriteSocketService(this.socketChannel);
    this.readSocketService = new ReadSocketService(this.socketChannel);
    this.haService.getConnectionCount().incrementAndGet();
}

在构造方法内进行了对socketChannel的一些配置，还创建了一个WriteSocketService和一个ReadSocketService，这两个是后续处理消息同步的基础

在创建完HAConnection后，调用其start方法：

public void start() {
    this.readSocketService.start();
    this.writeSocketService.start();
}

这里会启动两个线程，分别处理读取slave发送的数据，以及向slave发送数据

 

到这里，先不急着分析master了，来看看slave端
slave在启动时，会启动HAClient的线程：

public void run() {
    log.info(this.getServiceName() + " service started");

    while (!this.isStopped()) {
        try {
            if (this.connectMaster()) {

                if (this.isTimeToReportOffset()) {
                    boolean result = this.reportSlaveMaxOffset(this.currentReportedOffset);
                    if (!result) {
                        this.closeMaster();
                    }
                }

                this.selector.select(1000);

                boolean ok = this.processReadEvent();
                if (!ok) {
                    this.closeMaster();
                }

                if (!reportSlaveMaxOffsetPlus()) {
                    continue;
                }

                long interval =
                    HAService.this.getDefaultMessageStore().getSystemClock().now()
                        - this.lastWriteTimestamp;
                if (interval > HAService.this.getDefaultMessageStore().getMessageStoreConfig()
                    .getHaHousekeepingInterval()) {
                    log.warn("HAClient, housekeeping, found this connection[" + this.masterAddress
                        + "] expired, " + interval);
                    this.closeMaster();
                    log.warn("HAClient, master not response some time, so close connection");
                }
            } else {
                this.waitForRunning(1000 * 5);
            }
        } catch (Exception e) {
            log.warn(this.getServiceName() + " service has exception. ", e);
            this.waitForRunning(1000 * 5);
        }
    }

    log.info(this.getServiceName() + " service end");
}

在这个while循环中，首先通过connectMaster检查是否和master连接了

connectMaster方法：

private boolean connectMaster() throws ClosedChannelException {
    if (null == socketChannel) {
        String addr = this.masterAddress.get();
        if (addr != null) {

            SocketAddress socketAddress = RemotingUtil.string2SocketAddress(addr);
            if (socketAddress != null) {
                this.socketChannel = RemotingUtil.connect(socketAddress);
                if (this.socketChannel != null) {
                    this.socketChannel.register(this.selector, SelectionKey.OP_READ);
                }
            }
        }

        this.currentReportedOffset = HAService.this.defaultMessageStore.getMaxPhyOffset();

        this.lastWriteTimestamp = System.currentTimeMillis();
    }

    return this.socketChannel != null;
}

若是socketChannel为null，意味着并没有产生连接，或者连接断开
需要重新根据masterAddress建立网络连接

只要是需要建立连接，都需要通过defaultMessageStore的getMaxPhyOffset方法，获取本地最大的Offset，由currentReportedOffset保存，后续用于向master报告；以及保存了一个时间戳lastWriteTimestamp，用于之后的校对

当确保与master的连接建立成功后，通过isTimeToReportOffset方法，检查是否需要向master报告当前的最大Offset

isTimeToReportOffset方法：

private boolean isTimeToReportOffset() {
    long interval =
        HAService.this.defaultMessageStore.getSystemClock().now() - this.lastWriteTimestamp;
    boolean needHeart = interval > HAService.this.defaultMessageStore.getMessageStoreConfig()
        .getHaSendHeartbeatInterval();

    return needHeart;
}

这里就通过lastWriteTimestamp和当前时间检查，判断是否达到了报告时间间隔HaSendHeartbeatInterval，默认5s

若是达到了，就需要通过reportSlaveMaxOffset方法，将记录的currentReportedOffset这个最大的offset发送给master

reportSlaveMaxOffset方法：

private boolean reportSlaveMaxOffset(final long maxOffset) {
    this.reportOffset.position(0);
    this.reportOffset.limit(8);
    this.reportOffset.putLong(maxOffset);
    this.reportOffset.position(0);
    this.reportOffset.limit(8);

    for (int i = 0; i < 3 && this.reportOffset.hasRemaining(); i++) {
        try {
            this.socketChannel.write(this.reportOffset);
        } catch (IOException e) {
            log.error(this.getServiceName()
                + "reportSlaveMaxOffset this.socketChannel.write exception", e);
            return false;
        }
    }

    return !this.reportOffset.hasRemaining();
}

其中reportOffset是专门用来缓存offset的ByteBuffer

private final ByteBuffer reportOffset = ByteBuffer.allocate(8);

