protected override void GenerateApplicationTable(HtmlBlock.Block html, UserGroupInformation
callerUGI, ICollection <ApplicationAttemptReport> attempts)
{
// Application Attempt Table
Hamlet.TBODY <Hamlet.TABLE <Org.Apache.Hadoop.Yarn.Webapp.Hamlet.Hamlet> > tbody = html
.Table("#attempts").Thead().Tr().Th(".id", "Attempt ID").Th(".started", "Started"
).Th(".node", "Node").Th(".logs", "Logs").Th(".blacklistednodes", "Blacklisted Nodes"
).().().Tbody();
RMApp rmApp = this.rm.GetRMContext().GetRMApps()[this.appID];
if (rmApp == null)
{
return;
}
StringBuilder attemptsTableData = new StringBuilder("[\n");
foreach (ApplicationAttemptReport appAttemptReport in attempts)
{
RMAppAttempt rmAppAttempt = rmApp.GetRMAppAttempt(appAttemptReport.GetApplicationAttemptId
());
if (rmAppAttempt == null)
{
continue;
}
AppAttemptInfo attemptInfo = new AppAttemptInfo(this.rm, rmAppAttempt, rmApp.GetUser
(), WebAppUtils.GetHttpSchemePrefix(conf));
string blacklistedNodesCount = "N/A";
ICollection <string> nodes = RMAppAttemptBlock.GetBlacklistedNodes(rm, rmAppAttempt
.GetAppAttemptId());
if (nodes != null)
{
blacklistedNodesCount = nodes.Count.ToString();
}
string nodeLink = attemptInfo.GetNodeHttpAddress();
if (nodeLink != null)
{
nodeLink = WebAppUtils.GetHttpSchemePrefix(conf) + nodeLink;
}
string logsLink = attemptInfo.GetLogsLink();
attemptsTableData.Append("[\"<a href='").Append(Url("appattempt", rmAppAttempt.GetAppAttemptId
().ToString())).Append("'>").Append(rmAppAttempt.GetAppAttemptId().ToString()).Append
("</a>\",\"").Append(attemptInfo.GetStartTime()).Append("\",\"<a ").Append(nodeLink
== null ? "#" : "href='" + nodeLink).Append("'>").Append(nodeLink == null ? "N/A"
: StringEscapeUtils.EscapeJavaScript(StringEscapeUtils.EscapeHtml(nodeLink))).Append
("</a>\",\"<a ").Append(logsLink == null ? "#" : "href='" + logsLink).Append("'>"
).Append(logsLink == null ? "N/A" : "Logs").Append("</a>\",").Append("\"").Append
(blacklistedNodesCount).Append("\"],\n");
}
if (attemptsTableData[attemptsTableData.Length - 2] == ',')
{
attemptsTableData.Delete(attemptsTableData.Length - 2, attemptsTableData.Length -
1);
}
attemptsTableData.Append("]");
html.Script().$type("text/javascript").("var attemptsTableData=" + attemptsTableData
).();
tbody.().();
}
public virtual void TestAMLaunchAndCleanup()
{
Logger rootLogger = LogManager.GetRootLogger();
rootLogger.SetLevel(Level.Debug);
TestApplicationMasterLauncher.MyContainerManagerImpl containerManager = new TestApplicationMasterLauncher.MyContainerManagerImpl
();
MockRMWithCustomAMLauncher rm = new MockRMWithCustomAMLauncher(containerManager);
rm.Start();
MockNM nm1 = rm.RegisterNode("127.0.0.1:1234", 5120);
RMApp app = rm.SubmitApp(2000);
// kick the scheduling
nm1.NodeHeartbeat(true);
int waitCount = 0;
while (containerManager.launched == false && waitCount++ < 20)
{
Log.Info("Waiting for AM Launch to happen..");
Sharpen.Thread.Sleep(1000);
}
NUnit.Framework.Assert.IsTrue(containerManager.launched);
RMAppAttempt attempt = app.GetCurrentAppAttempt();
ApplicationAttemptId appAttemptId = attempt.GetAppAttemptId();
NUnit.Framework.Assert.AreEqual(appAttemptId.ToString(), containerManager.attemptIdAtContainerManager
);
NUnit.Framework.Assert.AreEqual(app.GetSubmitTime(), containerManager.submitTimeAtContainerManager
);
NUnit.Framework.Assert.AreEqual(app.GetRMAppAttempt(appAttemptId).GetMasterContainer
