Dubbo 缺省协议采用单一长连接和 NIO 异步通讯,适合于小数据量大并发的服务调用,以及服务消费者机器数远大于服务提供者机器数的情况。
反之,Dubbo 缺省协议不适合传送大数据量的服务,比如传文件,传视频等,除非请求量很低。
缺省协议,使用基于netty3.2.5+hessian3.2.1交互。
dubbo协议中对字节流的处理在com.alibaba.dubbo.remoting.exchange.codec.ExchangeCodec类中,包含了对request请求的编码和解码,response响应的编码和解码。
dubbo协议采用固定长度的消息头(16字节)和不定长度的消息体来进行数据传输,消息头定义了一些通讯框架netty在IO线程处理时需要的信息。具体dubbo协议的报文格式如下:
消息头报文格式:
com.alibaba.dubbo.remoting.exchange.Response
body length:消息体 body 长度, int 类型,即记录Body Content有多少个字节。
消息头的编码:
类路径:com.alibaba.dubbo.remoting.exchange.codec.ExchangeCodec
protected void encodeRequest(Channel channel, ChannelBuffer buffer, Request req) throws IOException {
Serialization serialization = getSerialization(channel);
// header.
byte[] header = new byte[HEADER_LENGTH];//16个字节
// set magic number.
Bytes.short2bytes(MAGIC, header); //2个字节的包头
// set request and serialization flag.
header[2] = (byte) (FLAG_REQUEST | serialization.getContentTypeId());
if (req.isTwoWay()) header[2] |= FLAG_TWOWAY;
if (req.isEvent()) header[2] |= FLAG_EVENT;
// set request id.
Bytes.long2bytes(req.getId(), header, 4); //invoke ID 8位
// encode request data.
int savedWriteIndex = buffer.writerIndex();
buffer.writerIndex(savedWriteIndex + HEADER_LENGTH);
ChannelBufferOutputStream bos = new ChannelBufferOutputStream(buffer);
ObjectOutput out = serialization.serialize(channel.getUrl(), bos);
if (req.isEvent()) {
encodeEventData(channel, out, req.getData());
} else {
encodeRequestData(channel, out, req.getData());
}
out.flushBuffer();
bos.flush();
bos.close();
int len = bos.writtenBytes();
checkPayload(channel, len);
Bytes.int2bytes(len, header, 12);
// write
buffer.writerIndex(savedWriteIndex);
buffer.writeBytes(header); // write header.
buffer.writerIndex(savedWriteIndex + HEADER_LENGTH + len);
}
消息头的解码:
protected Object decode(Channel channel, ChannelBuffer buffer, int readable, byte[] header) throws IOException {
// check magic number.
if (readable > 0 && header[0] != MAGIC_HIGH
|| readable > 1 && header[1] != MAGIC_LOW) {
int length = header.length;
if (header.length < readable) {
header = Bytes.copyOf(header, readable);
buffer.readBytes(header, length, readable - length);
}
for (int i = 1; i < header.length - 1; i++) {
if (header[i] == MAGIC_HIGH && header[i + 1] == MAGIC_LOW) {
buffer.readerIndex(buffer.readerIndex() - header.length + i);
header = Bytes.copyOf(header, i);
break;
}
}
return super.decode(channel, buffer, readable, header);
}
// check length.
if (readable < HEADER_LENGTH) {
return DecodeResult.NEED_MORE_INPUT;
}
// get data length.
int len = Bytes.bytes2int(header, 12);
checkPayload(channel, len);
int tt = len + HEADER_LENGTH;
if (readable < tt) {
return DecodeResult.NEED_MORE_INPUT;
}
// limit input stream.