将maxOffset存放在reportOffset中，然后通过socketChannel的write方法，完成向master的发送

其中hasRemaining方法用来检查当前位置是否已经达到缓冲区极限limit，确保reportOffset 中的内容能被完全发送出去

发送成功后，会调用selector的select方法，在超时时间内进行NIO的轮询，等待master的回送

通过这我们可以看出slave在和master建立连接后，会定时向master报告自己当前的offset

 

来看看master收到offset后是如何处理的：

在master端会通过前面提到的ReadSocketService线程进行处理：

public void run() {
    HAConnection.log.info(this.getServiceName() + " service started");

    while (!this.isStopped()) {
        try {
            this.selector.select(1000);
            boolean ok = this.processReadEvent();
            if (!ok) {
                HAConnection.log.error("processReadEvent error");
                break;
            }

            long interval = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now() - this.lastReadTimestamp;
            if (interval > HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig().getHaHousekeepingInterval()) {
                log.warn("ha housekeeping, found this connection[" + HAConnection.this.clientAddr + "] expired, " + interval);
                break;
            }
        } catch (Exception e) {
            HAConnection.log.error(this.getServiceName() + " service has exception.", e);
            break;
        }
    }

    this.makeStop();

    writeSocketService.makeStop();

    haService.removeConnection(HAConnection.this);

    HAConnection.this.haService.getConnectionCount().decrementAndGet();

    SelectionKey sk = this.socketChannel.keyFor(this.selector);
    if (sk != null) {
        sk.cancel();
    }

    try {
        this.selector.close();
        this.socketChannel.close();
    } catch (IOException e) {
        HAConnection.log.error("", e);
    }

    HAConnection.log.info(this.getServiceName() + " service end");
}

这里的while循环中首先也是通过selector的select方法，在超时时间内进行NIO的轮询

轮询结束后的进一步的处理由processReadEvent来完成：

private boolean processReadEvent() {
        int readSizeZeroTimes = 0;

        if (!this.byteBufferRead.hasRemaining()) {
            this.byteBufferRead.flip();
            this.processPostion = 0;
        }

        while (this.byteBufferRead.hasRemaining()) {
            try {
                int readSize = this.socketChannel.read(this.byteBufferRead);
                if (readSize > 0) {
                    readSizeZeroTimes = 0;
                    this.lastReadTimestamp = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now();
                    if ((this.byteBufferRead.position() - this.processPostion) >= 8) {
                        int pos = this.byteBufferRead.position() - (this.byteBufferRead.position() % 8);
                        long readOffset = this.byteBufferRead.getLong(pos - 8);
                        this.processPostion = pos;

                        HAConnection.this.slaveAckOffset = readOffset;
                        if (HAConnection.this.slaveRequestOffset < 0) {
                            HAConnection.this.slaveRequestOffset = readOffset;
                            log.info("slave[" + HAConnection.this.clientAddr + "] request offset " + readOffset);
                        }

                        HAConnection.this.haService.notifyTransferSome(HAConnection.this.slaveAckOffset);
                    }
                } else if (readSize == 0) {
                    if (++readSizeZeroTimes >= 3) {
                        break;
                    }
                } else {
                    log.error("read socket[" + HAConnection.this.clientAddr + "] < 0");
                    return false;
                }
            } catch (IOException e) {
                log.error("processReadEvent exception", e);
                return false;
            }
        }

        return true;
    }
}

这个方法其实就是通过socketChannel的read方法，将slave发送过来的数据存入byteBufferRead中
在确保发送过来的数据能达到8字节时，取出long类型的offset值，然后交给HAConnection的slaveAckOffset成员进行保存

其中slaveRequestOffset是用来处理第一次连接时的同步

notifyTransferSome方法是作为同步master时，进行相应的唤醒操作，异步master则没有要求，在后面具体分析

也就是说ReadSocketService这个线程，只是不断地读取并更新slave发送来的offset数据

 

