Rocketmq 通信协议及编解码

一.通信协议的格式

    1.length：4字节整数，二三四部分长度总和；（length - 4 - headLength）

    2.header length：4字节整数，第三部分header data长度；

    3.header data：存放Json序列化的数据；

    4.body data：应用自定义二进制序列化的数据。

二.消息的编码过程

public ByteBuffer encode() {
    // 1> header length size
    int length = 4;    //消息头及序列化的长度 占用4个字节

    // 2> header data length
    //将消息头编码成byte[]
    byte[] headerData = this.headerEncode(); 
    //计算头部长度 
    length += headerData.length;              

    // 3> body data length
    if (this.body != null) {
        //消息主体长度
        length += body.length;                
    }
    //分配ByteBuffer, 这边加了4, 
    //这是因为在消息总长度的计算中没有将存储头部长度的4个字节计算在内
    ByteBuffer result = ByteBuffer.allocate(4 + length);  

    // length
    //将消息总长度放入ByteBuffer
    result.putInt(length);   

    // header length
    //将消息头长度放入ByteBuffer
    result.put(markProtocolType(headerData.length, serializeTypeCurrentRPC)); 

    // header data
    //将消息头数据放入ByteBuffer
    result.put(headerData);    

    // body data;
    if (this.body != null) {
        //将消息主体放入ByteBuffer
        result.put(this.body); 
    }
    //重置ByteBuffer的position位置
    result.flip();     

    return result;
}

    /**
     * markProtocolType方法是将RPC类型和headerData长度编码放到一个byte[4]数组中
     *
     * @param source
     * @param type
     * @return
     */
    public static byte[] markProtocolType(int source, SerializeType type) {
        byte[] result = new byte[4];

        result[0] = type.getCode();
        //右移16位后再和255与->“16-24位”
        result[1] = (byte) ((source >> 16) & 0xFF);
        //右移8位后再和255与->“8-16位”
        result[2] = (byte) ((source >> 8) & 0xFF);
        //右移0位后再和255与->“8-0位”
        result[3] = (byte) (source & 0xFF);
        return result;
    }

对于头部数据 - header data 的编码，在 RemotingCommand 中 headerEncode 方法处理： 

    private byte[] headerEncode() {
        this.makeCustomHeaderToNet();
        if (SerializeType.ROCKETMQ == serializeTypeCurrentRPC) {
            return RocketMQSerializable.rocketMQProtocolEncode(this);
        } else {
            return RemotingSerializable.encode(this);
        }
    }

三.消息的解码过程

public static RemotingCommand decode(final ByteBuffer byteBuffer) {
        //获取byteBuffer的总长度
        int length = byteBuffer.limit();

        //获取前4个字节，组装int类型，该长度为总长度 包含2、3、4
        int oriHeaderLen = byteBuffer.getInt();

        //获取消息头的长度，这里和0xFFFFFF做与运算，编码时候的长度即为24位
        int headerLength = getHeaderLength(oriHeaderLen);

        byte[] headerData = new byte[headerLength];
        byteBuffer.get(headerData);

        RemotingCommand cmd = headerDecode(headerData, getProtocolType(oriHeaderLen));

        int bodyLength = length - 4 - headerLength;// 报文体的数据，减去了第二、三部分的长度
        byte[] bodyData = null;
        if (bodyLength > 0) {
            bodyData = new byte[bodyLength];
            byteBuffer.get(bodyData);
        }
        cmd.body = bodyData;

        return cmd;
    }

四.消息的通信方式和通信流程 

 4.1、Client发送请求消息的具体实现

当客户端调用异步通信接口—invokeAsync时候，先由RemotingClient的实现类—NettyRemotingClient根据addr获取相应的channel（如果本地缓存中没有则创建），随后调用invokeAsyncImpl方法，将数据流转给抽象类NettyRemotingAbstract处理（真正做完发送请求动作的是在NettyRemotingAbstract抽象类的invokeAsyncImpl方法里面）。具体发送请求消息的源代码如下所示：

/**
     * invokeAsync（异步调用）
     * 
     * @param channel
     * @param request
     * @param timeoutMillis
     * @param invokeCallback
     * @throws InterruptedException
     * @throws RemotingTooMuchRequestException
     * @throws RemotingTimeoutException
     * @throws RemotingSendRequestException
     */
    public void invokeAsyncImpl(final Channel channel, final RemotingCommand request, final long timeoutMillis,
        final InvokeCallback invokeCallback)
        throws InterruptedException, RemotingTooMuchRequestException, RemotingTimeoutException, RemotingSendRequestException {
        //相当于request ID, RemotingCommand会为每一个request产生一个request ID, 从0开始, 每次加1

