首先来看看start()方法的源码，这个方法有点长，这里拆开来一一分析，首先是注释：
/**
* Start up the connection, including the MainLoop thread.
* Sends the protocol
* version negotiation header, and runs through
* Connection.Start/.StartOk, Connection.Tune/.TuneOk, and then
* calls Connection.Open and waits for the OpenOk. Sets heart-beat
* and frame max values after tuning has taken place.
* @throws IOException if an error is encountered
* either before, or during, protocol negotiation;
* sub-classes {@link ProtocolVersionMismatchException} and
* {@link PossibleAuthenticationFailureException} will be thrown in the
* corresponding circumstances. {@link AuthenticationFailureException}
* will be thrown if the broker closes the connection with ACCESS_REFUSED.
* If an exception is thrown, connection resources allocated can all be
* garbage collected when the connection object is no longer referenced.
*/
首先来看看方法上的注释说了什么：
- 方法的作用是启动连接（start up the connection）, 包括启动MainLoop线程，这个MainLoop线程主要是和broker进行通信交互处理通信帧（Frame）的一个线程（非常的重要！！！）。
- 这个方法会在建立连接的初始化阶段（negotiation）会进行Connection.Start/.StartOk, Connection.Tune/.TuneOk, 调用Connection.Open之后再等待Conenction.OpenOk(这里的都是指AMQP协议层面的)，这个可以参考本文中第一张使用wireshark抓包的网络截图，一一对应的关系。
- 通过broker回复的Connection.Tune帧（帧中包含Channel-Max, Frame-Max, Heartbeat三个参数）设置channelMax, frameMax以及Heartbeat的参数值。
- 一些异常情况。
public void start()
throws IOException, TimeoutException {
initializeConsumerWorkService();
initializeHeartbeatSender();
this._running = true;
// Make sure that the first thing we do is to send the header,
// which should cause any socket errors to show up for us, rather
// than risking them pop out in the MainLoop
AMQChannel.SimpleBlockingRpcContinuation connStartBlocker =
new AMQChannel.SimpleBlockingRpcContinuation();
// We enqueue an RPC continuation here without sending an RPC
// request, since the protocol specifies that after sending
// the version negotiation header, the client (connection
// initiator) is to wait for a connection.start method to
// arrive.
_channel0.enqueueRpc(connStartBlocker);
首先是初始化工作线程池（initializeConsumerWorkService）和初始化心跳线程（initializeHeartbeatSender）并设置运行状态为true(this._isrunning=true，这个值会在MainLoop线程中有用，控制MainLoop线程是否继续运行)。
“AMQChannel.SimpleBlockingRpcContinuation connStartBlocker = new AMQChannel.SimpleBlockingRpcContinuation();”这句代码，从命名上来说像是rpc, 其实这么理解也没错。RabbitMQ-Client这个版本（3.5.3）的客户端与broker端的通信是采用java原生socket.当然后面也改成了NIO，这个自然是后话。RabbitMQ-Client程序中会对各种帧进行处理，处理的方式也不是单一化的，这里举Connection.Start这个类型的报文做分析。当broker发送Connection.Start至client端，client收到之后进行处理（MainLoop线程中），然后将此报文存入SimpleBlockingRpcContinuation中，照着SimpleBlockingRpcContinuation深究下去，其就是一个容量为1的BlockingQueue,也就是当MainLoop主导的线程将收到的Connection.Start存入其中，然后AMQConnction类的start()线程在等待（start()方法下面的代码）：
connStart = (AMQP.Connection.Start) connStartBlocker.getReply(HANDSHAKE_TIMEOUT/2).getMethod();
然后继续处理。这看上去也算是个rpc，等待别的线程（这个线程同样在等待broker的返回）处理完毕。
AMQCommand(这个之后会讲到), 下面的“_channel0.enqueueRpc(connStartBlocker)”将这个rpc任务放入Channel中，如果深入代码看的话，channel中当前至多只能enqueue一个rpc，如果当前的rpc没有处理完再enqueue的话会被阻塞（wait()）直到处理完当前的rpc才能enqueue下一个rpc。
try {
// The following two lines are akin to AMQChannel's
// transmit() method for this pseudo-RPC.