再来看看WriteSocketService线程是如何进行向slave的发送：

public void run() {
    HAConnection.log.info(this.getServiceName() + " service started");

    while (!this.isStopped()) {
        try {
            this.selector.select(1000);

            if (-1 == HAConnection.this.slaveRequestOffset) {
                Thread.sleep(10);
                continue;
            }

            if (-1 == this.nextTransferFromWhere) {
                if (0 == HAConnection.this.slaveRequestOffset) {
                    long masterOffset = HAConnection.this.haService.getDefaultMessageStore().getCommitLog().getMaxOffset();
                    masterOffset =
                        masterOffset
                            - (masterOffset % HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig()
                            .getMapedFileSizeCommitLog());

                    if (masterOffset < 0) {
                        masterOffset = 0;
                    }

                    this.nextTransferFromWhere = masterOffset;
                } else {
                    this.nextTransferFromWhere = HAConnection.this.slaveRequestOffset;
                }

                log.info("master transfer data from " + this.nextTransferFromWhere + " to slave[" + HAConnection.this.clientAddr
                    + "], and slave request " + HAConnection.this.slaveRequestOffset);
            }

            if (this.lastWriteOver) {

                long interval =
                    HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now() - this.lastWriteTimestamp;

                if (interval > HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig()
                    .getHaSendHeartbeatInterval()) {

                    // Build Header
                    this.byteBufferHeader.position(0);
                    this.byteBufferHeader.limit(headerSize);
                    this.byteBufferHeader.putLong(this.nextTransferFromWhere);
                    this.byteBufferHeader.putInt(0);
                    this.byteBufferHeader.flip();

                    this.lastWriteOver = this.transferData();
                    if (!this.lastWriteOver)
                        continue;
                }
            } else {
                this.lastWriteOver = this.transferData();
                if (!this.lastWriteOver)
                    continue;
            }

            SelectMappedBufferResult selectResult =
                HAConnection.this.haService.getDefaultMessageStore().getCommitLogData(this.nextTransferFromWhere);
            if (selectResult != null) {
                int size = selectResult.getSize();
                if (size > HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig().getHaTransferBatchSize()) {
                    size = HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig().getHaTransferBatchSize();
                }

                long thisOffset = this.nextTransferFromWhere;
                this.nextTransferFromWhere += size;

                selectResult.getByteBuffer().limit(size);
                this.selectMappedBufferResult = selectResult;

                // Build Header
                this.byteBufferHeader.position(0);
                this.byteBufferHeader.limit(headerSize);
                this.byteBufferHeader.putLong(thisOffset);
                this.byteBufferHeader.putInt(size);
                this.byteBufferHeader.flip();

                this.lastWriteOver = this.transferData();
            } else {

                HAConnection.this.haService.getWaitNotifyObject().allWaitForRunning(100);
            }
        } catch (Exception e) {

            HAConnection.log.error(this.getServiceName() + " service has exception.", e);
            break;
        }
    }

    HAConnection.this.haService.getWaitNotifyObject().removeFromWaitingThreadTable();

    if (this.selectMappedBufferResult != null) {
        this.selectMappedBufferResult.release();
    }

    this.makeStop();

    readSocketService.makeStop();

    haService.removeConnection(HAConnection.this);

    SelectionKey sk = this.socketChannel.keyFor(this.selector);
    if (sk != null) {
        sk.cancel();
    }

    try {
        this.selector.close();
        this.socketChannel.close();
    } catch (IOException e) {
        HAConnection.log.error("", e);
    }

    HAConnection.log.info(this.getServiceName() + " service end");
}

这里一开始会对slaveRequestOffset进行一次判断，当且仅当slaveRequestOffset初始化的时候是才是-1

也就是说当slave还没有发送过来offset时，WriteSocketService线程只会干等

当slave发送来offset后
首先对nextTransferFromWhere进行了判断，nextTransferFromWhere和slaveRequestOffset一样，在初始化的时候为-1
也就代表着master和slave刚刚建立连接，并没有进行过一次消息的同步！

