org.apache.hadoop.yarn.ipc.YarnRPC#getServer

Source File: ResourceLocalizationService.java From big-c with Apache License 2.0

6 votes

Server createServer() {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  if (UserGroupInformation.isSecurityEnabled()) {
    secretManager = new LocalizerTokenSecretManager();      
  }
  
  Server server = rpc.getServer(LocalizationProtocol.class, this,
      localizationServerAddress, conf, secretManager, 
      conf.getInt(YarnConfiguration.NM_LOCALIZER_CLIENT_THREAD_COUNT, 
          YarnConfiguration.DEFAULT_NM_LOCALIZER_CLIENT_THREAD_COUNT));
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    server.refreshServiceAcl(conf, new NMPolicyProvider());
  }
  
  return server;
}

Source File: SharedCacheUploaderService.java From big-c with Apache License 2.0

6 votes

@Override
protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  this.metrics = SharedCacheUploaderMetrics.getInstance();

  YarnRPC rpc = YarnRPC.create(conf);
  this.server =
      rpc.getServer(SCMUploaderProtocol.class, this, bindAddress,
          conf, null, // Secret manager null for now (security not supported)
          conf.getInt(YarnConfiguration.SCM_UPLOADER_SERVER_THREAD_COUNT,
              YarnConfiguration.DEFAULT_SCM_UPLOADER_SERVER_THREAD_COUNT));

  // TODO (YARN-2774): Enable service authorization

  this.server.start();
  bindAddress =
      conf.updateConnectAddr(YarnConfiguration.SCM_UPLOADER_SERVER_ADDRESS,
          server.getListenerAddress());

  super.serviceStart();
}

Source File: SharedCacheUploaderService.java From hadoop with Apache License 2.0

6 votes

@Override
protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  this.metrics = SharedCacheUploaderMetrics.getInstance();

  YarnRPC rpc = YarnRPC.create(conf);
  this.server =
      rpc.getServer(SCMUploaderProtocol.class, this, bindAddress,
          conf, null, // Secret manager null for now (security not supported)
          conf.getInt(YarnConfiguration.SCM_UPLOADER_SERVER_THREAD_COUNT,
              YarnConfiguration.DEFAULT_SCM_UPLOADER_SERVER_THREAD_COUNT));

  // TODO (YARN-2774): Enable service authorization

  this.server.start();
  bindAddress =
      conf.updateConnectAddr(YarnConfiguration.SCM_UPLOADER_SERVER_ADDRESS,
          server.getListenerAddress());

  super.serviceStart();
}

Source File: ClientProtocolService.java From hadoop with Apache License 2.0

6 votes

@Override
protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  this.metrics = ClientSCMMetrics.getInstance();

  YarnRPC rpc = YarnRPC.create(conf);
  this.server =
      rpc.getServer(ClientSCMProtocol.class, this,
          clientBindAddress,
          conf, null, // Secret manager null for now (security not supported)
          conf.getInt(YarnConfiguration.SCM_CLIENT_SERVER_THREAD_COUNT,
              YarnConfiguration.DEFAULT_SCM_CLIENT_SERVER_THREAD_COUNT));

  // TODO (YARN-2774): Enable service authorization

  this.server.start();
  clientBindAddress =
      conf.updateConnectAddr(YarnConfiguration.SCM_CLIENT_SERVER_ADDRESS,
          server.getListenerAddress());

  super.serviceStart();
}

Source File: SCMAdminProtocolService.java From hadoop with Apache License 2.0

6 votes

@Override
protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  this.server =
      rpc.getServer(SCMAdminProtocol.class, this,
          clientBindAddress,
          conf, null, // Secret manager null for now (security not supported)
          conf.getInt(YarnConfiguration.SCM_ADMIN_CLIENT_THREAD_COUNT,
              YarnConfiguration.DEFAULT_SCM_ADMIN_CLIENT_THREAD_COUNT));

  // TODO: Enable service authorization (see YARN-2774)

  this.server.start();
  clientBindAddress =
      conf.updateConnectAddr(YarnConfiguration.SCM_ADMIN_ADDRESS,
          server.getListenerAddress());

  super.serviceStart();
}

Source File: ResourceLocalizationService.java From hadoop with Apache License 2.0

