/// <exception cref="System.Exception"/>
protected internal virtual void RollNMTokenMasterKey(NMTokenSecretManagerInRM nmTokenSecretManagerRM
, NMTokenSecretManagerInNM nmTokenSecretManagerNM)
{
int oldKeyId = nmTokenSecretManagerRM.GetCurrentKey().GetKeyId();
nmTokenSecretManagerRM.RollMasterKey();
int interval = 40;
while (nmTokenSecretManagerNM.GetCurrentKey().GetKeyId() == oldKeyId && interval--
> 0)
{
Sharpen.Thread.Sleep(1000);
}
nmTokenSecretManagerRM.ActivateNextMasterKey();
NUnit.Framework.Assert.IsTrue((nmTokenSecretManagerNM.GetCurrentKey().GetKeyId()
== nmTokenSecretManagerRM.GetCurrentKey().GetKeyId()));
}
/// <summary>This tests whether a containerId is serialized/deserialized with epoch.</summary>
/// <exception cref="System.IO.IOException"/>
/// <exception cref="System.Exception"/>
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
private void TestContainerTokenWithEpoch(Configuration conf)
{
Log.Info("Running test for serializing/deserializing containerIds");
NMTokenSecretManagerInRM nmTokenSecretManagerInRM = yarnCluster.GetResourceManager
().GetRMContext().GetNMTokenSecretManager();
ApplicationId appId = ApplicationId.NewInstance(1, 1);
ApplicationAttemptId appAttemptId = ApplicationAttemptId.NewInstance(appId, 0);
ContainerId cId = ContainerId.NewContainerId(appAttemptId, (5L << 40) | 3L);
NodeManager nm = yarnCluster.GetNodeManager(0);
NMTokenSecretManagerInNM nmTokenSecretManagerInNM = nm.GetNMContext().GetNMTokenSecretManager
();
string user = "******";
WaitForNMToReceiveNMTokenKey(nmTokenSecretManagerInNM, nm);
NodeId nodeId = nm.GetNMContext().GetNodeId();
// Both id should be equal.
NUnit.Framework.Assert.AreEqual(nmTokenSecretManagerInNM.GetCurrentKey().GetKeyId
(), nmTokenSecretManagerInRM.GetCurrentKey().GetKeyId());
// Creating a normal Container Token
RMContainerTokenSecretManager containerTokenSecretManager = yarnCluster.GetResourceManager
().GetRMContext().GetContainerTokenSecretManager();
Resource r = Resource.NewInstance(1230, 2);
Org.Apache.Hadoop.Yarn.Api.Records.Token containerToken = containerTokenSecretManager
.CreateContainerToken(cId, nodeId, user, r, Priority.NewInstance(0), 0);
ContainerTokenIdentifier containerTokenIdentifier = new ContainerTokenIdentifier(
);
byte[] tokenIdentifierContent = ((byte[])containerToken.GetIdentifier().Array());
DataInputBuffer dib = new DataInputBuffer();
dib.Reset(tokenIdentifierContent, tokenIdentifierContent.Length);
containerTokenIdentifier.ReadFields(dib);
NUnit.Framework.Assert.AreEqual(cId, containerTokenIdentifier.GetContainerID());
NUnit.Framework.Assert.AreEqual(cId.ToString(), containerTokenIdentifier.GetContainerID
().ToString());
Org.Apache.Hadoop.Yarn.Api.Records.Token nmToken = nmTokenSecretManagerInRM.CreateNMToken
(appAttemptId, nodeId, user);
YarnRPC rpc = YarnRPC.Create(conf);
TestStartContainer(rpc, appAttemptId, nodeId, containerToken, nmToken, false);
IList <ContainerId> containerIds = new List <ContainerId>();
containerIds.AddItem(cId);
ContainerManagementProtocol proxy = GetContainerManagementProtocolProxy(rpc, nmToken
, nodeId, user);
GetContainerStatusesResponse res = proxy.GetContainerStatuses(GetContainerStatusesRequest
.NewInstance(containerIds));
NUnit.Framework.Assert.IsNotNull(res.GetContainerStatuses()[0]);
NUnit.Framework.Assert.AreEqual(cId, res.GetContainerStatuses()[0].GetContainerId
());
NUnit.Framework.Assert.AreEqual(cId.ToString(), res.GetContainerStatuses()[0].GetContainerId
().ToString());
}
/// <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);
}
/// <summary>
/// This tests a malice user getting a proper token but then messing with it by
/// tampering with containerID/Resource etc..