().GetId().ToString(), containerManager.containerIdAtContainerManager);
NUnit.Framework.Assert.AreEqual(nm1.GetNodeId().ToString(), containerManager.nmHostAtContainerManager
);
NUnit.Framework.Assert.AreEqual(YarnConfiguration.DefaultRmAmMaxAttempts, containerManager
.maxAppAttempts);
MockAM am = new MockAM(rm.GetRMContext(), rm.GetApplicationMasterService(), appAttemptId
);
am.RegisterAppAttempt();
am.UnregisterAppAttempt();
//complete the AM container to finish the app normally
nm1.NodeHeartbeat(attempt.GetAppAttemptId(), 1, ContainerState.Complete);
am.WaitForState(RMAppAttemptState.Finished);
waitCount = 0;
while (containerManager.cleanedup == false && waitCount++ < 20)
{
Log.Info("Waiting for AM Cleanup to happen..");
Sharpen.Thread.Sleep(1000);
}
NUnit.Framework.Assert.IsTrue(containerManager.cleanedup);
am.WaitForState(RMAppAttemptState.Finished);
rm.Stop();
}
/// <exception cref="System.Exception"/>
public virtual void WaitForState(RMAppAttemptState finalState)
{
RMApp app = context.GetRMApps()[attemptId.GetApplicationId()];
RMAppAttempt attempt = app.GetRMAppAttempt(attemptId);
int timeoutSecs = 0;
while (!finalState.Equals(attempt.GetAppAttemptState()) && timeoutSecs++ < 40)
{
System.Console.Out.WriteLine("AppAttempt : " + attemptId + " State is : " + attempt
.GetAppAttemptState() + " Waiting for state : " + finalState);
Sharpen.Thread.Sleep(1000);
}
System.Console.Out.WriteLine("AppAttempt State is : " + attempt.GetAppAttemptState
());
NUnit.Framework.Assert.AreEqual("AppAttempt state is not correct (timedout)", finalState
, attempt.GetAppAttemptState());
}
public virtual void SetUp()
{
CapacityScheduler spyCs = new CapacityScheduler();
cs = Org.Mockito.Mockito.Spy(spyCs);
scheduler = cs;
rmContext = TestUtils.GetMockRMContext();
spyRMContext = Org.Mockito.Mockito.Spy(rmContext);
ConcurrentMap <ApplicationId, RMApp> spyApps = Org.Mockito.Mockito.Spy(new ConcurrentHashMap
<ApplicationId, RMApp>());
RMApp rmApp = Org.Mockito.Mockito.Mock <RMApp>();
Org.Mockito.Mockito.When(rmApp.GetRMAppAttempt((ApplicationAttemptId)Matchers.Any
())).ThenReturn(null);
Org.Mockito.Mockito.DoReturn(rmApp).When(spyApps)[(ApplicationId)Matchers.Any()];
Org.Mockito.Mockito.When(spyRMContext.GetRMApps()).ThenReturn(spyApps);
Org.Mockito.Mockito.When(spyRMContext.GetScheduler()).ThenReturn(scheduler);
CapacitySchedulerConfiguration csConf = new CapacitySchedulerConfiguration();
ReservationSystemTestUtil.SetupQueueConfiguration(csConf);
cs.SetConf(csConf);
csContext = Org.Mockito.Mockito.Mock <CapacitySchedulerContext>();
Org.Mockito.Mockito.When(csContext.GetConfiguration()).ThenReturn(csConf);
Org.Mockito.Mockito.When(csContext.GetConf()).ThenReturn(csConf);
Org.Mockito.Mockito.When(csContext.GetMinimumResourceCapability()).ThenReturn(minAlloc
);
Org.Mockito.Mockito.When(csContext.GetMaximumResourceCapability()).ThenReturn(maxAlloc
);
Org.Mockito.Mockito.When(csContext.GetClusterResource()).ThenReturn(Resources.CreateResource
(100 * 16 * Gb, 100 * 32));
Org.Mockito.Mockito.When(scheduler.GetClusterResource()).ThenReturn(Resources.CreateResource
(125 * Gb, 125));
Org.Mockito.Mockito.When(csContext.GetResourceCalculator()).ThenReturn(new DefaultResourceCalculator
());
RMContainerTokenSecretManager containerTokenSecretManager = new RMContainerTokenSecretManager
(csConf);