6 votes

Server createServer() {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  if (UserGroupInformation.isSecurityEnabled()) {
    secretManager = new LocalizerTokenSecretManager();      
  }
  
  Server server = rpc.getServer(LocalizationProtocol.class, this,
      localizationServerAddress, conf, secretManager, 
      conf.getInt(YarnConfiguration.NM_LOCALIZER_CLIENT_THREAD_COUNT, 
          YarnConfiguration.DEFAULT_NM_LOCALIZER_CLIENT_THREAD_COUNT));
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    server.refreshServiceAcl(conf, new NMPolicyProvider());
  }
  
  return server;
}

Source File: TestClientRedirect.java From big-c with Apache License 2.0

6 votes

public void start(Configuration conf) {
  YarnRPC rpc = YarnRPC.create(conf);
  //TODO : use fixed port ??
  InetSocketAddress address = NetUtils.createSocketAddr(hostAddress);
  InetAddress hostNameResolved = null;
  try {
    address.getAddress();
    hostNameResolved = InetAddress.getLocalHost();
  } catch (UnknownHostException e) {
    throw new YarnRuntimeException(e);
  }

  server =
      rpc.getServer(protocol, this, address,
          conf, null, 1);
  server.start();
  this.bindAddress = NetUtils.getConnectAddress(server);
   super.start();
   amRunning = true;
}

Source File: MRClientService.java From big-c with Apache License 2.0

5 votes

protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  InetSocketAddress address = new InetSocketAddress(0);

  server =
      rpc.getServer(MRClientProtocol.class, protocolHandler, address,
          conf, appContext.getClientToAMTokenSecretManager(),
          conf.getInt(MRJobConfig.MR_AM_JOB_CLIENT_THREAD_COUNT, 
              MRJobConfig.DEFAULT_MR_AM_JOB_CLIENT_THREAD_COUNT),
              MRJobConfig.MR_AM_JOB_CLIENT_PORT_RANGE);
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    refreshServiceAcls(conf, new MRAMPolicyProvider());
  }

  server.start();
  this.bindAddress = NetUtils.createSocketAddrForHost(appContext.getNMHostname(),
      server.getListenerAddress().getPort());
  LOG.info("Instantiated MRClientService at " + this.bindAddress);
  try {
    // Explicitly disabling SSL for map reduce task as we can't allow MR users
    // to gain access to keystore file for opening SSL listener. We can trust
    // RM/NM to issue SSL certificates but definitely not MR-AM as it is
    // running in user-land.
    webApp =
        WebApps.$for("mapreduce", AppContext.class, appContext, "ws")
          .withHttpPolicy(conf, Policy.HTTP_ONLY).start(new AMWebApp());
  } catch (Exception e) {
    LOG.error("Webapps failed to start. Ignoring for now:", e);
  }
  super.serviceStart();
}

Source File: HistoryClientService.java From hadoop with Apache License 2.0

5 votes

protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  initializeWebApp(conf);
  InetSocketAddress address = conf.getSocketAddr(
      JHAdminConfig.MR_HISTORY_BIND_HOST,
      JHAdminConfig.MR_HISTORY_ADDRESS,
      JHAdminConfig.DEFAULT_MR_HISTORY_ADDRESS,
      JHAdminConfig.DEFAULT_MR_HISTORY_PORT);

  server =
      rpc.getServer(HSClientProtocol.class, protocolHandler, address,
          conf, jhsDTSecretManager,
          conf.getInt(JHAdminConfig.MR_HISTORY_CLIENT_THREAD_COUNT,
              JHAdminConfig.DEFAULT_MR_HISTORY_CLIENT_THREAD_COUNT));

  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION,
      false)) {
    server.refreshServiceAcl(conf, new ClientHSPolicyProvider());
  }
  
  server.start();
  this.bindAddress = conf.updateConnectAddr(JHAdminConfig.MR_HISTORY_BIND_HOST,
                                            JHAdminConfig.MR_HISTORY_ADDRESS,
                                            JHAdminConfig.DEFAULT_MR_HISTORY_ADDRESS,
                                            server.getListenerAddress());
  LOG.info("Instantiated HistoryClientService at " + this.bindAddress);

  super.serviceStart();
}

Source File: ApplicationHistoryClientService.java From big-c with Apache License 2.0

