public virtual void TestReconnect()
{
string hostname1 = "localhost1";
Resource capability = BuilderUtils.NewResource(1024, 1);
RegisterNodeManagerRequest request1 = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
NodeId nodeId1 = NodeId.NewInstance(hostname1, 0);
request1.SetNodeId(nodeId1);
request1.SetHttpPort(0);
request1.SetResource(capability);
resourceTrackerService.RegisterNodeManager(request1);
NUnit.Framework.Assert.AreEqual(RMNodeEventType.Started, rmNodeEvents[0].GetType(
));
rmNodeEvents.Clear();
resourceTrackerService.RegisterNodeManager(request1);
NUnit.Framework.Assert.AreEqual(RMNodeEventType.Reconnected, rmNodeEvents[0].GetType
());
rmNodeEvents.Clear();
resourceTrackerService.RegisterNodeManager(request1);
capability = BuilderUtils.NewResource(1024, 2);
request1.SetResource(capability);
NUnit.Framework.Assert.AreEqual(RMNodeEventType.Reconnected, rmNodeEvents[0].GetType
());
}
public virtual void TestNodeRegistrationVersionLessThanRM()
{
WriteToHostsFile("host2");
Configuration conf = new Configuration();
conf.Set(YarnConfiguration.RmNodesIncludeFilePath, hostFile.GetAbsolutePath());
conf.Set(YarnConfiguration.RmNodemanagerMinimumVersion, "EqualToRM");
rm = new MockRM(conf);
rm.Start();
string nmVersion = "1.9.9";
ResourceTrackerService resourceTrackerService = rm.GetResourceTrackerService();
RegisterNodeManagerRequest req = Org.Apache.Hadoop.Yarn.Util.Records.NewRecord <RegisterNodeManagerRequest
>();
NodeId nodeId = NodeId.NewInstance("host2", 1234);
Resource capability = BuilderUtils.NewResource(1024, 1);
req.SetResource(capability);
req.SetNodeId(nodeId);
req.SetHttpPort(1234);
req.SetNMVersion(nmVersion);
// trying to register a invalid node.
RegisterNodeManagerResponse response = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Shutdown, response.GetNodeAction());
NUnit.Framework.Assert.IsTrue("Diagnostic message did not contain: 'Disallowed NodeManager "
+ "Version " + nmVersion + ", is less than the minimum version'", response.GetDiagnosticsMessage
().Contains("Disallowed NodeManager Version " + nmVersion + ", is less than the minimum version "
));
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public override RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
this._enclosing.ResetStartFailoverFlag(true);
// make sure failover has been triggered
NUnit.Framework.Assert.IsTrue(this._enclosing.WaittingForFailOver());
return(base.RegisterNodeManager(request));
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
if (exception)
{
throw new YarnException("testMessage");
}
return(recordFactory.NewRecordInstance <RegisterNodeManagerResponse>());
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
RegisterNodeManagerResponse response = recordFactory.NewRecordInstance <RegisterNodeManagerResponse
>();
MasterKey masterKey = new MasterKeyPBImpl();
masterKey.SetKeyId(123);
masterKey.SetBytes(ByteBuffer.Wrap(new byte[] { 123 }));
response.SetContainerTokenMasterKey(masterKey);
response.SetNMTokenMasterKey(masterKey);
return(response);
}
public virtual void TestCompareRMNodeAfterReconnect()
{
Configuration yarnConf = new YarnConfiguration();
CapacityScheduler scheduler = new CapacityScheduler();
scheduler.SetConf(yarnConf);
ConfigurationProvider configurationProvider = ConfigurationProviderFactory.GetConfigurationProvider
(yarnConf);
configurationProvider.Init(yarnConf);
context.SetConfigurationProvider(configurationProvider);
RMNodeLabelsManager nlm = new RMNodeLabelsManager();
nlm.Init(yarnConf);
nlm.Start();
context.SetNodeLabelManager(nlm);
scheduler.SetRMContext(context);
scheduler.Init(yarnConf);
scheduler.Start();
dispatcher.Register(typeof(SchedulerEventType), scheduler);
string hostname1 = "localhost1";
Resource capability = BuilderUtils.NewResource(4096, 4);
RegisterNodeManagerRequest request1 = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
NodeId nodeId1 = NodeId.NewInstance(hostname1, 0);
request1.SetNodeId(nodeId1);
request1.SetHttpPort(0);
request1.SetResource(capability);
resourceTrackerService.RegisterNodeManager(request1);
NUnit.Framework.Assert.IsNotNull(context.GetRMNodes()[nodeId1]);
// verify Scheduler and RMContext use same RMNode reference.