containerTokenSecretManager.RollMasterKey();
Org.Mockito.Mockito.When(csContext.GetContainerTokenSecretManager()).ThenReturn(containerTokenSecretManager
);
cs.SetRMContext(spyRMContext);
cs.Init(csConf);
cs.Start();
SetupPlanFollower();
}
public virtual void SetUp()
{
conf = CreateConfiguration();
ReservationSystemTestUtil.SetupFSAllocationFile(AllocFile);
ReservationSystemTestUtil testUtil = new ReservationSystemTestUtil();
// Setup
rmContext = TestUtils.GetMockRMContext();
spyRMContext = Org.Mockito.Mockito.Spy(rmContext);
fs = ReservationSystemTestUtil.SetupFairScheduler(testUtil, spyRMContext, conf, 125
);
scheduler = fs;
ConcurrentMap <ApplicationId, RMApp> spyApps = Org.Mockito.Mockito.Spy(new ConcurrentHashMap
<ApplicationId, RMApp>());
RMApp rmApp = Org.Mockito.Mockito.Mock <RMApp>();
Org.Mockito.Mockito.When(rmApp.GetRMAppAttempt((ApplicationAttemptId)Matchers.Any
())).ThenReturn(null);
Org.Mockito.Mockito.DoReturn(rmApp).When(spyApps)[(ApplicationId)Matchers.Any()];
Org.Mockito.Mockito.When(spyRMContext.GetRMApps()).ThenReturn(spyApps);
ReservationSystemTestUtil.SetupFSAllocationFile(AllocFile);
SetupPlanFollower();
}
internal virtual void HandleNMContainerStatus(NMContainerStatus containerStatus,
NodeId nodeId)
{
ApplicationAttemptId appAttemptId = containerStatus.GetContainerId().GetApplicationAttemptId
();
RMApp rmApp = rmContext.GetRMApps()[appAttemptId.GetApplicationId()];
if (rmApp == null)
{
Log.Error("Received finished container : " + containerStatus.GetContainerId() + " for unknown application "
+ appAttemptId.GetApplicationId() + " Skipping.");
return;
}
if (rmApp.GetApplicationSubmissionContext().GetUnmanagedAM())
{
if (Log.IsDebugEnabled())
{
Log.Debug("Ignoring container completion status for unmanaged AM " + rmApp.GetApplicationId
());
}
return;
}
RMAppAttempt rmAppAttempt = rmApp.GetRMAppAttempt(appAttemptId);
Container masterContainer = rmAppAttempt.GetMasterContainer();
if (masterContainer.GetId().Equals(containerStatus.GetContainerId()) && containerStatus
.GetContainerState() == ContainerState.Complete)
{
ContainerStatus status = ContainerStatus.NewInstance(containerStatus.GetContainerId
(), containerStatus.GetContainerState(), containerStatus.GetDiagnostics(), containerStatus
.GetContainerExitStatus());
// sending master container finished event.
RMAppAttemptContainerFinishedEvent evt = new RMAppAttemptContainerFinishedEvent(appAttemptId
, status, nodeId);
rmContext.GetDispatcher().GetEventHandler().Handle(evt);
}
}
/// <exception cref="System.Exception"/>
public virtual void TestAMRestartWithExistingContainers()
{
YarnConfiguration conf = new YarnConfiguration();
conf.SetInt(YarnConfiguration.RmAmMaxAttempts, 2);
MockRM rm1 = new MockRM(conf);
rm1.Start();
RMApp app1 = rm1.SubmitApp(200, "name", "user", new Dictionary <ApplicationAccessType
, string>(), false, "default", -1, null, "MAPREDUCE", false, true);
MockNM nm1 = new MockNM("127.0.0.1:1234", 10240, rm1.GetResourceTrackerService());
nm1.RegisterNode();
MockNM nm2 = new MockNM("127.0.0.1:2351", 4089, rm1.GetResourceTrackerService());
nm2.RegisterNode();
MockAM am1 = MockRM.LaunchAndRegisterAM(app1, rm1, nm1);
int NumContainers = 3;
// allocate NUM_CONTAINERS containers
am1.Allocate("127.0.0.1", 1024, NumContainers, new AList <ContainerId>());
nm1.NodeHeartbeat(true);
// wait for containers to be allocated.