5 votes

protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  InetSocketAddress address = conf.getSocketAddr(
      YarnConfiguration.TIMELINE_SERVICE_BIND_HOST,
      YarnConfiguration.TIMELINE_SERVICE_ADDRESS,
      YarnConfiguration.DEFAULT_TIMELINE_SERVICE_ADDRESS,
      YarnConfiguration.DEFAULT_TIMELINE_SERVICE_PORT);

  Preconditions.checkArgument(conf.getInt(
      YarnConfiguration.TIMELINE_SERVICE_HANDLER_THREAD_COUNT,
      YarnConfiguration.DEFAULT_TIMELINE_SERVICE_CLIENT_THREAD_COUNT) > 0,
      "%s property value should be greater than zero",
      YarnConfiguration.TIMELINE_SERVICE_HANDLER_THREAD_COUNT);

  server =
      rpc.getServer(ApplicationHistoryProtocol.class, this,
        address, conf, null, conf.getInt(
          YarnConfiguration.TIMELINE_SERVICE_HANDLER_THREAD_COUNT,
          YarnConfiguration.DEFAULT_TIMELINE_SERVICE_CLIENT_THREAD_COUNT));

  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, false)) {
    refreshServiceAcls(conf, new TimelinePolicyProvider());
  }

  server.start();
  this.bindAddress =
      conf.updateConnectAddr(YarnConfiguration.TIMELINE_SERVICE_BIND_HOST,
                             YarnConfiguration.TIMELINE_SERVICE_ADDRESS,
                             YarnConfiguration.DEFAULT_TIMELINE_SERVICE_ADDRESS,
                             server.getListenerAddress());
  LOG.info("Instantiated ApplicationHistoryClientService at "
      + this.bindAddress);

  super.serviceStart();
}

Source File: MRClientService.java From hadoop with Apache License 2.0

5 votes

protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  InetSocketAddress address = new InetSocketAddress(0);

  server =
      rpc.getServer(MRClientProtocol.class, protocolHandler, address,
          conf, appContext.getClientToAMTokenSecretManager(),
          conf.getInt(MRJobConfig.MR_AM_JOB_CLIENT_THREAD_COUNT, 
              MRJobConfig.DEFAULT_MR_AM_JOB_CLIENT_THREAD_COUNT),
              MRJobConfig.MR_AM_JOB_CLIENT_PORT_RANGE);
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    refreshServiceAcls(conf, new MRAMPolicyProvider());
  }

  server.start();
  this.bindAddress = NetUtils.createSocketAddrForHost(appContext.getNMHostname(),
      server.getListenerAddress().getPort());
  LOG.info("Instantiated MRClientService at " + this.bindAddress);
  try {
    // Explicitly disabling SSL for map reduce task as we can't allow MR users
    // to gain access to keystore file for opening SSL listener. We can trust
    // RM/NM to issue SSL certificates but definitely not MR-AM as it is
    // running in user-land.
    webApp =
        WebApps.$for("mapreduce", AppContext.class, appContext, "ws")
          .withHttpPolicy(conf, Policy.HTTP_ONLY).start(new AMWebApp());
  } catch (Exception e) {
    LOG.error("Webapps failed to start. Ignoring for now:", e);
  }
  super.serviceStart();
}

Source File: ApplicationHistoryClientService.java From hadoop with Apache License 2.0

5 votes

protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  InetSocketAddress address = conf.getSocketAddr(
      YarnConfiguration.TIMELINE_SERVICE_BIND_HOST,
      YarnConfiguration.TIMELINE_SERVICE_ADDRESS,
      YarnConfiguration.DEFAULT_TIMELINE_SERVICE_ADDRESS,
      YarnConfiguration.DEFAULT_TIMELINE_SERVICE_PORT);