/// </summary>
/// <remarks>
/// This tests a malice user getting a proper token but then messing with it by
/// tampering with containerID/Resource etc.. His/her containers should be
/// rejected.
/// </remarks>
/// <exception cref="System.IO.IOException"/>
/// <exception cref="System.Exception"/>
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
private void TestContainerToken(Configuration conf)
{
Log.Info("Running test for malice user");
/*
* We need to check for containerToken (authorization).
* Here we will be assuming that we have valid NMToken
* 1) ContainerToken used is expired.
* 2) ContainerToken is tampered (resource is modified).
*/
NMTokenSecretManagerInRM nmTokenSecretManagerInRM = yarnCluster.GetResourceManager
().GetRMContext().GetNMTokenSecretManager();
ApplicationId appId = ApplicationId.NewInstance(1, 1);
ApplicationAttemptId appAttemptId = ApplicationAttemptId.NewInstance(appId, 0);
ContainerId cId = ContainerId.NewContainerId(appAttemptId, 0);
NodeManager nm = yarnCluster.GetNodeManager(0);
NMTokenSecretManagerInNM nmTokenSecretManagerInNM = nm.GetNMContext().GetNMTokenSecretManager
();
string user = "******";
WaitForNMToReceiveNMTokenKey(nmTokenSecretManagerInNM, nm);
NodeId nodeId = nm.GetNMContext().GetNodeId();
// Both id should be equal.
NUnit.Framework.Assert.AreEqual(nmTokenSecretManagerInNM.GetCurrentKey().GetKeyId
(), nmTokenSecretManagerInRM.GetCurrentKey().GetKeyId());
RMContainerTokenSecretManager containerTokenSecretManager = yarnCluster.GetResourceManager
().GetRMContext().GetContainerTokenSecretManager();
Resource r = Resource.NewInstance(1230, 2);
Org.Apache.Hadoop.Yarn.Api.Records.Token containerToken = containerTokenSecretManager
.CreateContainerToken(cId, nodeId, user, r, Priority.NewInstance(0), 0);
ContainerTokenIdentifier containerTokenIdentifier = GetContainerTokenIdentifierFromToken
(containerToken);
// Verify new compatible version ContainerTokenIdentifier can work successfully.