NUnit.Framework.Assert.IsTrue(scheduler.GetSchedulerNode(nodeId1).GetRMNode() ==
context.GetRMNodes()[nodeId1]);
NUnit.Framework.Assert.AreEqual(context.GetRMNodes()[nodeId1].GetTotalCapability(
), capability);
Resource capability1 = BuilderUtils.NewResource(2048, 2);
request1.SetResource(capability1);
resourceTrackerService.RegisterNodeManager(request1);
NUnit.Framework.Assert.IsNotNull(context.GetRMNodes()[nodeId1]);
// verify Scheduler and RMContext use same RMNode reference
// after reconnect.
NUnit.Framework.Assert.IsTrue(scheduler.GetSchedulerNode(nodeId1).GetRMNode() ==
context.GetRMNodes()[nodeId1]);
// verify RMNode's capability is changed.
NUnit.Framework.Assert.AreEqual(context.GetRMNodes()[nodeId1].GetTotalCapability(
), capability1);
nlm.Stop();
scheduler.Stop();
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
YarnServerCommonServiceProtos.RegisterNodeManagerRequestProto requestProto = ((RegisterNodeManagerRequestPBImpl
)request).GetProto();
try
{
return(new RegisterNodeManagerResponsePBImpl(proxy.RegisterNodeManager(null, requestProto
)));
}
catch (ServiceException e)
{
RPCUtil.UnwrapAndThrowException(e);
return(null);
}
}
public virtual void TestRPCResponseId()
{
string node = "localhost";
Resource capability = BuilderUtils.NewResource(1024, 1);
RegisterNodeManagerRequest request = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
nodeId = NodeId.NewInstance(node, 1234);
request.SetNodeId(nodeId);
request.SetHttpPort(0);
request.SetResource(capability);
RegisterNodeManagerRequest request1 = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
request1.SetNodeId(nodeId);
request1.SetHttpPort(0);
request1.SetResource(capability);
resourceTrackerService.RegisterNodeManager(request1);
NodeStatus nodeStatus = recordFactory.NewRecordInstance <NodeStatus>();
nodeStatus.SetNodeId(nodeId);
NodeHealthStatus nodeHealthStatus = recordFactory.NewRecordInstance <NodeHealthStatus
>();
nodeHealthStatus.SetIsNodeHealthy(true);
nodeStatus.SetNodeHealthStatus(nodeHealthStatus);
NodeHeartbeatRequest nodeHeartBeatRequest = recordFactory.NewRecordInstance <NodeHeartbeatRequest
>();
nodeHeartBeatRequest.SetNodeStatus(nodeStatus);
nodeStatus.SetResponseId(0);
NodeHeartbeatResponse response = resourceTrackerService.NodeHeartbeat(nodeHeartBeatRequest
);
NUnit.Framework.Assert.IsTrue(response.GetResponseId() == 1);
nodeStatus.SetResponseId(response.GetResponseId());
response = resourceTrackerService.NodeHeartbeat(nodeHeartBeatRequest);
NUnit.Framework.Assert.IsTrue(response.GetResponseId() == 2);
/* try calling with less response id */
response = resourceTrackerService.NodeHeartbeat(nodeHeartBeatRequest);
NUnit.Framework.Assert.IsTrue(response.GetResponseId() == 2);
nodeStatus.SetResponseId(0);
response = resourceTrackerService.NodeHeartbeat(nodeHeartBeatRequest);
NUnit.Framework.Assert.IsTrue(NodeAction.Resync.Equals(response.GetNodeAction()));
NUnit.Framework.Assert.AreEqual("Too far behind rm response id:2 nm response id:0"
, response.GetDiagnosticsMessage());
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
RegisterNodeManagerResponse response;
try
{
response = rt.RegisterNodeManager(request);
}
catch (YarnException e)
{
MiniYARNCluster.Log.Info("Exception in node registration from " + request.GetNodeId
().ToString(), e);
throw;
}
return(response);
}
public virtual void TestNMExpiry()
{
string hostname1 = "localhost1";
string hostname2 = "localhost2";
string hostname3 = "localhost3";
Resource capability = BuilderUtils.NewResource(1024, 1);
RegisterNodeManagerRequest request1 = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
NodeId nodeId1 = NodeId.NewInstance(hostname1, 0);
request1.SetNodeId(nodeId1);
request1.SetHttpPort(0);
request1.SetResource(capability);
resourceTrackerService.RegisterNodeManager(request1);