  Preconditions.checkArgument(conf.getInt(
      YarnConfiguration.TIMELINE_SERVICE_HANDLER_THREAD_COUNT,
      YarnConfiguration.DEFAULT_TIMELINE_SERVICE_CLIENT_THREAD_COUNT) > 0,
      "%s property value should be greater than zero",
      YarnConfiguration.TIMELINE_SERVICE_HANDLER_THREAD_COUNT);

  server =
      rpc.getServer(ApplicationHistoryProtocol.class, this,
        address, conf, null, conf.getInt(
          YarnConfiguration.TIMELINE_SERVICE_HANDLER_THREAD_COUNT,
          YarnConfiguration.DEFAULT_TIMELINE_SERVICE_CLIENT_THREAD_COUNT));

  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, false)) {
    refreshServiceAcls(conf, new TimelinePolicyProvider());
  }

  server.start();
  this.bindAddress =
      conf.updateConnectAddr(YarnConfiguration.TIMELINE_SERVICE_BIND_HOST,
                             YarnConfiguration.TIMELINE_SERVICE_ADDRESS,
                             YarnConfiguration.DEFAULT_TIMELINE_SERVICE_ADDRESS,
                             server.getListenerAddress());
  LOG.info("Instantiated ApplicationHistoryClientService at "
      + this.bindAddress);

  super.serviceStart();
}

Source File: ClientRMService.java From hadoop with Apache License 2.0

5 votes

@Override
protected void serviceStart() throws Exception {
  Configuration conf = getConfig();
  YarnRPC rpc = YarnRPC.create(conf);
  this.server =   
    rpc.getServer(ApplicationClientProtocol.class, this,
          clientBindAddress,
          conf, this.rmDTSecretManager,
          conf.getInt(YarnConfiguration.RM_CLIENT_THREAD_COUNT, 
              YarnConfiguration.DEFAULT_RM_CLIENT_THREAD_COUNT));
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    InputStream inputStream =
        this.rmContext.getConfigurationProvider()
            .getConfigurationInputStream(conf,
                YarnConfiguration.HADOOP_POLICY_CONFIGURATION_FILE);
    if (inputStream != null) {
      conf.addResource(inputStream);
    }
    refreshServiceAcls(conf, RMPolicyProvider.getInstance());
  }
  
  this.server.start();
  clientBindAddress = conf.updateConnectAddr(YarnConfiguration.RM_BIND_HOST,
                                             YarnConfiguration.RM_ADDRESS,
                                             YarnConfiguration.DEFAULT_RM_ADDRESS,
                                             server.getListenerAddress());
  super.serviceStart();
}

Source File: TestClientRedirect.java From big-c with Apache License 2.0

5 votes

@Override
protected void serviceStart() throws Exception {
  // All the clients to appsManager are supposed to be authenticated via
  // Kerberos if security is enabled, so no secretManager.
  YarnRPC rpc = YarnRPC.create(getConfig());
  Configuration clientServerConf = new Configuration(getConfig());
  this.server = rpc.getServer(ApplicationClientProtocol.class, this,
      clientBindAddress, clientServerConf, null, 1);
  this.server.start();
  super.serviceStart();
}

Source File: ContainerManagerImpl.java From big-c with Apache License 2.0

4 votes

@Override
protected void serviceStart() throws Exception {

  // Enqueue user dirs in deletion context

  Configuration conf = getConfig();
  final InetSocketAddress initialAddress = conf.getSocketAddr(
      YarnConfiguration.NM_BIND_HOST,
      YarnConfiguration.NM_ADDRESS,
      YarnConfiguration.DEFAULT_NM_ADDRESS,
      YarnConfiguration.DEFAULT_NM_PORT);
  boolean usingEphemeralPort = (initialAddress.getPort() == 0);
  if (context.getNMStateStore().canRecover() && usingEphemeralPort) {
    throw new IllegalArgumentException("Cannot support recovery with an "
        + "ephemeral server port. Check the setting of "
        + YarnConfiguration.NM_ADDRESS);
  }
  // If recovering then delay opening the RPC service until the recovery
  // of resources and containers have completed, otherwise requests from
  // clients during recovery can interfere with the recovery process.
  final boolean delayedRpcServerStart =
      context.getNMStateStore().canRecover();

  Configuration serverConf = new Configuration(conf);

  // always enforce it to be token-based.
  serverConf.set(
    CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHENTICATION,
    SaslRpcServer.AuthMethod.TOKEN.toString());
  