IList <Container> containers = am1.Allocate(new AList <ResourceRequest>(), new AList
<ContainerId>()).GetAllocatedContainers();
while (containers.Count != NumContainers)
{
nm1.NodeHeartbeat(true);
Sharpen.Collections.AddAll(containers, am1.Allocate(new AList <ResourceRequest>(),
new AList <ContainerId>()).GetAllocatedContainers());
Sharpen.Thread.Sleep(200);
}
// launch the 2nd container, for testing running container transferred.
nm1.NodeHeartbeat(am1.GetApplicationAttemptId(), 2, ContainerState.Running);
ContainerId containerId2 = ContainerId.NewContainerId(am1.GetApplicationAttemptId
(), 2);
rm1.WaitForState(nm1, containerId2, RMContainerState.Running);
// launch the 3rd container, for testing container allocated by previous
// attempt is completed by the next new attempt/
nm1.NodeHeartbeat(am1.GetApplicationAttemptId(), 3, ContainerState.Running);
ContainerId containerId3 = ContainerId.NewContainerId(am1.GetApplicationAttemptId
(), 3);
rm1.WaitForState(nm1, containerId3, RMContainerState.Running);
// 4th container still in AQUIRED state. for testing Acquired container is
// always killed.
ContainerId containerId4 = ContainerId.NewContainerId(am1.GetApplicationAttemptId
(), 4);
rm1.WaitForState(nm1, containerId4, RMContainerState.Acquired);
// 5th container is in Allocated state. for testing allocated container is
// always killed.
am1.Allocate("127.0.0.1", 1024, 1, new AList <ContainerId>());
nm1.NodeHeartbeat(true);
ContainerId containerId5 = ContainerId.NewContainerId(am1.GetApplicationAttemptId
(), 5);
rm1.WaitForContainerAllocated(nm1, containerId5);
rm1.WaitForState(nm1, containerId5, RMContainerState.Allocated);
// 6th container is in Reserved state.
am1.Allocate("127.0.0.1", 6000, 1, new AList <ContainerId>());
ContainerId containerId6 = ContainerId.NewContainerId(am1.GetApplicationAttemptId
(), 6);
nm1.NodeHeartbeat(true);
SchedulerApplicationAttempt schedulerAttempt = ((AbstractYarnScheduler)rm1.GetResourceScheduler
()).GetCurrentAttemptForContainer(containerId6);
while (schedulerAttempt.GetReservedContainers().IsEmpty())
{
System.Console.Out.WriteLine("Waiting for container " + containerId6 + " to be reserved."
);
nm1.NodeHeartbeat(true);
Sharpen.Thread.Sleep(200);
}
// assert containerId6 is reserved.
NUnit.Framework.Assert.AreEqual(containerId6, schedulerAttempt.GetReservedContainers
()[0].GetContainerId());
// fail the AM by sending CONTAINER_FINISHED event without registering.
nm1.NodeHeartbeat(am1.GetApplicationAttemptId(), 1, ContainerState.Complete);
am1.WaitForState(RMAppAttemptState.Failed);
// wait for some time. previous AM's running containers should still remain
// in scheduler even though am failed
Sharpen.Thread.Sleep(3000);
rm1.WaitForState(nm1, containerId2, RMContainerState.Running);
// acquired/allocated containers are cleaned up.
NUnit.Framework.Assert.IsNull(rm1.GetResourceScheduler().GetRMContainer(containerId4
));
NUnit.Framework.Assert.IsNull(rm1.GetResourceScheduler().GetRMContainer(containerId5
));
// wait for app to start a new attempt.
rm1.WaitForState(app1.GetApplicationId(), RMAppState.Accepted);
// assert this is a new AM.
ApplicationAttemptId newAttemptId = app1.GetCurrentAppAttempt().GetAppAttemptId();
NUnit.Framework.Assert.IsFalse(newAttemptId.Equals(am1.GetApplicationAttemptId())
);
// launch the new AM
RMAppAttempt attempt2 = app1.GetCurrentAppAttempt();
nm1.NodeHeartbeat(true);
MockAM am2 = rm1.SendAMLaunched(attempt2.GetAppAttemptId());
RegisterApplicationMasterResponse registerResponse = am2.RegisterAppAttempt();
// Assert two containers are running: container2 and container3;
NUnit.Framework.Assert.AreEqual(2, registerResponse.GetContainersFromPreviousAttempts
().Count);
bool containerId2Exists = false;
bool containerId3Exists = false;
foreach (Container container in registerResponse.GetContainersFromPreviousAttempts
())
{
if (container.GetId().Equals(containerId2))
{
containerId2Exists = true;
}
if (container.GetId().Equals(containerId3))
{
containerId3Exists = true;
}
}
NUnit.Framework.Assert.IsTrue(containerId2Exists && containerId3Exists);
rm1.WaitForState(app1.GetApplicationId(), RMAppState.Running);
// complete container by sending the container complete event which has earlier
// attempt's attemptId
nm1.NodeHeartbeat(am1.GetApplicationAttemptId(), 3, ContainerState.Complete);
// Even though the completed container containerId3 event was sent to the
// earlier failed attempt, new RMAppAttempt can also capture this container
// info.