ChannelBufferInputStream is = new ChannelBufferInputStream(buffer, len);
try {
return decodeBody(channel, is, header);
} finally {
if (is.available() > 0) {
try {
if (logger.isWarnEnabled()) {
logger.warn("Skip input stream " + is.available());
}
StreamUtils.skipUnusedStream(is);
} catch (IOException e) {
logger.warn(e.getMessage(), e);
}
}
}
}
body:
消息体解析:
类路径:com.alibaba.dubbo.rpc.protocol.dubbo.DubboCodec
编码request
@Override
protected void encodeRequestData(Channel channel, ObjectOutput out, Object data) throws IOException {
RpcInvocation inv = (RpcInvocation) data;
out.writeUTF(inv.getAttachment(Constants.DUBBO_VERSION_KEY, DUBBO_VERSION));
out.writeUTF(inv.getAttachment(Constants.PATH_KEY));
out.writeUTF(inv.getAttachment(Constants.VERSION_KEY));
out.writeUTF(inv.getMethodName());
out.writeUTF(ReflectUtils.getDesc(inv.getParameterTypes()));
Object[] args = inv.getArguments();
if (args != null)
for (int i = 0; i < args.length; i++) {
out.writeObject(encodeInvocationArgument(channel, inv, i));
}
out.writeObject(inv.getAttachments());
}
解码request:
类路径:DecodeableRpcInvocation
public Object decode(Channel channel, InputStream input) throws IOException {
ObjectInput in = CodecSupport.getSerialization(channel.getUrl(), serializationType)
.deserialize(channel.getUrl(), input);
setAttachment(Constants.DUBBO_VERSION_KEY, in.readUTF());
setAttachment(Constants.PATH_KEY, in.readUTF());
setAttachment(Constants.VERSION_KEY, in.readUTF());
setMethodName(in.readUTF());
try {
Object[] args;
Class>[] pts;
String desc = in.readUTF();
if (desc.length() == 0) {
pts = DubboCodec.EMPTY_CLASS_ARRAY;
args = DubboCodec.EMPTY_OBJECT_ARRAY;
} else {
pts = ReflectUtils.desc2classArray(desc);
args = new Object[pts.length];
for (int i = 0; i < args.length; i++) {
try {
args[i] = in.readObject(pts[i]);
} catch (Exception e) {
if (log.isWarnEnabled()) {
log.warn("Decode argument failed: " + e.getMessage(), e);
}
}
}
}
setParameterTypes(pts);
Map map = (Map) in.readObject(Map.class);
if (map != null && map.size() > 0) {
Map attachment = getAttachments();
if (attachment == null) {
attachment = new HashMap();
}
attachment.putAll(map);
setAttachments(attachment);
}
//decode argument ,may be callback
for (int i = 0; i < args.length; i++) {
args[i] = decodeInvocationArgument(channel, this, pts, i, args[i]);
}
setArguments(args);
} catch (ClassNotFoundException e) {
throw new IOException(StringUtils.toString("Read invocation data failed.", e));
}
return this;
}
编码response
类路径:com.alibaba.dubbo.rpc.protocol.dubbo.DubboCodec
@Override
protected void encodeResponseData(Channel channel, ObjectOutput out, Object data) throws IOException {
Result result = (Result) data;
Throwable th = result.getException();
if (th == null) {
Object ret = result.getValue();
if (ret == null) {
out.writeByte(RESPONSE_NULL_VALUE);
} else {
out.writeByte(RESPONSE_VALUE);
out.writeObject(ret);
}
} else {
out.writeByte(RESPONSE_WITH_EXCEPTION);
out.writeObject(th);
}
}
解码response
类路径:com.alibaba.dubbo.rpc.protocol.dubbo.DecodeableRpcResult
public Object decode(Channel channel, InputStream input) throws IOException {
ObjectInput in = CodecSupport.getSerialization(channel.getUrl(), serializationType)
.deserialize(channel.getUrl(), input);
byte flag = in.readByte();
switch (flag) {
case DubboCodec.RESPONSE_NULL_VALUE:
break;
case DubboCodec.RESPONSE_VALUE:
try {
Type[] returnType = RpcUtils.getReturnTypes(invocation);
setValue(returnType == null || returnType.length == 0 ? in.readObject() :
(returnType.length == 1 ? in.readObject((Class>) returnType[0])
: in.readObject((Class>) returnType[0], returnType[1])));
} catch (ClassNotFoundException e) {
throw new IOException(StringUtils.toString("Read response data failed.", e));
}
break;
case DubboCodec.RESPONSE_WITH_EXCEPTION:
try {
Object obj = in.readObject();
if (obj instanceof Throwable == false)
throw new IOException("Response data error, expect Throwable, but get " + obj);
setException((Throwable) obj);
} catch (ClassNotFoundException e) {
throw new IOException(StringUtils.toString("Read response data failed.", e));
}
break;
default:
throw new IOException("Unknown result flag, expect '0' '1' '2', get " + flag);
}
return this;
}
dubbo协议的问题:
参考: https://www.cnblogs.com/lishijia/p/5492847.html
1.为什么要消费者比提供者个数多?