  YarnRPC rpc = YarnRPC.create(conf);

  server =
      rpc.getServer(ContainerManagementProtocol.class, this, initialAddress, 
          serverConf, this.context.getNMTokenSecretManager(),
          conf.getInt(YarnConfiguration.NM_CONTAINER_MGR_THREAD_COUNT, 
              YarnConfiguration.DEFAULT_NM_CONTAINER_MGR_THREAD_COUNT));
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    refreshServiceAcls(conf, new NMPolicyProvider());
  }
  
  LOG.info("Blocking new container-requests as container manager rpc" +
  		" server is still starting.");
  this.setBlockNewContainerRequests(true);

  String bindHost = conf.get(YarnConfiguration.NM_BIND_HOST);
  String nmAddress = conf.getTrimmed(YarnConfiguration.NM_ADDRESS);
  String hostOverride = null;
  if (bindHost != null && !bindHost.isEmpty()
      && nmAddress != null && !nmAddress.isEmpty()) {
    //a bind-host case with an address, to support overriding the first
    //hostname found when querying for our hostname with the specified
    //address, combine the specified address with the actual port listened
    //on by the server
    hostOverride = nmAddress.split(":")[0];
  }

  // setup node ID
  InetSocketAddress connectAddress;
  if (delayedRpcServerStart) {
    connectAddress = NetUtils.getConnectAddress(initialAddress);
  } else {
    server.start();
    connectAddress = NetUtils.getConnectAddress(server);
  }
  NodeId nodeId = buildNodeId(connectAddress, hostOverride);
  ((NodeManager.NMContext)context).setNodeId(nodeId);
  this.context.getNMTokenSecretManager().setNodeId(nodeId);
  this.context.getContainerTokenSecretManager().setNodeId(nodeId);

  // start remaining services
  super.serviceStart();

  if (delayedRpcServerStart) {
    waitForRecoveredContainers();
    server.start();

    // check that the node ID is as previously advertised
    connectAddress = NetUtils.getConnectAddress(server);
    NodeId serverNode = buildNodeId(connectAddress, hostOverride);
    if (!serverNode.equals(nodeId)) {
      throw new IOException("Node mismatch after server started, expected '"
          + nodeId + "' but found '" + serverNode + "'");
    }
  }

  LOG.info("ContainerManager started at " + connectAddress);
  LOG.info("ContainerManager bound to " + initialAddress);
}

Source File: TestContainerLaunchRPC.java From big-c with Apache License 2.0

4 votes

private void testRPCTimeout(String rpcClass) throws Exception {
  Configuration conf = new Configuration();
  // set timeout low for the test
  conf.setInt("yarn.rpc.nm-command-timeout", 3000);

  conf.set(YarnConfiguration.IPC_RPC_IMPL, rpcClass);
  YarnRPC rpc = YarnRPC.create(conf);
  String bindAddr = "localhost:0";
  InetSocketAddress addr = NetUtils.createSocketAddr(bindAddr);
  Server server = rpc.getServer(ContainerManagementProtocol.class,
      new DummyContainerManager(), addr, conf, null, 1);
  server.start();
  try {

    ContainerManagementProtocol proxy = (ContainerManagementProtocol) rpc.getProxy(
        ContainerManagementProtocol.class,
        server.getListenerAddress(), conf);
    ContainerLaunchContext containerLaunchContext = recordFactory
        .newRecordInstance(ContainerLaunchContext.class);

    ApplicationId applicationId = ApplicationId.newInstance(0, 0);
    ApplicationAttemptId applicationAttemptId =
        ApplicationAttemptId.newInstance(applicationId, 0);
    ContainerId containerId =
        ContainerId.newContainerId(applicationAttemptId, 100);
    NodeId nodeId = NodeId.newInstance("localhost", 1234);
    Resource resource = Resource.newInstance(1234, 2);
    ContainerTokenIdentifier containerTokenIdentifier =
        new ContainerTokenIdentifier(containerId, "localhost", "user",
          resource, System.currentTimeMillis() + 10000, 42, 42,
          Priority.newInstance(0), 0);
    Token containerToken =
        TestRPC.newContainerToken(nodeId, "password".getBytes(),
          containerTokenIdentifier);