        final int opaque = request.getOpaque();
        boolean acquired = this.semaphoreAsync.tryAcquire(timeoutMillis, TimeUnit.MILLISECONDS);
        if (acquired) {
            final SemaphoreReleaseOnlyOnce once = new SemaphoreReleaseOnlyOnce(this.semaphoreAsync);
            //根据request ID构建ResponseFuture
            final ResponseFuture responseFuture = new ResponseFuture(opaque, timeoutMillis, invokeCallback, once);
            //将ResponseFuture放入responseTable
            this.responseTable.put(opaque, responseFuture);
            try {
                //使用Netty的channel发送请求数据
                channel.writeAndFlush(request).addListener(new ChannelFutureListener() {
                    //消息发送后执行
                    @Override
                    public void operationComplete(ChannelFuture f) throws Exception {
                        if (f.isSuccess()) {
                            //如果发送消息成功给Server，那么这里直接Set后return
                            responseFuture.setSendRequestOK(true);
                            return;
                        } else {
                            responseFuture.setSendRequestOK(false);
                        }

                        responseFuture.putResponse(null);
                        responseTable.remove(opaque);
                        try {
                            //执行回调
                            executeInvokeCallback(responseFuture);
                        } catch (Throwable e) {
                            log.warn("excute callback in writeAndFlush addListener, and callback throw", e);
                        } finally {
                            //释放信号量
                            responseFuture.release();
                        }

                        log.warn("send a request command to channel <{}> failed.", RemotingHelper.parseChannelRemoteAddr(channel));
                    }
                });
            } catch (Exception e) {
                //异常处理
                responseFuture.release();
                log.warn("send a request command to channel <" + RemotingHelper.parseChannelRemoteAddr(channel) + "> Exception", e);
                throw new RemotingSendRequestException(RemotingHelper.parseChannelRemoteAddr(channel), e);
            }
        } else {
            if (timeoutMillis <= 0) {
                throw new RemotingTooMuchRequestException("invokeAsyncImpl invoke too fast");
            } else {
                String info =
                    String.format("invokeAsyncImpl tryAcquire semaphore timeout, %dms, waiting thread nums: %d semaphoreAsyncValue: %d",
                        timeoutMillis,
                        this.semaphoreAsync.getQueueLength(),
                        this.semaphoreAsync.availablePermits()
                    );
                log.warn(info);
                throw new RemotingTimeoutException(info);
            }
        }
    }

 1）responseTable—保存请求码与响应关联映射

protected final ConcurrentHashMap<Integer /* opaque */, ResponseFuture> responseTable 

 

opaque表示请求发起方在同个连接上不同的请求标识代码，每次发送一个消息的时候，可以选择同步阻塞/异步非阻塞的方式。无论是哪种通信方式，都会保存请求操作码至ResponseFuture的Map映射—responseTable中

 2）ResponseFuture—保存返回响应（包括回调执行方法和信号量）

 

public ResponseFuture(int opaque, long timeoutMillis, InvokeCallback invokeCallback,
        SemaphoreReleaseOnlyOnce once) {
        this.opaque = opaque;
        this.timeoutMillis = timeoutMillis;
        this.invokeCallback = invokeCallback;
        this.once = once;
    }

 3）异常发送流程处理—定时扫描responseTable本地缓存

在发送消息时候，如果遇到异常情况（比如服务端没有response返回给客户端或者response因网络而丢失），上面所述的responseTable的本地缓存Map将会出现堆积情况。这个时候需要一个定时任务来专门做responseTable的清理回收。在RocketMQ的客户端/服务端启动时候会产生一个频率为1s调用一次来的定时任务检查所有的responseTable缓存中的responseFuture变量，判断是否已经得到返回, 并进行相应的处理

public void scanResponseTable() {
        final List<ResponseFuture> rfList = new LinkedList<ResponseFuture>();
        Iterator<Entry<Integer, ResponseFuture>> it = this.responseTable.entrySet().iterator();
        while (it.hasNext()) {
            Entry<Integer, ResponseFuture> next = it.next();
            ResponseFuture rep = next.getValue();

            if ((rep.getBeginTimestamp() + rep.getTimeoutMillis() + 1000) <= System.currentTimeMillis()) {
                rep.release();
                it.remove();
                rfList.add(rep);
                log.warn("remove timeout request, " + rep);
            }
        }

        for (ResponseFuture rf : rfList) {
            try {
                executeInvokeCallback(rf);
            } catch (Throwable e) {
                log.warn("scanResponseTable, operationComplete Exception", e);
            }
        }
    }