_frameHandler.setTimeout(HANDSHAKE_TIMEOUT);
_frameHandler.sendHeader();
} catch (IOException ioe) {
_frameHandler.close();
throw ioe;
}
接下来“_frameHandler.sendHeader()”主要是发送Protocol-Header 0-9-1帧（可参考下图），这个客户端与broker建立连接的AMQP协议的第一帧，帧中的内容包括AMQP的版本号。这里发_frameHandler就是前面Connection提到的SocketFrameHandler对象，我们来看看sendHeader()做了什么：
//本段代码在SocketFrameHandler类中
public void sendHeader(int major, int minor, int revision) throws IOException {
synchronized (_outputStream) {
_outputStream.write("AMQP".getBytes("US-ASCII"));
_outputStream.write(0);
_outputStream.write(major);
_outputStream.write(minor);
_outputStream.write(revision);
_outputStream.flush();
}
}
public void sendHeader() throws IOException {
sendHeader(AMQP.PROTOCOL.MAJOR, AMQP.PROTOCOL.MINOR, AMQP.PROTOCOL.REVISION);
}
// start the main loop going
MainLoop loop = new MainLoop();
final String name = "AMQP Connection " + getHostAddress() + ":" + getPort();
mainLoopThread = Environment.newThread(threadFactory, loop, name);
mainLoopThread.start();
// after this point clear-up of MainLoop is triggered by closing the frameHandler.
下面就是最重要的MainLoop线程了。这里先跳过，接下去看看start()方法，之后就是Connection.Start/.StartOk, Connection.Tune/.TuneOk, Connection.Open/.OpenOk的来回negotiation，以及设置channelMax, frameMax和heartbeat的参数值。当然在设置frameMax之前还初始化了ChannelManager，至于ChannelManager可以简单的理解为管理Channel的一个类，具体实现细节可以参考（[[三]RabbitMQ-客户端源码之ChannelManager][RabbitMQ-_ChannelManager]）
AMQP.Connection.Start connStart = null;
AMQP.Connection.Tune connTune = null;
try {
connStart =
(AMQP.Connection.Start) connStartBlocker.getReply(HANDSHAKE_TIMEOUT/2).getMethod();
_serverProperties = Collections.unmodifiableMap(connStart.getServerProperties());
Version serverVersion =
new Version(connStart.getVersionMajor(),
connStart.getVersionMinor());
if (!Version.checkVersion(clientVersion, serverVersion)) {
throw new ProtocolVersionMismatchException(clientVersion,
serverVersion);
}
String[] mechanisms = connStart.getMechanisms().toString().split(" ");
SaslMechanism sm = this.saslConfig.getSaslMechanism(mechanisms);
if (sm == null) {
throw new IOException("No compatible authentication mechanism found - " +
"server offered [" + connStart.getMechanisms() + "]");
}
LongString challenge = null;
LongString response = sm.handleChallenge(null, this.username, this.password);
do {
Method method = (challenge == null)
? new AMQP.Connection.StartOk.Builder()
.clientProperties(_clientProperties)
.mechanism(sm.getName())
.response(response)
.build()
: new AMQP.Connection.SecureOk.Builder().response(response).build();
try {
Method serverResponse = _channel0.rpc(method, HANDSHAKE_TIMEOUT/2).getMethod();
if (serverResponse instanceof AMQP.Connection.Tune) {
connTune = (AMQP.Connection.Tune) serverResponse;
} else {
challenge = ((AMQP.Connection.Secure) serverResponse).getChallenge();
response = sm.handleChallenge(challenge, this.username, this.password);
}
} catch (ShutdownSignalException e) {
Method shutdownMethod = e.getReason();
if (shutdownMethod instanceof AMQP.Connection.Close) {
AMQP.Connection.Close shutdownClose = (AMQP.Connection.Close) shutdownMethod;
if (shutdownClose.getReplyCode() == AMQP.ACCESS_REFUSED) {
throw new AuthenticationFailureException(shutdownClose.getReplyText());
}
}
throw new PossibleAuthenticationFailureException(e);
}
} while (connTune == null);
} catch (TimeoutException te) {
_frameHandler.close();
throw te;
} catch (ShutdownSignalException sse) {
_frameHandler.close();
throw AMQChannel.wrap(sse);
} catch(IOException ioe) {
_frameHandler.close();
throw ioe;
}
try {
int channelMax =
negotiateChannelMax(this.requestedChannelMax,
connTune.getChannelMax());
_channelManager = instantiateChannelManager(channelMax, threadFactory);
int frameMax =
negotiatedMaxValue(this.requestedFrameMax,
connTune.getFrameMax());
this._frameMax = frameMax;
int heartbeat =
negotiatedMaxValue(this.requestedHeartbeat,
connTune.getHeartbeat());
setHeartbeat(heartbeat);
_channel0.transmit(new AMQP.Connection.TuneOk.Builder()
.channelMax(channelMax)
.frameMax(frameMax)
.heartbeat(heartbeat)
.build());
_channel0.exnWrappingRpc(new AMQP.Connection.Open.Builder()
.virtualHost(_virtualHost)
.build());
} catch (IOException ioe) {
_heartbeatSender.shutdown();
_frameHandler.close();
throw ioe;
} catch (ShutdownSignalException sse) {
_heartbeatSender.shutdown();
_frameHandler.close();
throw AMQChannel.wrap(sse);
}
// We can now respond to errors having finished tailoring the connection
this._inConnectionNegotiation = false;
return;
}
接着回顾MainLoop, 在start()方法中关于MainLoop的代码主要有：
// start the main loop going
MainLoop loop = new MainLoop();
final String name = "AMQP Connection " + getHostAddress() + ":" + getPort();
mainLoopThread = Environment.newThread(threadFactory, loop, name);
mainLoopThread.start();
// after this point clear-up of MainLoop is triggered by closing the frameHandler.