    StartContainerRequest scRequest =
        StartContainerRequest.newInstance(containerLaunchContext,
          containerToken);
    List<StartContainerRequest> list = new ArrayList<StartContainerRequest>();
    list.add(scRequest);
    StartContainersRequest allRequests =
        StartContainersRequest.newInstance(list);
    try {
      proxy.startContainers(allRequests);
    } catch (Exception e) {
      LOG.info(StringUtils.stringifyException(e));
      Assert.assertEquals("Error, exception is not: "
          + SocketTimeoutException.class.getName(),
          SocketTimeoutException.class.getName(), e.getClass().getName());
      return;
    }
  } finally {
    server.stop();
  }

  Assert.fail("timeout exception should have occurred!");
}

Source File: TestContainerLauncher.java From hadoop with Apache License 2.0

4 votes

@Test(timeout = 15000)
public void testSlowNM() throws Exception {

  conf = new Configuration();
  int maxAttempts = 1;
  conf.setInt(MRJobConfig.MAP_MAX_ATTEMPTS, maxAttempts);
  conf.setBoolean(MRJobConfig.JOB_UBERTASK_ENABLE, false);
  // set timeout low for the test
  conf.setInt("yarn.rpc.nm-command-timeout", 3000);
  conf.set(YarnConfiguration.IPC_RPC_IMPL, HadoopYarnProtoRPC.class.getName());
  YarnRPC rpc = YarnRPC.create(conf);
  String bindAddr = "localhost:0";
  InetSocketAddress addr = NetUtils.createSocketAddr(bindAddr);
  NMTokenSecretManagerInNM tokenSecretManager =
      new NMTokenSecretManagerInNM();
  MasterKey masterKey = Records.newRecord(MasterKey.class);
  masterKey.setBytes(ByteBuffer.wrap("key".getBytes()));
  tokenSecretManager.setMasterKey(masterKey);
  conf.set(CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHENTICATION,
    "token");
  server =
      rpc.getServer(ContainerManagementProtocol.class,
        new DummyContainerManager(), addr, conf, tokenSecretManager, 1);
  server.start();

  MRApp app = new MRAppWithSlowNM(tokenSecretManager);

  try {
  Job job = app.submit(conf);
  app.waitForState(job, JobState.RUNNING);

  Map<TaskId, Task> tasks = job.getTasks();
  Assert.assertEquals("Num tasks is not correct", 1, tasks.size());

  Task task = tasks.values().iterator().next();
  app.waitForState(task, TaskState.SCHEDULED);

  Map<TaskAttemptId, TaskAttempt> attempts = tasks.values().iterator()
      .next().getAttempts();
    Assert.assertEquals("Num attempts is not correct", maxAttempts,
        attempts.size());

  TaskAttempt attempt = attempts.values().iterator().next();
    app.waitForInternalState((TaskAttemptImpl) attempt,
        TaskAttemptStateInternal.ASSIGNED);

  app.waitForState(job, JobState.FAILED);

  String diagnostics = attempt.getDiagnostics().toString();
  LOG.info("attempt.getDiagnostics: " + diagnostics);

    Assert.assertTrue(diagnostics.contains("Container launch failed for "
        + "container_0_0000_01_000000 : "));
    Assert
        .assertTrue(diagnostics
            .contains("java.net.SocketTimeoutException: 3000 millis timeout while waiting for channel"));

  } finally {
    server.stop();
  app.stop();
}
}

Source File: ContainerManagerImpl.java From hadoop with Apache License 2.0

4 votes

@Override
protected void serviceStart() throws Exception {

  // Enqueue user dirs in deletion context

  Configuration conf = getConfig();
  final InetSocketAddress initialAddress = conf.getSocketAddr(
      YarnConfiguration.NM_BIND_HOST,
      YarnConfiguration.NM_ADDRESS,
      YarnConfiguration.DEFAULT_NM_ADDRESS,
      YarnConfiguration.DEFAULT_NM_PORT);
  boolean usingEphemeralPort = (initialAddress.getPort() == 0);
  if (context.getNMStateStore().canRecover() && usingEphemeralPort) {
    throw new IllegalArgumentException("Cannot support recovery with an "
        + "ephemeral server port. Check the setting of "
        + YarnConfiguration.NM_ADDRESS);
  }
  // If recovering then delay opening the RPC service until the recovery
  // of resources and containers have completed, otherwise requests from
  // clients during recovery can interfere with the recovery process.
  final boolean delayedRpcServerStart =
      context.getNMStateStore().canRecover();