因为Dubbo协议采用单一长连接,假设网络为千兆网卡(1024Mbit=128MByte),根据测试经验数据每条连接最多只能压满7MByte(不同的环境可能不一样,供参考),理论上1个服务提供者需要20个服务消费者才能压满网卡。
2.为什么不能传大数据量?
因dubbo协议采用单一长连接,如果每次请求的数据包大小为500KByte,假设网络为千兆网卡(1024Mbit=128MByte),每条连接最大7MByte(不同的环境可能不一样,供参考),
单个服务提供者的TPS(每秒处理事务数)最大为:128MByte / 500KByte = 262。
单个消费者调用单个服务提供者的TPS(每秒处理事务数)最大为:7MByte / 500KByte = 14。
如果能接受,可以考虑使用,否则网络将成为瓶颈。
3.为什么采用异步单一长连接?
因为服务的现状大都是服务提供者少,通常只有几台机器,
而服务的消费者多,可能整个网站都在访问该服务,
比如Morgan的提供者只有6台提供者,却有上百台消费者,每天有1.5亿次调用,
如果采用常规的hessian服务,服务提供者很容易就被压跨,
通过单一连接,保证单一消费者不会压死提供者,
长连接,减少连接握手验证等,
并使用异步IO,复用线程池,防止C10K(即单机1万个并发连接问题)问题。
4.单一长连接如何解决多线程并发请求?
socket通信的双方发送和接受数据不会被其它(线程)干扰,这种干扰不同于数数据包的“粘包问题”。其实说白了就相当于电话线路的场景。
粘包的解决思路:定义一个定长的数据包头,其中包含了完整数据包的长度,以此来完成服务器端拆包工作。
dubbo的解决思路:给包头添加一个标识ID,服务端响应请求时也要携带这个id,供客户端多线程领取对应的响应数据提供线索。
调用流程如下:
3. 向专门存放调用信息的全局ConcurrentHashMap里面put(ID, object)
4.将ID和打包的方法调用信息封装成一对象connRequest,使用IoSession.write(connRequest)异步发送出去
5.当前线程再使用callback的get()方法试图获取远程返回的结果,在get()内部,则使用synchronized获取回调对象callback的锁,再先检测是否已经获取到结果,如果没有,然后调用callback的wait()方法,释放callback上的锁,让当前线程处于等待状态。
6.服务端接收到请求并处理后,将结果(此结果中包含了前面的ID,即回传)发送给客户端,客户端socket连接上专门监听消息的线程收到消息,分析结果,取到ID,再从前面的ConcurrentHashMap里面get(ID),从而找到callback,将方法调用结果设置到callback对象里。
7.监听线程接着使用synchronized获取回调对象callback的锁(因为前面调用过wait(),那个线程已释放callback的锁了),再notifyAll(),唤醒前面处于等待状态的线程继续执行(callback的get()方法继续执行就能拿到调用结果了),至此,整个过程结束。