此时会对修改了的slaveRequestOffset进行判断
若是等于0，说明slave没有任何消息的历史记录，那么此时master会取得自身的MaxOffset，根据这个MaxOffset，通过：

masterOffset =  masterOffset
                - (masterOffset % HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig()
                .getMapedFileSizeCommitLog() /* 1G */);

计算出最后一个文件开始的offset
也就是说，当slave没有消息的历史记录，master只会从本地最后一个CommitLog文件开始的地方，将消息数据发送给slave

若是slave有数据，就从slave发送来的offset的位置起，进行发送，通过nextTransferFromWhere记录这个offset值

接着对lastWriteOver进行了判断，lastWriteOver是一个状态量，用来表示上次发送是否传输完毕，初始化是true

若是true，这里会进行一次时间检查，lastWriteTimestamp记录最后一次发送的时间
一次来判断是否超过了时间间隔haSendHeartbeatInterval（默认5s）
也就是说至少有5s，master没有向slave发送任何消息
那么此时就会发送一个心跳包

其中byteBufferHeader是一个12字节的ByteBuffer：

private final int headerSize = 8 + 4;
private final ByteBuffer byteBufferHeader = ByteBuffer.allocate(headerSize);

这里就简单地构造了一个心跳包，后续通过transferData方法来完成数据的发送

若是 lastWriteOver为false，则表示上次数据没有发送完，就需要通过transferData方法，将剩余数据继续发送，只要没发送完，只会重复循环，直到发完

先继续往下看，下面就是发送具体的消息数据了：
首先根据nextTransferFromWhere，也就是刚才保存的offset，通过DefaultMessageStore的getCommitLogData方法，其实际上调用的是CommitLog的getData方法，这个方法在

【RocketMQ中Broker的启动源码分析（二）】中关于消息调度（ReputMessageService）时详细介绍过

根据offset找到对应的CommitLog文件，将其从offset对应起始处所有数据读入ByteBuffer中，由SelectMappedBufferResult封装

这里若是master已将将所有本地数据同步给了slave，那么得到的SelectMappedBufferResult就会为null，会调用：

HAConnection.this.haService.getWaitNotifyObject().allWaitForRunning(100);

将自身阻塞，超时等待100ms，要么一直等到超时时间到了，要么就会在后面所讲的同步双传中被同步master唤醒

 

在得到SelectMappedBufferResult后，这里会对读取到的数据大小进行一次判断，若是大于haTransferBatchSize（默认32K），将size改为32K，实际上就是对发送数据大小的限制，大于32K会切割，每次最多只允许发送32k

通过thisOffset记录nextTransferFromWhere即offset
更新nextTransferFromWhere值，以便下一次定位
还会将读取到的数据结果selectResult交给selectMappedBufferResult保存

然后构建消息头，这里就和心跳包格式一样，前八字节存放offset，后四字节存放数据大小

最后调用transferData方法，进行发送：

private boolean transferData() throws Exception {
    int writeSizeZeroTimes = 0;
    // Write Header
    while (this.byteBufferHeader.hasRemaining()) {
        int writeSize = this.socketChannel.write(this.byteBufferHeader);
        if (writeSize > 0) {
            writeSizeZeroTimes = 0;
            this.lastWriteTimestamp = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now();
        } else if (writeSize == 0) {
            if (++writeSizeZeroTimes >= 3) {
                break;
            }
        } else {
            throw new Exception("ha master write header error < 0");
        }
    }

    if (null == this.selectMappedBufferResult) {
        return !this.byteBufferHeader.hasRemaining();
    }

    writeSizeZeroTimes = 0;

    // Write Body
    if (!this.byteBufferHeader.hasRemaining()) {
        while (this.selectMappedBufferResult.getByteBuffer().hasRemaining()) {
            int writeSize = this.socketChannel.write(this.selectMappedBufferResult.getByteBuffer());
            if (writeSize > 0) {
                writeSizeZeroTimes = 0;
                this.lastWriteTimestamp = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now();
            } else if (writeSize == 0) {
                if (++writeSizeZeroTimes >= 3) {
                    break;
                }
            } else {
                throw new Exception("ha master write body error < 0");
            }
        }
    }

    boolean result = !this.byteBufferHeader.hasRemaining() && !this.selectMappedBufferResult.getByteBuffer().hasRemaining();

    if (!this.selectMappedBufferResult.getByteBuffer().hasRemaining()) {
        this.selectMappedBufferResult.release();
        this.selectMappedBufferResult = null;
    }

    return result;
}

首先将byteBufferHeader中的12字节消息头通过socketChannel的write方法发送出去
然后将selectMappedBufferResult中的ByteBuffer的消息数据发送出去