  Configuration serverConf = new Configuration(conf);

  // always enforce it to be token-based.
  serverConf.set(
    CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHENTICATION,
    SaslRpcServer.AuthMethod.TOKEN.toString());
  
  YarnRPC rpc = YarnRPC.create(conf);

  server =
      rpc.getServer(ContainerManagementProtocol.class, this, initialAddress, 
          serverConf, this.context.getNMTokenSecretManager(),
          conf.getInt(YarnConfiguration.NM_CONTAINER_MGR_THREAD_COUNT, 
              YarnConfiguration.DEFAULT_NM_CONTAINER_MGR_THREAD_COUNT));
  
  // Enable service authorization?
  if (conf.getBoolean(
      CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHORIZATION, 
      false)) {
    refreshServiceAcls(conf, new NMPolicyProvider());
  }
  
  LOG.info("Blocking new container-requests as container manager rpc" +
  		" server is still starting.");
  this.setBlockNewContainerRequests(true);

  String bindHost = conf.get(YarnConfiguration.NM_BIND_HOST);
  String nmAddress = conf.getTrimmed(YarnConfiguration.NM_ADDRESS);
  String hostOverride = null;
  if (bindHost != null && !bindHost.isEmpty()
      && nmAddress != null && !nmAddress.isEmpty()) {
    //a bind-host case with an address, to support overriding the first
    //hostname found when querying for our hostname with the specified
    //address, combine the specified address with the actual port listened
    //on by the server
    hostOverride = nmAddress.split(":")[0];
  }

  // setup node ID
  InetSocketAddress connectAddress;
  if (delayedRpcServerStart) {
    connectAddress = NetUtils.getConnectAddress(initialAddress);
  } else {
    server.start();
    connectAddress = NetUtils.getConnectAddress(server);
  }
  NodeId nodeId = buildNodeId(connectAddress, hostOverride);
  ((NodeManager.NMContext)context).setNodeId(nodeId);
  this.context.getNMTokenSecretManager().setNodeId(nodeId);
  this.context.getContainerTokenSecretManager().setNodeId(nodeId);

  // start remaining services
  super.serviceStart();

  if (delayedRpcServerStart) {
    waitForRecoveredContainers();
    server.start();

    // check that the node ID is as previously advertised
    connectAddress = NetUtils.getConnectAddress(server);
    NodeId serverNode = buildNodeId(connectAddress, hostOverride);
    if (!serverNode.equals(nodeId)) {
      throw new IOException("Node mismatch after server started, expected '"
          + nodeId + "' but found '" + serverNode + "'");
    }
  }

  LOG.info("ContainerManager started at " + connectAddress);
  LOG.info("ContainerManager bound to " + initialAddress);
}

Source File: TestContainerLaunchRPC.java From hadoop with Apache License 2.0

4 votes

private void testRPCTimeout(String rpcClass) throws Exception {
  Configuration conf = new Configuration();
  // set timeout low for the test
  conf.setInt("yarn.rpc.nm-command-timeout", 3000);

  conf.set(YarnConfiguration.IPC_RPC_IMPL, rpcClass);
  YarnRPC rpc = YarnRPC.create(conf);
  String bindAddr = "localhost:0";
  InetSocketAddress addr = NetUtils.createSocketAddr(bindAddr);
  Server server = rpc.getServer(ContainerManagementProtocol.class,
      new DummyContainerManager(), addr, conf, null, 1);
  server.start();
  try {

    ContainerManagementProtocol proxy = (ContainerManagementProtocol) rpc.getProxy(
        ContainerManagementProtocol.class,
        server.getListenerAddress(), conf);
    ContainerLaunchContext containerLaunchContext = recordFactory
        .newRecordInstance(ContainerLaunchContext.class);