 4.2、Server端接收消息并进行处理的具体实现

Server端接收消息的处理入口在NettyServerHandler类的channelRead0方法中，其中调用了processMessageReceived方法（这里省略了Netty服务端消息流转的大部分流程和逻辑）。其中服务端最为重要的处理请求方法实现如下：

public void processRequestCommand(final ChannelHandlerContext ctx, final RemotingCommand cmd) {
    //根据RemotingCommand中的code获取processor和ExecutorService
    final Pair<NettyRequestProcessor, ExecutorService> matched = this.processorTable.get(cmd.getCode());
    final Pair<NettyRequestProcessor, ExecutorService> pair = null == matched ? this.defaultRequestProcessor : matched;
    final int opaque = cmd.getOpaque();

    if (pair != null) {
        Runnable run = new Runnable() {
            @Override
            public void run() {
                try {
                    //rpc hook
                    RPCHook rpcHook = NettyRemotingAbstract.this.getRPCHook();
                    if (rpcHook != null) {
                        rpcHook.doBeforeRequest(RemotingHelper.parseChannelRemoteAddr(ctx.channel()), cmd);
                    }
                    //processor处理请求
                    final RemotingCommand response = pair.getObject1().processRequest(ctx, cmd);
                    //rpc hook
                    if (rpcHook != null) {
                        rpcHook.doAfterResponse(RemotingHelper.parseChannelRemoteAddr(ctx.channel()), cmd, response);
                    }

                    if (!cmd.isOnewayRPC()) {
                        if (response != null) {
                            response.setOpaque(opaque);
                            response.markResponseType();
                            try {
                                ctx.writeAndFlush(response);
                            } catch (Throwable e) {
                                PLOG.error("process request over, but response failed", e);
                                PLOG.error(cmd.toString());
                                PLOG.error(response.toString());
                            }
                        } else {

                        }
                    }
                } catch (Throwable e) {
                    if (!"com.aliyun.openservices.ons.api.impl.authority.exception.AuthenticationException"
                        .equals(e.getClass().getCanonicalName())) {
                        PLOG.error("process request exception", e);
                        PLOG.error(cmd.toString());
                    }

                    if (!cmd.isOnewayRPC()) {
                        final RemotingCommand response = RemotingCommand.createResponseCommand(RemotingSysResponseCode.SYSTEM_ERROR, //
                            RemotingHelper.exceptionSimpleDesc(e));
                        response.setOpaque(opaque);
                        ctx.writeAndFlush(response);
                    }
                }
            }
        };

        if (pair.getObject1().rejectRequest()) {
            final RemotingCommand response = RemotingCommand.createResponseCommand(RemotingSysResponseCode.SYSTEM_BUSY,
                "[REJECTREQUEST]system busy, start flow control for a while");
            response.setOpaque(opaque);
            ctx.writeAndFlush(response);
            return;
        }

        try {
            //封装requestTask
            final RequestTask requestTask = new RequestTask(run, ctx.channel(), cmd);
            //想线程池提交requestTask
            pair.getObject2().submit(requestTask);
        } catch (RejectedExecutionException e) {
            if ((System.currentTimeMillis() % 10000) == 0) {
                PLOG.warn(RemotingHelper.parseChannelRemoteAddr(ctx.channel()) //
                    + ", too many requests and system thread pool busy, RejectedExecutionException " //
                    + pair.getObject2().toString() //
                    + " request code: " + cmd.getCode());
            }

            if (!cmd.isOnewayRPC()) {
                final RemotingCommand response = RemotingCommand.createResponseCommand(RemotingSysResponseCode.SYSTEM_BUSY,
                    "[OVERLOAD]system busy, start flow control for a while");
                response.setOpaque(opaque);
                ctx.writeAndFlush(response);
            }
        }
    } else {
        String error = " request type " + cmd.getCode() + " not supported";
        //构建response
        final RemotingCommand response =
            RemotingCommand.createResponseCommand(RemotingSysResponseCode.REQUEST_CODE_NOT_SUPPORTED, error);
        response.setOpaque(opaque);
        ctx.writeAndFlush(response);
        PLOG.error(RemotingHelper.parseChannelRemoteAddr(ctx.channel()) + error);
    }
}

 

上面的请求处理方法中根据RemotingCommand的请求业务码来匹配到相应的业务处理器；然后生成一个新的线程提交至对应的业务线程池进行异步处理。
（1）processorTable—请求业务码与业务处理、业务线程池的映射变量

    protected final HashMap<Integer/* request code */, Pair<NettyRequestProcessor, ExecutorService>> processorTable =
        new HashMap<Integer, Pair<NettyRequestProcessor, ExecutorService>>(64);

我想RocketMQ这种做法是为了给不同类型的请求业务码指定不同的处理器Processor处理，同时消息实际的处理并不是在当前线程，而是被封装成task放到业务处理器Processor对应的线程池中完成异步执行。(在RocketMQ中能看到很多地方都是这样的处理，这样的设计能够最大程度的保证异步，保证每个线程都专注处理自己负责的东西）

参考文章：https://www.jianshu.com/p/d5da161efc33