Hadoop2源码分析－YARN RPC 示例介绍

1.概述

　　之前在《Hadoop2源码分析－RPC探索实战》一文当中介绍了Hadoop的RPC机制，今天给大家分享关于YARN的RPC的机制。下面是今天的分享目录：

• YARN的RPC介绍
• YARN的RPC示例
• 截图预览

　　下面开始今天的内容分享。

2.YARN的RPC介绍

　　我们知道在Hadoop的RPC当中，其主要由RPC，Client及Server这三个大类组成，分别实现对外提供编程接口、客户端实现及服务端实现。如下图所示：

　　图中是Hadoop的RPC的一个类的关系图，大家可以到《Hadoop2源码分析－RPC探索实战》一文中，通过代码示例去理解他们之间的关系，这里就不多做赘述了。接下来，我们去看Yarn的RPC。

　　Yarn对外提供的是YarnRPC这个类，这是一个抽象类，通过阅读YarnRPC的源码可以知道，实际的实现由参数yarn.ipc.rpc.class设定，默认情况下，其值为：org.apache.hadoop.yarn.ipc.HadoopYarnProtoRPC，部分代码如下：

• YarnRPC：

public abstract class YarnRPC {
   // ......

    public static YarnRPC create(Configuration conf) {
    LOG.debug("Creating YarnRPC for " +
        conf.get(YarnConfiguration.IPC_RPC_IMPL));
    String clazzName = conf.get(YarnConfiguration.IPC_RPC_IMPL);
    if (clazzName == null) {
      clazzName = YarnConfiguration.DEFAULT_IPC_RPC_IMPL;
    }
    try {
      return (YarnRPC) Class.forName(clazzName).newInstance();
    } catch (Exception e) {
      throw new YarnRuntimeException(e);
    }
  }

}

• YarnConfiguration类：

public class YarnConfiguration extends Configuration {

  //Configurations
  public static final String YARN_PREFIX = "yarn.";

  ////////////////////////////////
  // IPC Configs
  ////////////////////////////////
  public static final String IPC_PREFIX = YARN_PREFIX + "ipc.";
  /** RPC class implementation*/
  public static final String IPC_RPC_IMPL =
    IPC_PREFIX + "rpc.class";
  public static final String DEFAULT_IPC_RPC_IMPL =
    "org.apache.hadoop.yarn.ipc.HadoopYarnProtoRPC";
}

　　而HadoopYarnProtoRPC 通过 RPC 的 RpcFactoryProvider 生成客户端工厂(由参数 yarn.ipc.client.factory.class 指定，默认值是 org.apache.hadoop.yarn.factories.impl.pb.RpcClientFactoryPBImpl)和服务器工厂 (由参数 yarn.ipc.server.factory.class 指定，默认值是 org.apache.hadoop.yarn.factories.impl.pb.RpcServerFactoryPBImpl)，以根据通信协议的 Protocol Buffers 定义生成客户端对象和服务器对象。相关类的部分代码如下：

• HadoopYarnProtoRPC

public class HadoopYarnProtoRPC extends YarnRPC {

  private static final Log LOG = LogFactory.getLog(HadoopYarnProtoRPC.class);

  @Override
  public Object getProxy(Class protocol, InetSocketAddress addr,
      Configuration conf) {
    LOG.debug("Creating a HadoopYarnProtoRpc proxy for protocol " + protocol);
    return RpcFactoryProvider.getClientFactory(conf).getClient(protocol, 1,
        addr, conf);
  }

  @Override
  public void stopProxy(Object proxy, Configuration conf) {
    RpcFactoryProvider.getClientFactory(conf).stopClient(proxy);
  }

  @Override
  public Server getServer(Class protocol, Object instance,
      InetSocketAddress addr, Configuration conf,
      SecretManager<? extends TokenIdentifier> secretManager,
      int numHandlers, String portRangeConfig) {
    LOG.debug("Creating a HadoopYarnProtoRpc server for protocol " + protocol +
        " with " + numHandlers + " handlers");

    return RpcFactoryProvider.getServerFactory(conf).getServer(protocol,
        instance, addr, conf, secretManager, numHandlers, portRangeConfig);

  }

}

• RpcFactoryProvider

public class RpcFactoryProvider {

  // ......

  public static RpcClientFactory getClientFactory(Configuration conf) {
    String clientFactoryClassName = conf.get(
        YarnConfiguration.IPC_CLIENT_FACTORY_CLASS,
        YarnConfiguration.DEFAULT_IPC_CLIENT_FACTORY_CLASS);
    return (RpcClientFactory) getFactoryClassInstance(clientFactoryClassName);
  }

  //......

