当前位置: 首页 > >

hadoop rpc服务端初始化和调用过程详解

发布时间:

为什么80%的码农都做不了架构师?>>> ??



本文主要描述了hadoop rpc服务端的初始化和调用过程,相比客户端的初始化,rpc服务端感觉会简单点,但是调用过程却比客户端复杂一些。本文还是以namenode为例,namenode会在执行main方法的时候,创建一个namenode实例,及完成一系列的初始化过程,其中就包括了rpc的初始化过程。



rpc服务端的初始化

上面已经提到我们这里主要借用了namenode的远程服务,先来看看相关代码:


public class NameNode implements NameNodeStatusMXBean {
public static void main(String argv[]) throws Exception {
NameNode namenode = createNameNode(argv, null);
}

protected NameNode(Configuration conf, NamenodeRole role)throws IOException {
initialize(conf);
}

protected void initialize(Configuration conf) throws IOException {
rpcServer = createRpcServer(conf);

startCommonServices(conf); //相当重要
}

protected NameNodeRpcServer createRpcServer(Configuration conf)throws IOException {
return new NameNodeRpcServer(conf, this);
}
}

我们的linux的终端执行hadoop的启动命令的时候,最终的命令是调用NameNode的main方法,所以我们追踪代码的切入点是NameNode的main方法,方法比较简单,就是调用NameNode的构造函数创建一个NameNode,然后执行初始化方法initialize,这个方法相对来说,是我们关注的重点,包括rpc服务在内的初始化操作都放在这个方法里面。特定于rpc,他执行了两个相关的方法createRpcServer和startCommonServices,第一个方法见名思意,不多说,先简单介绍下后面的方法,该方法的作用就是启动namenode的rpc服务,稍后我给出代码。好的,从上面的代码可以看到,我们的rpcServer功能都放在了类NameNodeRpcServer里面,现在让我们来看看这个类里面相关的代码:


class NameNodeRpcServer implements NamenodeProtocols {
public NameNodeRpcServer(Configuration conf, NameNode nn)
throws IOException {
RPC.setProtocolEngine(conf, ClientNamenodeProtocolPB.class,
ProtobufRpcEngine.class);

ClientNamenodeProtocolServerSideTranslatorPB
clientProtocolServerTranslator =
new ClientNamenodeProtocolServerSideTranslatorPB(this);
BlockingService clientNNPbService = ClientNamenodeProtocol.
newReflectiveBlockingService(clientProtocolServerTranslator);

InetSocketAddress rpcAddr = nn.getRpcServerAddress(conf); // fs.defaultFS
String bindHost = nn.getRpcServerBindHost(conf);
if (bindHost == null) {
bindHost = rpcAddr.getHostName();
}
LOG.info("RPC server is binding to " + bindHost + ":" + rpcAddr.getPort());

this.clientRpcServer = new RPC.Builder(conf)
.setProtocol(
org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolPB.class)
.setInstance(clientNNPbService).setBindAddress(bindHost)
.setPort(rpcAddr.getPort()).setNumHandlers(handlerCount)
.setVerbose(false)
.setSecretManager(namesystem.getDelegationTokenSecretManager()).build();

// Add all the RPC protocols that the namenode implements
DFSUtil.addPBProtocol(conf, HAServiceProtocolPB.class, haPbService,
clientRpcServer);
DFSUtil.addPBProtocol(conf, NamenodeProtocolPB.class, NNPbService,
clientRpcServer);
DFSUtil.addPBProtocol(conf, DatanodeProtocolPB.class, dnProtoPbService,
clientRpcServer);
}
}

在NameNodeRpcServer的构造函数里面最重要的一件事情是实例化clientRpcServer,这里面我最想说明的是,NameNode宣称自己实现了三个协议:ClientProtocol、DatanodeProtocol和NamenodeProtocol,在服务端的实现基本上就靠ClientNamenodeProtocolServerSideTranslatorPB之类的类型了,特别在实例化ClientNamenodeProtocolServerSideTranslatorPB的时候有传入一个形参,这个形参就是NameNodeRpcServer实例,看代码:


public ClientNamenodeProtocolServerSideTranslatorPB(ClientProtocol server)
throws IOException {
this.server = server;
}