ContainerTokenIdentifierForTest newVersionTokenIdentifier = new ContainerTokenIdentifierForTest
(containerTokenIdentifier, "message");
byte[] password = containerTokenSecretManager.CreatePassword(newVersionTokenIdentifier
);
Org.Apache.Hadoop.Yarn.Api.Records.Token newContainerToken = BuilderUtils.NewContainerToken
(nodeId, password, newVersionTokenIdentifier);
Org.Apache.Hadoop.Yarn.Api.Records.Token nmToken = nmTokenSecretManagerInRM.CreateNMToken
(appAttemptId, nodeId, user);
YarnRPC rpc = YarnRPC.Create(conf);
NUnit.Framework.Assert.IsTrue(TestStartContainer(rpc, appAttemptId, nodeId, newContainerToken
, nmToken, false).IsEmpty());
// Creating a tampered Container Token
RMContainerTokenSecretManager tamperedContainerTokenSecretManager = new RMContainerTokenSecretManager
(conf);
tamperedContainerTokenSecretManager.RollMasterKey();
do
{
tamperedContainerTokenSecretManager.RollMasterKey();
tamperedContainerTokenSecretManager.ActivateNextMasterKey();
}while (containerTokenSecretManager.GetCurrentKey().GetKeyId() == tamperedContainerTokenSecretManager
.GetCurrentKey().GetKeyId());
ContainerId cId2 = ContainerId.NewContainerId(appAttemptId, 1);
// Creating modified containerToken
Org.Apache.Hadoop.Yarn.Api.Records.Token containerToken2 = tamperedContainerTokenSecretManager
.CreateContainerToken(cId2, nodeId, user, r, Priority.NewInstance(0), 0);
StringBuilder sb = new StringBuilder("Given Container ");
sb.Append(cId2);
sb.Append(" seems to have an illegally generated token.");
NUnit.Framework.Assert.IsTrue(TestStartContainer(rpc, appAttemptId, nodeId, containerToken2
, nmToken, true).Contains(sb.ToString()));
}
/// <exception cref="System.Exception"/>
private void TestNMTokens(Configuration conf)
{
NMTokenSecretManagerInRM nmTokenSecretManagerRM = yarnCluster.GetResourceManager(
).GetRMContext().GetNMTokenSecretManager();
NMTokenSecretManagerInNM nmTokenSecretManagerNM = yarnCluster.GetNodeManager(0).GetNMContext
().GetNMTokenSecretManager();
RMContainerTokenSecretManager containerTokenSecretManager = yarnCluster.GetResourceManager
().GetRMContext().GetContainerTokenSecretManager();
NodeManager nm = yarnCluster.GetNodeManager(0);
WaitForNMToReceiveNMTokenKey(nmTokenSecretManagerNM, nm);
// Both id should be equal.
NUnit.Framework.Assert.AreEqual(nmTokenSecretManagerNM.GetCurrentKey().GetKeyId()
, nmTokenSecretManagerRM.GetCurrentKey().GetKeyId());
/*
* Below cases should be tested.
* 1) If Invalid NMToken is used then it should be rejected.
* 2) If valid NMToken but belonging to another Node is used then that
* too should be rejected.
* 3) NMToken for say appAttempt-1 is used for starting/stopping/retrieving
* status for container with containerId for say appAttempt-2 should
* be rejected.
* 4) After start container call is successful nmtoken should have been
* saved in NMTokenSecretManagerInNM.
* 5) If start container call was successful (no matter if container is
* still running or not), appAttempt->NMToken should be present in
* NMTokenSecretManagerInNM's cache. Any future getContainerStatus call
* for containerId belonging to that application attempt using
* applicationAttempt's older nmToken should not get any invalid
* nmToken error. (This can be best tested if we roll over NMToken
* master key twice).
*/
YarnRPC rpc = YarnRPC.Create(conf);
string user = "******";
Resource r = Resource.NewInstance(1024, 1);
ApplicationId appId = ApplicationId.NewInstance(1, 1);
ApplicationAttemptId validAppAttemptId = ApplicationAttemptId.NewInstance(appId,
1);
ContainerId validContainerId = ContainerId.NewContainerId(validAppAttemptId, 0);
NodeId validNode = yarnCluster.GetNodeManager(0).GetNMContext().GetNodeId();
NodeId invalidNode = NodeId.NewInstance("InvalidHost", 1234);
Token validNMToken = nmTokenSecretManagerRM.CreateNMToken(validAppAttemptId, validNode
, user);
Token validContainerToken = containerTokenSecretManager.CreateContainerToken(validContainerId
, validNode, user, r, Priority.NewInstance(10), 1234);
ContainerTokenIdentifier identifier = BuilderUtils.NewContainerTokenIdentifier(validContainerToken
);
NUnit.Framework.Assert.AreEqual(Priority.NewInstance(10), identifier.GetPriority(
));
NUnit.Framework.Assert.AreEqual(1234, identifier.GetCreationTime());
StringBuilder sb;
// testInvalidNMToken ... creating NMToken using different secret manager.