若是selectMappedBufferResult等于null，说明是心跳包，只发送消息头
无论发送什么都会将时间记录在lastWriteTimestamp中，以便后续发送心跳包的判断

看到这里其实就会发现WriteSocketService线程开启后，只要slave向master发出了第一个offset后，WriteSocketService线程都会不断地将对应位置自己本地的CommitLog文件中的内容发送给slave，直到完全同步后，WriteSocketService线程才会稍微缓缓，进入阻塞100ms以及每隔五秒发一次心跳包的状态

但是只要当Producer向master发送来消息后，由刷盘线程完成持久化后，WriteSocketService线程又会忙碌起来，此时也才是体现同步双写和异步复制的时候

先不急着说这个，来看看slave接收到消息是如何处理的：

是在HAClient的线程中的processReadEvent方法处理的：

private boolean processReadEvent() {
    int readSizeZeroTimes = 0;
    while (this.byteBufferRead.hasRemaining()) {
        try {
            int readSize = this.socketChannel.read(this.byteBufferRead);
            if (readSize > 0) {
                lastWriteTimestamp = HAService.this.defaultMessageStore.getSystemClock().now();
                readSizeZeroTimes = 0;
                boolean result = this.dispatchReadRequest();
                if (!result) {
                    log.error("HAClient, dispatchReadRequest error");
                    return false;
                }
            } else if (readSize == 0) {
                if (++readSizeZeroTimes >= 3) {
                    break;
                }
            } else {
                log.info("HAClient, processReadEvent read socket < 0");
                return false;
            }
        } catch (IOException e) {
            log.info("HAClient, processReadEvent read socket exception", e);
            return false;
        }
    }

    return true;
}

在socketChannel通过read方法将master发送的数据读取到byteBufferRead缓冲区后，由dispatchReadRequest方法做进一步处理

dispatchReadRequest方法：

private boolean dispatchReadRequest() {
    final int msgHeaderSize = 8 + 4; // phyoffset + size
    int readSocketPos = this.byteBufferRead.position();

    while (true) {
        int diff = this.byteBufferRead.position() - this.dispatchPostion;
        if (diff >= msgHeaderSize) {
            long masterPhyOffset = this.byteBufferRead.getLong(this.dispatchPostion);
            int bodySize = this.byteBufferRead.getInt(this.dispatchPostion + 8);

            long slavePhyOffset = HAService.this.defaultMessageStore.getMaxPhyOffset();

            if (slavePhyOffset != 0) {
                if (slavePhyOffset != masterPhyOffset) {
                    log.error("master pushed offset not equal the max phy offset in slave, SLAVE: "
                        + slavePhyOffset + " MASTER: " + masterPhyOffset);
                    return false;
                }
            }

            if (diff >= (msgHeaderSize + bodySize)) {
                byte[] bodyData = new byte[bodySize];
                this.byteBufferRead.position(this.dispatchPostion + msgHeaderSize);
                this.byteBufferRead.get(bodyData);

                HAService.this.defaultMessageStore.appendToCommitLog(masterPhyOffset, bodyData);

                this.byteBufferRead.position(readSocketPos);
                this.dispatchPostion += msgHeaderSize + bodySize;

                if (!reportSlaveMaxOffsetPlus()) {
                    return false;
                }

                continue;
            }
        }

        if (!this.byteBufferRead.hasRemaining()) {
            this.reallocateByteBuffer();
        }

        break;
    }

    return true;
}

这里就首先将12字节的消息头取出来
masterPhyOffset：8字节offset ，bodySize ：4字节消息大小
根据master发来的masterPhyOffset会和自己本地的slavePhyOffset进行校验，以便安全备份