@Override
public GetBlockLocationsResponseProto getBlockLocations(
RpcController controller, GetBlockLocationsRequestProto req)
throws ServiceException {
try {
LocatedBlocks b = server.getBlockLocations(req.getSrc(), req.getOffset(),
req.getLength());
Builder builder = GetBlockLocationsResponseProto
.newBuilder();
if (b != null) {
builder.setLocations(PBHelper.convert(b)).build();
}
return builder.build();
} catch (IOException e) {
throw new ServiceException(e);
}
}

上面代码中的getBlockLocations也一定程度上说明了刚才的观点。


现在让我们回过头看看NameNode中initialize方法中执行的startCommonServices方法,这个方法用来启动clientRpcServer下面的线程,包括listener,handler、response,具体看代码:?


public class NameNode implements NameNodeStatusMXBean {
private void startCommonServices(Configuration conf) throws IOException {
rpcServer.start();
}
}

class NameNodeRpcServer implements NamenodeProtocols {
void start() {
clientRpcServer.start();
if (serviceRpcServer != null) {
serviceRpcServer.start();
}
}
}

public abstract class Server {
public synchronized void start() {
responder.start();
listener.start();
handlers = new Handler[handlerCount];

for (int i = 0; i < handlerCount; i++) {
handlers[i] = new Handler(i);
handlers[i].start();
}
}
}

代码看到这里,启动过程中rpc相关的代码就结束了。



rpc服务端的调用过程

现在让我们来看看rpc被调用的过程,先来认识下Server的关键结构:


public abstract class Server {
private Listener listener = null;
private Responder responder = null;
private Handler[] handlers = null;

private class Responder extends Thread {

}

private class Listener extends Thread {

}

private class Handler extends Thread {

}
}

在初始化的时候,就启动listener、responder和handlers下面的所有线程。


其中listener线程里面启动了一个socker服务,专门用来接受客户端的请求,handler下面的线程用来处理具体的请求,responder写请求结果,具体过程可以看下下面的代码:


public abstract class Server {
private Listener listener = null;
private Responder responder = null;
private Handler[] handlers = null;

private class Listener extends Thread {
public Listener() throws IOException {
address = new InetSocketAddress(bindAddress, port);
// Create a new server socket and set to non blocking mode
acceptChannel = ServerSocketChannel.open();
acceptChannel.configureBlocking(false);

// Bind the server socket to the local host and port
bind(acceptChannel.socket(), address, backlogLength, conf, portRangeConfig);
port = acceptChannel.socket().getLocalPort(); //Could be an ephemeral port
// create a selector;
selector= Selector.open();
readers = new Reader[readThreads];
for (int i = 0; i < readThreads; i++) {
Reader reader = new Reader(
"Socket Reader #" + (i + 1) + " for port " + port);
readers[i] = reader;
reader.start();
}

// Register accepts on the server socket with the selector.
acceptChannel.register(selector, SelectionKey.OP_ACCEPT);
this.setName("IPC Server listener on " + port);
this.setDaemon(true);
}

public void run() {
while (running) {
doAccept(key);
}
}

void doAccept(SelectionKey key) throws InterruptedException, IOException, OutOfMemoryError {
Reader reader = getReader();
Connection c = connectionManager.register(channel);
key.attach(c); // so closeCurrentConnection can get the object
reader.addConnection(c);
}

private class Reader extends Thread {
public void run() {
doRunLoop();
}

private synchronized void doRunLoop() {
while (running) {
Connection conn = pendingConnections.take();
conn.channel.register(readSelector, SelectionKey.OP_READ, conn);
}
readSelector.select();
doRead(key);
}