NMTokenSecretManagerInRM tempManager = new NMTokenSecretManagerInRM(conf);
tempManager.RollMasterKey();
do
{
tempManager.RollMasterKey();
tempManager.ActivateNextMasterKey();
}while (tempManager.GetCurrentKey().GetKeyId() == nmTokenSecretManagerRM.GetCurrentKey
().GetKeyId());
// Making sure key id is different.
// Testing that NM rejects the requests when we don't send any token.
if (UserGroupInformation.IsSecurityEnabled())
{
sb = new StringBuilder("Client cannot authenticate via:[TOKEN]");
}
else
{
sb = new StringBuilder("SIMPLE authentication is not enabled. Available:[TOKEN]"
);
}
string errorMsg = TestStartContainer(rpc, validAppAttemptId, validNode, validContainerToken
, null, true);
NUnit.Framework.Assert.IsTrue(errorMsg.Contains(sb.ToString()));
Token invalidNMToken = tempManager.CreateNMToken(validAppAttemptId, validNode, user
);
sb = new StringBuilder("Given NMToken for application : ");
sb.Append(validAppAttemptId.ToString()).Append(" seems to have been generated illegally."
);
NUnit.Framework.Assert.IsTrue(sb.ToString().Contains(TestStartContainer(rpc, validAppAttemptId
, validNode, validContainerToken, invalidNMToken, true)));
// valid NMToken but belonging to other node
invalidNMToken = nmTokenSecretManagerRM.CreateNMToken(validAppAttemptId, invalidNode
, user);
sb = new StringBuilder("Given NMToken for application : ");
sb.Append(validAppAttemptId).Append(" is not valid for current node manager.expected : "
).Append(validNode.ToString()).Append(" found : ").Append(invalidNode.ToString()
);
NUnit.Framework.Assert.IsTrue(sb.ToString().Contains(TestStartContainer(rpc, validAppAttemptId
, validNode, validContainerToken, invalidNMToken, true)));
// using correct tokens. nmtoken for app attempt should get saved.
conf.SetInt(YarnConfiguration.RmContainerAllocExpiryIntervalMs, 4 * 60 * 1000);
validContainerToken = containerTokenSecretManager.CreateContainerToken(validContainerId
, validNode, user, r, Priority.NewInstance(0), 0);
NUnit.Framework.Assert.IsTrue(TestStartContainer(rpc, validAppAttemptId, validNode
, validContainerToken, validNMToken, false).IsEmpty());
NUnit.Framework.Assert.IsTrue(nmTokenSecretManagerNM.IsAppAttemptNMTokenKeyPresent
(validAppAttemptId));
// using a new compatible version nmtoken, expect container can be started
// successfully.
ApplicationAttemptId validAppAttemptId2 = ApplicationAttemptId.NewInstance(appId,
2);
ContainerId validContainerId2 = ContainerId.NewContainerId(validAppAttemptId2, 0);
Token validContainerToken2 = containerTokenSecretManager.CreateContainerToken(validContainerId2
, validNode, user, r, Priority.NewInstance(0), 0);
Token validNMToken2 = nmTokenSecretManagerRM.CreateNMToken(validAppAttemptId2, validNode
, user);
// First, get a new NMTokenIdentifier.
NMTokenIdentifier newIdentifier = new NMTokenIdentifier();
byte[] tokenIdentifierContent = ((byte[])validNMToken2.GetIdentifier().Array());
DataInputBuffer dib = new DataInputBuffer();
dib.Reset(tokenIdentifierContent, tokenIdentifierContent.Length);
newIdentifier.ReadFields(dib);
// Then, generate a new version NMTokenIdentifier (NMTokenIdentifierNewForTest)
// with additional field of message.