之后就会将byteBufferRead中存放在消息头后面的消息数据取出来，调用appendToCommitLog方法持久化到的CommitLog中

public boolean appendToCommitLog(long startOffset, byte[] data) {
    if (this.shutdown) {
        log.warn("message store has shutdown, so appendToPhyQueue is forbidden");
        return false;
    }

    boolean result = this.commitLog.appendData(startOffset, data);
    if (result) {
        this.reputMessageService.wakeup();
    } else {
        log.error("appendToPhyQueue failed " + startOffset + " " + data.length);
    }

    return result;
}

实际上调用了commitLog的appendData方法将其写入磁盘，这个方法我在前面博客中介绍过

【RocketMQ中Broker的刷盘源码分析】

在完成写入后，需要唤醒reputMessageService消息调度，以便Consumer的消费
关于消息调度详见  【RocketMQ中Broker的启动源码分析（二）】

当然前面说过master还会发送心跳消息，但这里明显没对心跳消息进行处理，只是appendToCommitLog调用时，传入了一个大小为0的byte数组，显然有些不合理，想不通

在完成后，还会调用reportSlaveMaxOffsetPlus方法：

private boolean reportSlaveMaxOffsetPlus() {
    boolean result = true;
    long currentPhyOffset = HAService.this.defaultMessageStore.getMaxPhyOffset();
    if (currentPhyOffset > this.currentReportedOffset) {
        this.currentReportedOffset = currentPhyOffset;
        result = this.reportSlaveMaxOffset(this.currentReportedOffset);
        if (!result) {
            this.closeMaster();
            log.error("HAClient, reportSlaveMaxOffset error, " + this.currentReportedOffset);
        }
    }

    return result;
}

由于完成了写入，那么此时获取到的offset肯定比currentReportedOffset中保存的大，然后再次通过reportSlaveMaxOffset方法，将当前的offset报告给master

这其实上已经完成了异步master的异步复制过程

再来看看同步双写是如何实现的：
和刷盘一样，都是在Producer发送完消息，Broker进行完消息的存储后进行的

【RocketMQ中Broker的消息存储源码分析】

 

在CommitLog的handleHA方法：

public void handleHA(AppendMessageResult result, PutMessageResult putMessageResult, MessageExt messageExt) {
    if (BrokerRole.SYNC_MASTER == this.defaultMessageStore.getMessageStoreConfig().getBrokerRole()) {
        HAService service = this.defaultMessageStore.getHaService();
        if (messageExt.isWaitStoreMsgOK()) {
            // Determine whether to wait
            if (service.isSlaveOK(result.getWroteOffset() + result.getWroteBytes())) {
                GroupCommitRequest request = new GroupCommitRequest(result.getWroteOffset() + result.getWroteBytes());
                service.putRequest(request);
                service.getWaitNotifyObject().wakeupAll();
                boolean flushOK =
                    request.waitForFlush(this.defaultMessageStore.getMessageStoreConfig().getSyncFlushTimeout());
                if (!flushOK) {
                    log.error("do sync transfer other node, wait return, but failed, topic: " + messageExt.getTopic() + " tags: "
                        + messageExt.getTags() + " client address: " + messageExt.getBornHostNameString());
                    putMessageResult.setPutMessageStatus(PutMessageStatus.FLUSH_SLAVE_TIMEOUT);
                }
            }
            // Slave problem
            else {
                // Tell the producer, slave not available
                putMessageResult.setPutMessageStatus(PutMessageStatus.SLAVE_NOT_AVAILABLE);
            }
        }
    }

}

这里就会检查Broker的类型，看以看到只对SYNC_MASTER即同步master进行了操作

这个操作过程其实就和同步刷盘类似

【RocketMQ中Broker的刷盘源码分析】

 

根据Offset+WroteBytes创建一条记录GroupCommitRequest，然后会将添加在List中
然后调用getWaitNotifyObject的wakeupAll方法，把阻塞中的所有WriteSocketService线程唤醒
因为master和slave是一对多的关系，那么这里就会有多个slave连接，也就有多个WriteSocketService线程，保证消息能同步到所有slave中