void doRead(SelectionKey key) throws InterruptedException {
Connection c = (Connection)key.attachment();
count = c.readAndProcess();
}
}
}

public class Connection {
public int readAndProcess(){
processOneRpc(data.array());
}

private void processOneRpc(byte[] buf){
processRpcRequest(header, dis);
}

private void processRpcRequest(RpcRequestHeaderProto header,
DataInputStream dis) throws WrappedRpcServerException,
InterruptedException {
Call call = new Call(header.getCallId(), header.getRetryCount(),
rpcRequest, this, ProtoUtil.convert(header.getRpcKind()), header
.getClientId().toByteArray());
callQueue.put(call);
}
}

private class Handler extends Thread {
public void run() {
final Call call = callQueue.take();
value = call(call.rpcKind, call.connection.protocolName, call.rpcRequest,
call.timestamp);

setupResponse(buf, call, returnStatus, detailedErr,
value, errorClass, error);

responder.doRespond(call);
}
}

private class Responder extends Thread {
void doRespond(Call call) throws IOException {
processResponse(call.connection.responseQueue, true);
}

private boolean processResponse(LinkedList responseQueue,
boolean inHandler) throws IOException {
int numBytes = channelWrite(channel, call.rpcResponse);

done = true;
}
}
}

这里给出了一个比较完整版Server的rpc调用过程,从listener都构造函数开始,在他的构造函数中起了几个reader线程,当监听器收到访问请求的时候,由reader请请求中读取数据,reader中实际上调用的是connection的readAndProcess方法,在这个方法中,会往RPC server中的callQueue添加call对象,之后,handler这个家伙从队列中取出当前call,具体的处理过程,用到了Server类的call方法,这地方有些玄机,仔细跟过代码的人才知道,因为server的实例类不再是org.apache.hadoop.ipc.Server,而是Protobuf的一个实现类,org.apache.hadoop.ipc.RPC.Server,而且call方法是被重写过的,代码如下:


@Override
public Writable call(RPC.RpcKind rpcKind, String protocol,
Writable rpcRequest, long receiveTime) throws Exception {
return getRpcInvoker(rpcKind).call(this, protocol, rpcRequest,
receiveTime);
}

继续追踪下,差不多就可以到底了:


public class ProtobufRpcEngine implements RpcEngine {
public static class Server extends RPC.Server {
static class ProtoBufRpcInvoker implements RpcInvoker {
public Writable call(RPC.Server server, String protocol,
Writable writableRequest, long receiveTime) throws Exception {
ProtoClassProtoImpl protocolImpl = getProtocolImpl(server, protoName,clientVersion);
BlockingService service = (BlockingService) protocolImpl.protocolImpl;

result = service.callBlockingMethod(methodDescriptor, null, param);

return new RpcResponseWrapper(result);
}
}
}

这部分的代码也正是hadoop rpc与protobuf结合的地方,这地方在补充一点,protbufImpl就是NameNodeRpcServer初始化的时候,已经准备了,而且看懂ProtoBufRpcInvoker下的call方法,确实也是需要结合NameNodeRpcServer初始化过程来理解的。我朦朦胧胧的懂了。而且这地方的深入会让你看到一些本质的东西,举例的话,你会跟踪到ClientNamenodeProtocolServerSideTranslatorPB,然后是NameNodeRpcServer,再然后是FSNamesystem,最后你发现,服务端对文件系统的操作出自FSNamesystem。


继续回到handler中的run方法,call方法调用完了,就轮到Responder处理返回结果了。


整个过程就是这样了,需要说明点,上面写都东西有些可以确认没问题了,有些是个人结合书的一些总结,不一定对,仅供参考。








转载于:https://my.oschina.net/psuyun/blog/372652






相关资源:TongWeb技术文档合集



友情链接: 传奇百科网 招聘百科网 非凡百科网 游艇百科网 口红百科网 创业百科网 软木百科网