RegisterNodeManagerRequest request2 = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
NodeId nodeId2 = NodeId.NewInstance(hostname2, 0);
request2.SetNodeId(nodeId2);
request2.SetHttpPort(0);
request2.SetResource(capability);
resourceTrackerService.RegisterNodeManager(request2);
int waitCount = 0;
while (ClusterMetrics.GetMetrics().GetNumLostNMs() != 2 && waitCount++ < 20)
{
lock (this)
{
Sharpen.Runtime.Wait(this, 100);
}
}
NUnit.Framework.Assert.AreEqual(2, ClusterMetrics.GetMetrics().GetNumLostNMs());
request3 = recordFactory.NewRecordInstance <RegisterNodeManagerRequest>();
NodeId nodeId3 = NodeId.NewInstance(hostname3, 0);
request3.SetNodeId(nodeId3);
request3.SetHttpPort(0);
request3.SetResource(capability);
resourceTrackerService.RegisterNodeManager(request3);
/* test to see if hostanme 3 does not expire */
stopT = false;
new TestNMExpiry.ThirdNodeHeartBeatThread(this).Start();
NUnit.Framework.Assert.AreEqual(2, ClusterMetrics.GetMetrics().GetNumLostNMs());
stopT = true;
}
/// <exception cref="System.Exception"/>
public virtual void TestResourceTrackerOnHA()
{
NodeId nodeId = NodeId.NewInstance("localhost", 0);
Resource resource = Resource.NewInstance(2048, 4);
// make sure registerNodeManager works when failover happens
RegisterNodeManagerRequest request = RegisterNodeManagerRequest.NewInstance(nodeId
, 0, resource, YarnVersionInfo.GetVersion(), null, null);
resourceTracker.RegisterNodeManager(request);
NUnit.Framework.Assert.IsTrue(WaitForNodeManagerToConnect(10000, nodeId));
// restart the failover thread, and make sure nodeHeartbeat works
failoverThread = CreateAndStartFailoverThread();
NodeStatus status = NodeStatus.NewInstance(NodeId.NewInstance("localhost", 0), 0,
null, null, null);
NodeHeartbeatRequest request2 = NodeHeartbeatRequest.NewInstance(status, null, null
);
resourceTracker.NodeHeartbeat(request2);
}
public virtual void TestNodeRegistrationWithMinimumAllocations()
{
Configuration conf = new Configuration();
conf.Set(YarnConfiguration.RmSchedulerMinimumAllocationMb, "2048");
conf.Set(YarnConfiguration.RmSchedulerMinimumAllocationVcores, "4");
rm = new MockRM(conf);
rm.Start();
ResourceTrackerService resourceTrackerService = rm.GetResourceTrackerService();
RegisterNodeManagerRequest req = Org.Apache.Hadoop.Yarn.Util.Records.NewRecord <RegisterNodeManagerRequest
>();
NodeId nodeId = BuilderUtils.NewNodeId("host", 1234);
req.SetNodeId(nodeId);
Resource capability = BuilderUtils.NewResource(1024, 1);
req.SetResource(capability);
RegisterNodeManagerResponse response1 = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Shutdown, response1.GetNodeAction());
capability.SetMemory(2048);
capability.SetVirtualCores(1);
req.SetResource(capability);
RegisterNodeManagerResponse response2 = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Shutdown, response2.GetNodeAction());
capability.SetMemory(1024);
capability.SetVirtualCores(4);
req.SetResource(capability);
RegisterNodeManagerResponse response3 = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Shutdown, response3.GetNodeAction());
capability.SetMemory(2048);
capability.SetVirtualCores(4);
req.SetResource(capability);
RegisterNodeManagerResponse response4 = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Normal, response4.GetNodeAction());
}
public virtual void TestResourceTrackerPBClientImpl()
{
RegisterNodeManagerRequest request = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
NUnit.Framework.Assert.IsNotNull(client.RegisterNodeManager(request));
TestResourceTrackerPBClientImpl.ResourceTrackerTestImpl.exception = true;
try
{
client.RegisterNodeManager(request);
NUnit.Framework.Assert.Fail("there should be YarnException");
}
catch (YarnException e)
{
NUnit.Framework.Assert.IsTrue(e.Message.StartsWith("testMessage"));