在唤醒WriteSocketService线程工作后，调用request的waitForFlush方法，将自身阻塞，预示着同步复制的真正开启

在HAService开启时，还开启了一个GroupTransferService线程：

public void run() {
    log.info(this.getServiceName() + " service started");

    while (!this.isStopped()) {
        try {
            this.waitForRunning(10);
            this.doWaitTransfer();
        } catch (Exception e) {
            log.warn(this.getServiceName() + " service has exception. ", e);
        }
    }

    log.info(this.getServiceName() + " service end");
}

这里的工作原理和同步刷盘GroupCommitService基本一致，相似的地方我就不仔细分析了

GroupTransferService同样保存两张List：

private volatile List<CommitLog.GroupCommitRequest> requestsWrite = new ArrayList<>();
private volatile List<CommitLog.GroupCommitRequest> requestsRead = new ArrayList<>();

由这两张List做一个类似JVM新生代的复制算法
在handleHA方法中，就会将创建的GroupCommitRequest记录添加在requestsWrite这个List中

其中doWaitTransfer方法：

private void doWaitTransfer() {
    synchronized (this.requestsRead) {
        if (!this.requestsRead.isEmpty()) {
            for (CommitLog.GroupCommitRequest req : this.requestsRead) {
                boolean transferOK = HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
                for (int i = 0; !transferOK && i < 5; i++) {
                    this.notifyTransferObject.waitForRunning(1000);
                    transferOK = HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
                }

                if (!transferOK) {
                    log.warn("transfer messsage to slave timeout, " + req.getNextOffset());
                }

                req.wakeupCustomer(transferOK);
            }

            this.requestsRead.clear();
        }
    }
}

和刷盘一样，这里会通过复制算法，将requestsWrite和requestsRead进行替换，那么这里的requestsRead实际上就存放着刚才添加的记录

首先取出记录中的NextOffset和push2SlaveMaxOffset比较

push2SlaveMaxOffset值是通过slave发送过来的，在之前说过的ReadSocketService线程中的：

HAConnection.this.haService.notifyTransferSome(HAConnection.this.slaveAckOffset);

notifyTransferSome方法：

public void notifyTransferSome(final long offset) {
    for (long value = this.push2SlaveMaxOffset.get(); offset > value; ) {
        boolean ok = this.push2SlaveMaxOffset.compareAndSet(value, offset);
        if (ok) {
            this.groupTransferService.notifyTransferSome();
            break;
        } else {
            value = this.push2SlaveMaxOffset.get();
        }
    }
}

即便也多个slave连接，这里的push2SlaveMaxOffset永远会记录最大的那个offset

所以在doWaitTransfer中，根据当前NextOffset（完成写入后master本地的offset），进行判断

其实这里主要要考虑到WriteSocketService线程的工作原理，只要本地文件有更新，那么就会向slave发送数据，所以这里由于HA同步是发生在刷盘后的，那么就有可能在这个doWaitTransfer执行前，有slave已经将数据进行了同步，并且向master报告了自己offset，更新了push2SlaveMaxOffset的值

那么

boolean transferOK = HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
```

这个判断就会为真，意味着节点中已经有了备份，所以就会直接调用

req.wakeupCustomer(transferOK);

以此来唤醒刚才在handleHA方法中的阻塞

若是判断为假，就说明没有一个slave完成同步，就需要

for (int i = 0; !transferOK && i < 5; i++) {
    this.notifyTransferObject.waitForRunning(1000);
    transferOK = HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
}

通过waitForRunning进行阻塞，超时等待，最多五次等待，超过时间会向Producer发送FLUSH_SLAVE_TIMEOUT

若是在超时时间内，有slave完成了同步，并向master发送了offset后，在notifyTransferSome方法中：

public void notifyTransferSome(final long offset) {
    for (long value = this.push2SlaveMaxOffset.get(); offset > value; ) {
        boolean ok = this.push2SlaveMaxOffset.compareAndSet(value, offset);
        if (ok) {
            this.groupTransferService.notifyTransferSome();
            break;
        } else {
            value = this.push2SlaveMaxOffset.get();
        }
    }
}

就会更新push2SlaveMaxOffset，并通过notifyTransferSome唤醒上面所说的阻塞

然后再次判断push2SlaveMaxOffset和getNextOffset
成功后唤醒刚才在handleHA方法中的阻塞，同步master的主从复制也就结束
由于同步master的刷盘是在主从复制前发生的，所以同步双写意味着master和slave都会完成消息的持久化

 

至此，RocketMQ中Broker的HA策略分析到此结束