    ApplicationId applicationId = ApplicationId.newInstance(0, 0);
    ApplicationAttemptId applicationAttemptId =
        ApplicationAttemptId.newInstance(applicationId, 0);
    ContainerId containerId =
        ContainerId.newContainerId(applicationAttemptId, 100);
    NodeId nodeId = NodeId.newInstance("localhost", 1234);
    Resource resource = Resource.newInstance(1234, 2, 3);
    ContainerTokenIdentifier containerTokenIdentifier =
        new ContainerTokenIdentifier(containerId, "localhost", "user",
          resource, System.currentTimeMillis() + 10000, 42, 42,
          Priority.newInstance(0), 0);
    Token containerToken =
        TestRPC.newContainerToken(nodeId, "password".getBytes(),
          containerTokenIdentifier);

    StartContainerRequest scRequest =
        StartContainerRequest.newInstance(containerLaunchContext,
          containerToken);
    List<StartContainerRequest> list = new ArrayList<StartContainerRequest>();
    list.add(scRequest);
    StartContainersRequest allRequests =
        StartContainersRequest.newInstance(list);
    try {
      proxy.startContainers(allRequests);
    } catch (Exception e) {
      LOG.info(StringUtils.stringifyException(e));
      Assert.assertEquals("Error, exception is not: "
          + SocketTimeoutException.class.getName(),
          SocketTimeoutException.class.getName(), e.getClass().getName());
      return;
    }
  } finally {
    server.stop();
  }

  Assert.fail("timeout exception should have occurred!");
}

Source File: TestContainerLauncher.java From big-c with Apache License 2.0

4 votes

@Test(timeout = 15000)
public void testSlowNM() throws Exception {

  conf = new Configuration();
  int maxAttempts = 1;
  conf.setInt(MRJobConfig.MAP_MAX_ATTEMPTS, maxAttempts);
  conf.setBoolean(MRJobConfig.JOB_UBERTASK_ENABLE, false);
  // set timeout low for the test
  conf.setInt("yarn.rpc.nm-command-timeout", 3000);
  conf.set(YarnConfiguration.IPC_RPC_IMPL, HadoopYarnProtoRPC.class.getName());
  YarnRPC rpc = YarnRPC.create(conf);
  String bindAddr = "localhost:0";
  InetSocketAddress addr = NetUtils.createSocketAddr(bindAddr);
  NMTokenSecretManagerInNM tokenSecretManager =
      new NMTokenSecretManagerInNM();
  MasterKey masterKey = Records.newRecord(MasterKey.class);
  masterKey.setBytes(ByteBuffer.wrap("key".getBytes()));
  tokenSecretManager.setMasterKey(masterKey);
  conf.set(CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHENTICATION,
    "token");
  server =
      rpc.getServer(ContainerManagementProtocol.class,
        new DummyContainerManager(), addr, conf, tokenSecretManager, 1);
  server.start();

  MRApp app = new MRAppWithSlowNM(tokenSecretManager);

  try {
  Job job = app.submit(conf);
  app.waitForState(job, JobState.RUNNING);

  Map<TaskId, Task> tasks = job.getTasks();
  Assert.assertEquals("Num tasks is not correct", 1, tasks.size());

  Task task = tasks.values().iterator().next();
  app.waitForState(task, TaskState.SCHEDULED);

  Map<TaskAttemptId, TaskAttempt> attempts = tasks.values().iterator()
      .next().getAttempts();
    Assert.assertEquals("Num attempts is not correct", maxAttempts,
        attempts.size());

  TaskAttempt attempt = attempts.values().iterator().next();
    app.waitForInternalState((TaskAttemptImpl) attempt,
        TaskAttemptStateInternal.ASSIGNED);

  app.waitForState(job, JobState.FAILED);

  String diagnostics = attempt.getDiagnostics().toString();
  LOG.info("attempt.getDiagnostics: " + diagnostics);

    Assert.assertTrue(diagnostics.contains("Container launch failed for "
        + "container_0_0000_01_000000 : "));
    Assert
        .assertTrue(diagnostics
            .contains("java.net.SocketTimeoutException: 3000 millis timeout while waiting for channel"));

  } finally {
    server.stop();
  app.stop();
}
}

Java Code Examples for org.apache.hadoop.yarn.ipc.YarnRPC#getServer()