}
finally
{
TestResourceTrackerPBClientImpl.ResourceTrackerTestImpl.exception = false;
}
}
/// <exception cref="System.IO.IOException"/>
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
public NodeManager(string hostName, int containerManagerPort, int httpPort, string
rackName, Resource capability, ResourceManager resourceManager)
{
this.containerManagerAddress = hostName + ":" + containerManagerPort;
this.nodeHttpAddress = hostName + ":" + httpPort;
this.rackName = rackName;
this.resourceTrackerService = resourceManager.GetResourceTrackerService();
this.capability = capability;
Resources.AddTo(available, capability);
this.nodeId = NodeId.NewInstance(hostName, containerManagerPort);
RegisterNodeManagerRequest request = recordFactory.NewRecordInstance <RegisterNodeManagerRequest
>();
request.SetHttpPort(httpPort);
request.SetResource(capability);
request.SetNodeId(this.nodeId);
request.SetNMVersion(YarnVersionInfo.GetVersion());
resourceTrackerService.RegisterNodeManager(request);
this.resourceManager = resourceManager;
resourceManager.GetResourceScheduler().GetNodeReport(this.nodeId);
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public override RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
if (this._enclosing._enclosing.registerCount == 0)
{
try
{
NUnit.Framework.Assert.AreEqual(0, request.GetNMContainerStatuses().Count);
}
catch (Exception error)
{
Sharpen.Runtime.PrintStackTrace(error);
this._enclosing._enclosing._enclosing.assertionFailedInThread.Set(true);
}
this._enclosing._enclosing.GetNMContext().GetContainers()[testCompleteContainer.GetContainerId
()] = container;
this._enclosing._enclosing.GetNMContext().GetApplications()[testCompleteContainer
.GetContainerId().GetApplicationAttemptId().GetApplicationId()] = Org.Mockito.Mockito
.Mock <Org.Apache.Hadoop.Yarn.Server.Nodemanager.Containermanager.Application.Application
>();
}
else
{
IList <NMContainerStatus> statuses = request.GetNMContainerStatuses();
try
{
NUnit.Framework.Assert.AreEqual(1, statuses.Count);
NUnit.Framework.Assert.AreEqual(testCompleteContainer.GetContainerId(), statuses[
0].GetContainerId());
}
catch (Exception error)
{
Sharpen.Runtime.PrintStackTrace(error);
this._enclosing._enclosing._enclosing.assertionFailedInThread.Set(true);
}
}
this._enclosing._enclosing.registerCount++;
return(base.RegisterNodeManager(request));
}
/// <exception cref="System.Exception"/>
public virtual RegisterNodeManagerResponse RegisterNode(IList <NMContainerStatus>
containerReports, IList <ApplicationId> runningApplications)
{
RegisterNodeManagerRequest req = Org.Apache.Hadoop.Yarn.Util.Records.NewRecord <RegisterNodeManagerRequest
>();
req.SetNodeId(nodeId);
req.SetHttpPort(httpPort);
Resource resource = BuilderUtils.NewResource(memory, vCores);
req.SetResource(resource);
req.SetContainerStatuses(containerReports);
req.SetNMVersion(version);
req.SetRunningApplications(runningApplications);
RegisterNodeManagerResponse registrationResponse = resourceTracker.RegisterNodeManager
(req);
this.currentContainerTokenMasterKey = registrationResponse.GetContainerTokenMasterKey
();
this.currentNMTokenMasterKey = registrationResponse.GetNMTokenMasterKey();
return(registrationResponse);
}
public virtual void TestNodeRegistrationFailure()
{
WriteToHostsFile("host1");
Configuration conf = new Configuration();
conf.Set(YarnConfiguration.RmNodesIncludeFilePath, hostFile.GetAbsolutePath());
rm = new MockRM(conf);
rm.Start();
ResourceTrackerService resourceTrackerService = rm.GetResourceTrackerService();
RegisterNodeManagerRequest req = Org.Apache.Hadoop.Yarn.Util.Records.NewRecord <RegisterNodeManagerRequest
>();
NodeId nodeId = NodeId.NewInstance("host2", 1234);
req.SetNodeId(nodeId);
req.SetHttpPort(1234);
// trying to register a invalid node.