这段代码主要是初始化MainLoop线程对象，然后让其运行。没有什么特别之处，而特别之处在于MainLoop本身。
MainLoop类是AMQConnection类的私有内部类：
private class MainLoop implements Runnable {
/**
* Channel reader thread main loop. Reads a frame, and if it is
* not a heartbeat frame, dispatches it to the channel it refers to.
* Continues running until the "running" flag is set false by
* shutdown().
*/
public void run() {
try {
while (_running) {
Frame frame = _frameHandler.readFrame();
if (frame != null) {
_missedHeartbeats = 0;
if (frame.type == AMQP.FRAME_HEARTBEAT) {
// Ignore it: we've already just reset the heartbeat counter.
} else {
if (frame.channel == 0) { // the special channel
_channel0.handleFrame(frame);
} else {
if (isOpen()) {
// If we're still _running, but not isOpen(), then we
// must be quiescing, which means any inbound frames
// for non-zero channels (and any inbound commands on
// channel zero that aren't Connection.CloseOk) must
// be discarded.
ChannelManager cm = _channelManager;
if (cm != null) {
cm.getChannel(frame.channel).handleFrame(frame);
}
}
}
}
} else {
// Socket timeout waiting for a frame.
// Maybe missed heartbeat.
handleSocketTimeout();
}
}
} catch (EOFException ex) {
if (!_brokerInitiatedShutdown)
shutdown(null, false, ex, true);
} catch (Throwable ex) {
_exceptionHandler.handleUnexpectedConnectionDriverException(AMQConnection.this,
ex);
shutdown(null, false, ex, true);
} finally {
// Finally, shut down our underlying data connection.
_frameHandler.close();
_appContinuation.set(null);
notifyListeners();
}
}
}
MainLoop线程主要用来处理通信帧（Frame，有关Frame的细节将会在（[[四]RabbitMQ-客户端源码之Frame][RabbitMQ-_Frame]）中陈述）的，可以看到当AMQConnection调用start()方法后，_isrunning就设置为true，那么线程一直在运行（while(true)）。
MainLoop线程当读取到通信帧之后，判断是否是心跳帧，如果是则忽略继续监听。如果是其他帧，则判断其frame.channel值是否为0，frame.channel值为0代表的是特殊帧，这些特殊帧是和Connection有关的，而不是和Channel有关的（上面代码里的frame.channel就是Channel里的channel number, 一般Connection类型的帧的channel number为0，而其余Channel类别帧的channel number大于0。）
这里就分channel_number=0和channel_number !=0分别进行处理。
当channel_number=0即frame.channel=0则直接调用_channel0的handleFrame方法。
这个_channel0是在AMQConnection类中创建的私有变量：
private final AMQChannel _channel0 = new AMQChannel(this, 0) {
@Override public boolean processAsync(Command c) throws IOException {
return getConnection().processControlCommand(c);
}
};
调用AMQChannel的handleFrame方法（有关AMQChannel的更多实现细节可以参考：（[[五]RabbitMQ-客户端源码之AMQChannel][RabbitMQ-_AMQChannel]））：
public void handleFrame(Frame frame) throws IOException {
AMQCommand command = _command;
if (command.handleFrame(frame)) { // a complete command has rolled off the assembly line
_command = new AMQCommand(); // prepare for the next one
handleCompleteInboundCommand(command);
}
}
对于channel number为0的帧，AMQCommand的handleFrame方法都是返回true.（有关AMQCommand的实现细节可以参考：（[[六]RabbitMQ-客户端源码之AMQCommand][RabbitMQ-_AMQCommand]））
进而调用AMQChannel的handleCompleteInboundCommand(command)方法：
public void handleCompleteInboundCommand(AMQCommand command) throws IOException {
// First, offer the command to the asynchronous-command
// handling mechanism, which gets to act as a filter on the
// incoming command stream. If processAsync() returns true,
// the command has been dealt with by the filter and so should
// not be processed further. It will return true for
// asynchronous commands (deliveries/returns/other events),
// and false for commands that should be passed on to some
// waiting RPC continuation.
if (!processAsync(command)) {
// The filter decided not to handle/consume the command,
// so it must be some reply to an earlier RPC.
nextOutstandingRpc().handleCommand(command);
markRpcFinished();
}
}
进而调用AMQChannel的processAsync方法。这个方法在AMQChannel类中是一个抽象方法，而观察AMQConnection中的AMQChannel _channel0私有变量其正好实现了这个方法：
/** The special channel 0 (not managed by the_channelManager) */
private final AMQChannel _channel0 = new AMQChannel(this, 0) {
@Override public boolean processAsync(Command c) throws IOException {
return getConnection().processControlCommand(c);
}
};
ChannelN中也实现了processAsync方法。（有关ChannelN的实现细节可以参考：（[[八]RabbitMQ-客户端源码之ChannelN][RabbitMQ-_ChannelN]））
进而调用了AMQConnection的processControlCommand方法：
/**
* Handles incoming control commands on channel zero.