}

/** Factory to create client IPC classes.*/
  public static final String IPC_CLIENT_FACTORY_CLASS =
    IPC_PREFIX + "client.factory.class";
  public static final String DEFAULT_IPC_CLIENT_FACTORY_CLASS =
      "org.apache.hadoop.yarn.factories.impl.pb.RpcClientFactoryPBImpl";

　　在 YARN 中并未使用Hadoop自带的Writable来做序列化，而是使用 Protocol Buffers 作为默认的序列化机制，这带来的好处主要有以下几点：

• 继承Protocol Buffers的优点：Protocol Buffers已被实践证明其拥有高效性、可扩展性、紧凑性以及跨语言性等特点。
• 支持在线升级回滚：在Hadoop 2.x版本后，添加的HA方案，该方案能够进行主备切换，在不停止NNA节点服务的前提下，能够在线升级版本。

3.YARN的RPC示例

　　YARN 的工作流程是先定义通信协议接口ResourceTracker，它包含2个函数，具体代码如下所示：

• ResourceTracker：

public interface ResourceTracker {

  @Idempotent
  public RegisterNodeManagerResponse registerNodeManager(
      RegisterNodeManagerRequest request) throws YarnException,
      IOException;

  @AtMostOnce
  public NodeHeartbeatResponse nodeHeartbeat(NodeHeartbeatRequest request)
      throws YarnException, IOException;

}

　　这里ResourceTracker提供了Protocol Buffers定义和Java实现，其中设计的Protocol Buffers文件有：ResourceTracker.proto、yarn_server_common_service_protos.proto和yarn_server_common_protos.proto，文件路径在Hadoop的源码包的 hadoop-2.6.0-src/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-common/src/main/proto，这里就不贴出3个文件的具体代码类，大家可以到该目录去阅读这部分代码。这里需要注意的是，若是大家要编译这些文件需要安装 ProtoBuf 的编译环境，环境安装较为简单，这里给大家简要说明下。

　　首先是下载ProtoBuf的安装包，然后解压，进入到解压目录，编译安装。命令如下：

./configure --prefix=/home/work /protobuf/  

make && make install

最后编译 .proto 文件的命令：

protoc ./ResourceTracker.proto  --java_out=./

　　下面，我们去收取Hadoop源码到本地工程，运行调试相关代码。

• TestYarnServerApiClasses：

public class TestYarnServerApiClasses {

  // ......

  // 列举测试4个方法  

@Test
  public void testRegisterNodeManagerResponsePBImpl() {
    RegisterNodeManagerResponsePBImpl original =
        new RegisterNodeManagerResponsePBImpl();
    original.setContainerTokenMasterKey(getMasterKey());
    original.setNMTokenMasterKey(getMasterKey());
    original.setNodeAction(NodeAction.NORMAL);
    original.setDiagnosticsMessage("testDiagnosticMessage");

    RegisterNodeManagerResponsePBImpl copy =
        new RegisterNodeManagerResponsePBImpl(
            original.getProto());
    assertEquals(1, copy.getContainerTokenMasterKey().getKeyId());
    assertEquals(1, copy.getNMTokenMasterKey().getKeyId());
    assertEquals(NodeAction.NORMAL, copy.getNodeAction());
    assertEquals("testDiagnosticMessage", copy.getDiagnosticsMessage());

  }

@Test
  public void testNodeHeartbeatRequestPBImpl() {
    NodeHeartbeatRequestPBImpl original = new NodeHeartbeatRequestPBImpl();
    original.setLastKnownContainerTokenMasterKey(getMasterKey());
    original.setLastKnownNMTokenMasterKey(getMasterKey());
    original.setNodeStatus(getNodeStatus());
    NodeHeartbeatRequestPBImpl copy = new NodeHeartbeatRequestPBImpl(
        original.getProto());
    assertEquals(1, copy.getLastKnownContainerTokenMasterKey().getKeyId());
    assertEquals(1, copy.getLastKnownNMTokenMasterKey().getKeyId());
    assertEquals("localhost", copy.getNodeStatus().getNodeId().getHost());
  }

@Test
  public void testNodeHeartbeatResponsePBImpl() {
    NodeHeartbeatResponsePBImpl original = new NodeHeartbeatResponsePBImpl();

    original.setDiagnosticsMessage("testDiagnosticMessage");
    original.setContainerTokenMasterKey(getMasterKey());
    original.setNMTokenMasterKey(getMasterKey());
    original.setNextHeartBeatInterval(1000);
    original.setNodeAction(NodeAction.NORMAL);
    original.setResponseId(100);

    NodeHeartbeatResponsePBImpl copy = new NodeHeartbeatResponsePBImpl(
        original.getProto());
    assertEquals(100, copy.getResponseId());
    assertEquals(NodeAction.NORMAL, copy.getNodeAction());
    assertEquals(1000, copy.getNextHeartBeatInterval());
    assertEquals(1, copy.getContainerTokenMasterKey().getKeyId());
    assertEquals(1, copy.getNMTokenMasterKey().getKeyId());
    assertEquals("testDiagnosticMessage", copy.getDiagnosticsMessage());
  }