RegisterNodeManagerResponse response = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Shutdown, response.GetNodeAction());
NUnit.Framework.Assert.AreEqual("Disallowed NodeManager from host2, Sending SHUTDOWN signal to the NodeManager."
, response.GetDiagnosticsMessage());
}
public virtual void TestNodeRegistrationSuccess()
{
WriteToHostsFile("host2");
Configuration conf = new Configuration();
conf.Set(YarnConfiguration.RmNodesIncludeFilePath, hostFile.GetAbsolutePath());
rm = new MockRM(conf);
rm.Start();
ResourceTrackerService resourceTrackerService = rm.GetResourceTrackerService();
RegisterNodeManagerRequest req = Org.Apache.Hadoop.Yarn.Util.Records.NewRecord <RegisterNodeManagerRequest
>();
NodeId nodeId = NodeId.NewInstance("host2", 1234);
Resource capability = BuilderUtils.NewResource(1024, 1);
req.SetResource(capability);
req.SetNodeId(nodeId);
req.SetHttpPort(1234);
req.SetNMVersion(YarnVersionInfo.GetVersion());
// trying to register a invalid node.
RegisterNodeManagerResponse response = resourceTrackerService.RegisterNodeManager
(req);
NUnit.Framework.Assert.AreEqual(NodeAction.Normal, response.GetNodeAction());
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
// TODO Auto-generated method stub
return(null);
}
protected internal virtual void RegisterWithRM()
{
IList <NMContainerStatus> containerReports = GetNMContainerStatuses();
RegisterNodeManagerRequest request = RegisterNodeManagerRequest.NewInstance(nodeId
, httpPort, totalResource, nodeManagerVersionId, containerReports, GetRunningApplications
());
if (containerReports != null)
{
Log.Info("Registering with RM using containers :" + containerReports);
}
RegisterNodeManagerResponse regNMResponse = resourceTracker.RegisterNodeManager(request
);
this.rmIdentifier = regNMResponse.GetRMIdentifier();
// if the Resourcemanager instructs NM to shutdown.
if (NodeAction.Shutdown.Equals(regNMResponse.GetNodeAction()))
{
string message = "Message from ResourceManager: " + regNMResponse.GetDiagnosticsMessage
();
throw new YarnRuntimeException("Recieved SHUTDOWN signal from Resourcemanager ,Registration of NodeManager failed, "
+ message);
}
// if ResourceManager version is too old then shutdown
if (!minimumResourceManagerVersion.Equals("NONE"))
{
if (minimumResourceManagerVersion.Equals("EqualToNM"))
{
minimumResourceManagerVersion = nodeManagerVersionId;
}
string rmVersion = regNMResponse.GetRMVersion();
if (rmVersion == null)
{
string message = "The Resource Manager's did not return a version. " + "Valid version cannot be checked.";
throw new YarnRuntimeException("Shutting down the Node Manager. " + message);
}
if (VersionUtil.CompareVersions(rmVersion, minimumResourceManagerVersion) < 0)
{
string message = "The Resource Manager's version (" + rmVersion + ") is less than the minimum "
+ "allowed version " + minimumResourceManagerVersion;
throw new YarnRuntimeException("Shutting down the Node Manager on RM " + "version error, "
+ message);
}
}
MasterKey masterKey = regNMResponse.GetContainerTokenMasterKey();
// do this now so that its set before we start heartbeating to RM
// It is expected that status updater is started by this point and
// RM gives the shared secret in registration during
// StatusUpdater#start().