* @see ChannelN#processAsync
*/
@SuppressWarnings("unused")
public boolean processControlCommand(Command c) throws IOException
{
// Similar trick to ChannelN.processAsync used here, except
// we're interested in whole-connection quiescing.
// See the detailed comments in ChannelN.processAsync.
Method method = c.getMethod();
if (isOpen()) {
if (method instanceof AMQP.Connection.Close) {
handleConnectionClose(c);
return true;
} else if (method instanceof AMQP.Connection.Blocked) {
AMQP.Connection.Blocked blocked = (AMQP.Connection.Blocked) method;
try {
for (BlockedListener l : this.blockedListeners) {
l.handleBlocked(blocked.getReason());
}
} catch (Throwable ex) {
getExceptionHandler().handleBlockedListenerException(this, ex);
}
return true;
} else if (method instanceof AMQP.Connection.Unblocked) {
try {
for (BlockedListener l : this.blockedListeners) {
l.handleUnblocked();
}
} catch (Throwable ex) {
getExceptionHandler().handleBlockedListenerException(this, ex);
}
return true;
} else {
return false;
}
} else {
if (method instanceof AMQP.Connection.Close) {
// Already shutting down, so just send back a CloseOk.
try {
_channel0.quiescingTransmit(new AMQP.Connection.CloseOk.Builder().build());
} catch (IOException _e) { } // ignore
return true;
} else if (method instanceof AMQP.Connection.CloseOk) {
// It's our final "RPC". Time to shut down.
_running = false;
// If Close was sent from within the MainLoop we
// will not have a continuation to return to, so
// we treat this as processed in that case.
return !_channel0.isOutstandingRpc();
} else { // Ignore all others.
return true;
}
}
}
这个方法是用来处理AMQP控制命令的：Connection.Close/CloseOk, Connection.Blocked/.Unblocked。正常情况下（比如Connection.Start/.StartOk）直接返回false。
这样就会运行到 nextOutstandingRpc().handleCommand(command);这句代码，意思就是将从broker接受到的AMQCommand对象存入RpcContinuation对象，确切的来说是SimpleBlockingRpcContinuation这个对象中，更确切的来说是存放到容量为1的BlockingQueue中，等待其余的线程来“take()”。有关RpcContinuation或者SimpleBlockingRpcContinuation细节可以参考：[[五]RabbitMQ-客户端源码之AMQChannel][RabbitMQ-_AMQChannel]。
我们假设有一个可靠的面向流的网络传输层（TCP/IP或相当）。在单个套接字连接中，可以存在多个独立控制线程，这些称之为通道。每个帧都使用通道编号来编号。通过交织他们的帧，不同的通道共享连接。对于给定的通道，帧运行在一个严格的序列，这样可以用来驱动一个协议解析器（通常是一个状态机）。
当channel_number!=0则需要从ChannelManager中根据channel number找出相应的AMQChannel再调用handleFrame方法处理。
这里的ChannelManager从何而来？
这里就还是要到AMQChannel的start()方法来看，有这么一句：_channelManager = instantiateChannelManager(channelMax, threadFactory);
//AMQConnection类中
protected ChannelManager instantiateChannelManager(int channelMax, ThreadFactory threadFactory) {
return new ChannelManager(this._workService, channelMax, threadFactory);
}
//ChannelManager的构造方法
public ChannelManager(ConsumerWorkService workService, int channelMax, ThreadFactory threadFactory) {
if (channelMax == 0) {
// The framing encoding only allows for unsigned 16-bit integers
// for the channel number
channelMax = (1 << 16) - 1;
}
_channelMax = channelMax;
channelNumberAllocator = new IntAllocator(1, channelMax);
this.workService = workService;
this.threadFactory = threadFactory;
}
ChannelManager构造方法中的ConsumerWorkService参数就是AMQConnection中start()方法第一行代码初始化的ConsumerWorkService对象。
有关ChannelManager的实现细节可以参考：（[[三]RabbitMQ-客户端源码之ChannelManager][RabbitMQ-_ChannelManager]）
当channel number等于0的时候是调用AMQChannel，也可以说是AQMConnection的内部成员变量AMQChannel _channel0来处理。
版权归原作者 Java面试大全 所有， 如有侵权，请联系我们删除。