NMTokenIdentifierNewForTest newVersionIdentifier = new NMTokenIdentifierNewForTest
(newIdentifier, "message");
// check new version NMTokenIdentifier has correct info.
NUnit.Framework.Assert.AreEqual("The ApplicationAttemptId is changed after set to "
+ "newVersionIdentifier", validAppAttemptId2.GetAttemptId(), newVersionIdentifier
.GetApplicationAttemptId().GetAttemptId());
NUnit.Framework.Assert.AreEqual("The message is changed after set to newVersionIdentifier"
, "message", newVersionIdentifier.GetMessage());
NUnit.Framework.Assert.AreEqual("The NodeId is changed after set to newVersionIdentifier"
, validNode, newVersionIdentifier.GetNodeId());
// create new Token based on new version NMTokenIdentifier.
Token newVersionedNMToken = BaseNMTokenSecretManager.NewInstance(nmTokenSecretManagerRM
.RetrievePassword(newVersionIdentifier), newVersionIdentifier);
// Verify startContainer is successful and no exception is thrown.
NUnit.Framework.Assert.IsTrue(TestStartContainer(rpc, validAppAttemptId2, validNode
, validContainerToken2, newVersionedNMToken, false).IsEmpty());
NUnit.Framework.Assert.IsTrue(nmTokenSecretManagerNM.IsAppAttemptNMTokenKeyPresent
(validAppAttemptId2));
//Now lets wait till container finishes and is removed from node manager.
WaitForContainerToFinishOnNM(validContainerId);
sb = new StringBuilder("Attempt to relaunch the same container with id ");
sb.Append(validContainerId);
NUnit.Framework.Assert.IsTrue(TestStartContainer(rpc, validAppAttemptId, validNode
, validContainerToken, validNMToken, true).Contains(sb.ToString()));
// Container is removed from node manager's memory by this time.
// trying to stop the container. It should not throw any exception.
TestStopContainer(rpc, validAppAttemptId, validNode, validContainerId, validNMToken
, false);
// Rolling over master key twice so that we can check whether older keys
// are used for authentication.
RollNMTokenMasterKey(nmTokenSecretManagerRM, nmTokenSecretManagerNM);
// Key rolled over once.. rolling over again
RollNMTokenMasterKey(nmTokenSecretManagerRM, nmTokenSecretManagerNM);
// trying get container status. Now saved nmToken should be used for
// authentication... It should complain saying container was recently
// stopped.
sb = new StringBuilder("Container ");
sb.Append(validContainerId);
sb.Append(" was recently stopped on node manager");
NUnit.Framework.Assert.IsTrue(TestGetContainer(rpc, validAppAttemptId, validNode,
validContainerId, validNMToken, true).Contains(sb.ToString()));
// Now lets remove the container from nm-memory
nm.GetNodeStatusUpdater().ClearFinishedContainersFromCache();
// This should fail as container is removed from recently tracked finished
// containers.
sb = new StringBuilder("Container ");
sb.Append(validContainerId.ToString());
sb.Append(" is not handled by this NodeManager");
NUnit.Framework.Assert.IsTrue(TestGetContainer(rpc, validAppAttemptId, validNode,
validContainerId, validNMToken, false).Contains(sb.ToString()));
// using appAttempt-1 NMtoken for launching container for appAttempt-2 should
// succeed.
ApplicationAttemptId attempt2 = ApplicationAttemptId.NewInstance(appId, 2);
Token attempt1NMToken = nmTokenSecretManagerRM.CreateNMToken(validAppAttemptId, validNode
, user);
Token newContainerToken = containerTokenSecretManager.CreateContainerToken(ContainerId
.NewContainerId(attempt2, 1), validNode, user, r, Priority.NewInstance(0), 0);
NUnit.Framework.Assert.IsTrue(TestStartContainer(rpc, attempt2, validNode, newContainerToken
, attempt1NMToken, false).IsEmpty());
}