if (masterKey != null)
{
this.context.GetContainerTokenSecretManager().SetMasterKey(masterKey);
}
masterKey = regNMResponse.GetNMTokenMasterKey();
if (masterKey != null)
{
this.context.GetNMTokenSecretManager().SetMasterKey(masterKey);
}
Log.Info("Registered with ResourceManager as " + this.nodeId + " with total resource of "
+ this.totalResource);
Log.Info("Notifying ContainerManager to unblock new container-requests");
((ContainerManagerImpl)this.context.GetContainerManager()).SetBlockNewContainerRequests
(false);
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual RegisterNodeManagerResponse RegisterNodeManager(RegisterNodeManagerRequest
request)
{
NodeId nodeId = request.GetNodeId();
string host = nodeId.GetHost();
int cmPort = nodeId.GetPort();
int httpPort = request.GetHttpPort();
Resource capability = request.GetResource();
string nodeManagerVersion = request.GetNMVersion();
RegisterNodeManagerResponse response = recordFactory.NewRecordInstance <RegisterNodeManagerResponse
>();
if (!minimumNodeManagerVersion.Equals("NONE"))
{
if (minimumNodeManagerVersion.Equals("EqualToRM"))
{
minimumNodeManagerVersion = YarnVersionInfo.GetVersion();
}
if ((nodeManagerVersion == null) || (VersionUtil.CompareVersions(nodeManagerVersion
, minimumNodeManagerVersion)) < 0)
{
string message = "Disallowed NodeManager Version " + nodeManagerVersion + ", is less than the minimum version "
+ minimumNodeManagerVersion + " sending SHUTDOWN signal to " + "NodeManager.";
Log.Info(message);
response.SetDiagnosticsMessage(message);
response.SetNodeAction(NodeAction.Shutdown);
return(response);
}
}
// Check if this node is a 'valid' node
if (!this.nodesListManager.IsValidNode(host))
{
string message = "Disallowed NodeManager from " + host + ", Sending SHUTDOWN signal to the NodeManager.";
Log.Info(message);
response.SetDiagnosticsMessage(message);
response.SetNodeAction(NodeAction.Shutdown);
return(response);
}
// Check if this node has minimum allocations
if (capability.GetMemory() < minAllocMb || capability.GetVirtualCores() < minAllocVcores)
{
string message = "NodeManager from " + host + " doesn't satisfy minimum allocations, Sending SHUTDOWN"
+ " signal to the NodeManager.";
Log.Info(message);
response.SetDiagnosticsMessage(message);
response.SetNodeAction(NodeAction.Shutdown);
return(response);
}
response.SetContainerTokenMasterKey(containerTokenSecretManager.GetCurrentKey());
response.SetNMTokenMasterKey(nmTokenSecretManager.GetCurrentKey());
RMNode rmNode = new RMNodeImpl(nodeId, rmContext, host, cmPort, httpPort, Resolve
(host), capability, nodeManagerVersion);
RMNode oldNode = this.rmContext.GetRMNodes().PutIfAbsent(nodeId, rmNode);
if (oldNode == null)
{
this.rmContext.GetDispatcher().GetEventHandler().Handle(new RMNodeStartedEvent(nodeId
, request.GetNMContainerStatuses(), request.GetRunningApplications()));
}
else
{
Log.Info("Reconnect from the node at: " + host);
this.nmLivelinessMonitor.Unregister(nodeId);
// Reset heartbeat ID since node just restarted.
oldNode.ResetLastNodeHeartBeatResponse();
this.rmContext.GetDispatcher().GetEventHandler().Handle(new RMNodeReconnectEvent(
nodeId, rmNode, request.GetRunningApplications(), request.GetNMContainerStatuses
()));
}
// On every node manager register we will be clearing NMToken keys if
// present for any running application.
this.nmTokenSecretManager.RemoveNodeKey(nodeId);
this.nmLivelinessMonitor.Register(nodeId);
// Handle received container status, this should be processed after new
// RMNode inserted
if (!rmContext.IsWorkPreservingRecoveryEnabled())
{
if (!request.GetNMContainerStatuses().IsEmpty())
{
Log.Info("received container statuses on node manager register :" + request.GetNMContainerStatuses
());
foreach (NMContainerStatus status in request.GetNMContainerStatuses())
{
HandleNMContainerStatus(status, nodeId);
}
}
}
string message_1 = "NodeManager from node " + host + "(cmPort: " + cmPort + " httpPort: "
+ httpPort + ") " + "registered with capability: " + capability + ", assigned nodeId "
+ nodeId;
Log.Info(message_1);
response.SetNodeAction(NodeAction.Normal);
response.SetRMIdentifier(ResourceManager.GetClusterTimeStamp());
response.SetRMVersion(YarnVersionInfo.GetVersion());
return(response);
}