@Test
  public void testRegisterNodeManagerRequestPBImpl() {
    RegisterNodeManagerRequestPBImpl original = new RegisterNodeManagerRequestPBImpl();
    original.setHttpPort(8080);
    original.setNodeId(getNodeId());
    Resource resource = recordFactory.newRecordInstance(Resource.class);
    resource.setMemory(10000);
    resource.setVirtualCores(2);
    original.setResource(resource);
    RegisterNodeManagerRequestPBImpl copy = new RegisterNodeManagerRequestPBImpl(
        original.getProto());

    assertEquals(8080, copy.getHttpPort());
    assertEquals(9090, copy.getNodeId().getPort());
    assertEquals(10000, copy.getResource().getMemory());
    assertEquals(2, copy.getResource().getVirtualCores());

  }

}

• TestResourceTrackerPBClientImpl：

public class TestResourceTrackerPBClientImpl {

    private static ResourceTracker client;
    private static Server server;
    private final static org.apache.hadoop.yarn.factories.RecordFactory recordFactory = RecordFactoryProvider
            .getRecordFactory(null);

    @BeforeClass
    public static void start() {

        System.out.println("Start client test");

        InetSocketAddress address = new InetSocketAddress(0);
        Configuration configuration = new Configuration();
        ResourceTracker instance = new ResourceTrackerTestImpl();
        server = RpcServerFactoryPBImpl.get().getServer(ResourceTracker.class, instance, address, configuration, null,
                1);
        server.start();

        client = (ResourceTracker) RpcClientFactoryPBImpl.get().getClient(ResourceTracker.class, 1,
                NetUtils.getConnectAddress(server), configuration);

    }

    @AfterClass
    public static void stop() {

        System.out.println("Stop client");

        if (server != null) {
            server.stop();
        }
    }

    /**
     * Test the method registerNodeManager. Method should return a not null
     * result.
     *
     */
    @Test
    public void testResourceTrackerPBClientImpl() throws Exception {
        RegisterNodeManagerRequest request = recordFactory.newRecordInstance(RegisterNodeManagerRequest.class);
        assertNotNull(client.registerNodeManager(request));

        ResourceTrackerTestImpl.exception = true;
        try {
            client.registerNodeManager(request);
            fail("there should be YarnException");
        } catch (YarnException e) {
            assertTrue(e.getMessage().startsWith("testMessage"));
        } finally {
            ResourceTrackerTestImpl.exception = false;
        }

    }

    /**
     * Test the method nodeHeartbeat. Method should return a not null result.
     *
     */

    @Test
    public void testNodeHeartbeat() throws Exception {
        NodeHeartbeatRequest request = recordFactory.newRecordInstance(NodeHeartbeatRequest.class);
        assertNotNull(client.nodeHeartbeat(request));

        ResourceTrackerTestImpl.exception = true;
        try {
            client.nodeHeartbeat(request);
            fail("there  should be YarnException");
        } catch (YarnException e) {
            assertTrue(e.getMessage().startsWith("testMessage"));
        } finally {
            ResourceTrackerTestImpl.exception = false;
        }

    }

    public static class ResourceTrackerTestImpl implements ResourceTracker {

        public static boolean exception = false;

        public RegisterNodeManagerResponse registerNodeManager(RegisterNodeManagerRequest request)
                throws YarnException, IOException {
            if (exception) {
                throw new YarnException("testMessage");
            }
            return recordFactory.newRecordInstance(RegisterNodeManagerResponse.class);
        }

        public NodeHeartbeatResponse nodeHeartbeat(NodeHeartbeatRequest request) throws YarnException, IOException {
            if (exception) {
                throw new YarnException("testMessage");
            }
            return recordFactory.newRecordInstance(NodeHeartbeatResponse.class);
        }

    }
}

4.截图预览

　　接下来，我们使用JUnit去测试代码，截图预览如下所示：

• 对testRegisterNodeManagerRequestPBImpl()方法的一个DEBUG调试

• testResourceTrackerPBClientImpl()方法的DEBUG调试

　　这里由于设置exception的状态为true，在调用registerNodeManager()时，会打印一条测试异常信息。

if (exception) {
　　throw new YarnException("testMessage");
}

5.总结

　　在学习Hadoop YARN的RPC时，可以先了解Hadoop的RPC机制，这样在接触YARN的RPC的会比较好理解，YARN的RPC只是其中的一部分，后续会给大家分享更多关于YARN的内容。

6.结束语

　　这篇博客就和大家分享到这里，如果大家在研究学习的过程当中有什么问题，可以加群进行讨论或发送邮件给我，我会尽我所能为您解答，与君共勉！