// completed containerId4 is also transferred to the new attempt.
RMAppAttempt newAttempt = app1.GetRMAppAttempt(am2.GetApplicationAttemptId());
// 4 containers finished, acquired/allocated/reserved/completed.
WaitForContainersToFinish(4, newAttempt);
bool container3Exists = false;
bool container4Exists = false;
bool container5Exists = false;
bool container6Exists = false;
foreach (ContainerStatus status in newAttempt.GetJustFinishedContainers())
{
if (status.GetContainerId().Equals(containerId3))
{
// containerId3 is the container ran by previous attempt but finished by the
// new attempt.
container3Exists = true;
}
if (status.GetContainerId().Equals(containerId4))
{
// containerId4 is the Acquired Container killed by the previous attempt,
// it's now inside new attempt's finished container list.
container4Exists = true;
}
if (status.GetContainerId().Equals(containerId5))
{
// containerId5 is the Allocated container killed by previous failed attempt.
container5Exists = true;
}
if (status.GetContainerId().Equals(containerId6))
{
// containerId6 is the reserved container killed by previous failed attempt.
container6Exists = true;
}
}
NUnit.Framework.Assert.IsTrue(container3Exists && container4Exists && container5Exists &&
container6Exists);
// New SchedulerApplicationAttempt also has the containers info.
rm1.WaitForState(nm1, containerId2, RMContainerState.Running);
// record the scheduler attempt for testing.
SchedulerApplicationAttempt schedulerNewAttempt = ((AbstractYarnScheduler)rm1.GetResourceScheduler
()).GetCurrentAttemptForContainer(containerId2);
// finish this application
MockRM.FinishAMAndVerifyAppState(app1, rm1, nm1, am2);
// the 2nd attempt released the 1st attempt's running container, when the
// 2nd attempt finishes.
NUnit.Framework.Assert.IsFalse(schedulerNewAttempt.GetLiveContainers().Contains(containerId2
));
// all 4 normal containers finished.
System.Console.Out.WriteLine("New attempt's just finished containers: " + newAttempt
.GetJustFinishedContainers());
WaitForContainersToFinish(5, newAttempt);
rm1.Stop();
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual AllocateResponse Allocate(AllocateRequest request)
{
AMRMTokenIdentifier amrmTokenIdentifier = AuthorizeRequest();
ApplicationAttemptId appAttemptId = amrmTokenIdentifier.GetApplicationAttemptId();
ApplicationId applicationId = appAttemptId.GetApplicationId();
this.amLivelinessMonitor.ReceivedPing(appAttemptId);
/* check if its in cache */
ApplicationMasterService.AllocateResponseLock Lock = responseMap[appAttemptId];
if (Lock == null)
{
string message = "Application attempt " + appAttemptId + " doesn't exist in ApplicationMasterService cache.";
Log.Error(message);
throw new ApplicationAttemptNotFoundException(message);
}
lock (Lock)
{
AllocateResponse lastResponse = Lock.GetAllocateResponse();
if (!HasApplicationMasterRegistered(appAttemptId))
{
string message = "AM is not registered for known application attempt: " + appAttemptId
+ " or RM had restarted after AM registered . AM should re-register.";
Log.Info(message);
RMAuditLogger.LogFailure(this.rmContext.GetRMApps()[appAttemptId.GetApplicationId
()].GetUser(), RMAuditLogger.AuditConstants.AmAllocate, string.Empty, "ApplicationMasterService"
, message, applicationId, appAttemptId);
throw new ApplicationMasterNotRegisteredException(message);
}
if ((request.GetResponseId() + 1) == lastResponse.GetResponseId())
{
/* old heartbeat */
return(lastResponse);
}
else
{
if (request.GetResponseId() + 1 < lastResponse.GetResponseId())
{
string message = "Invalid responseId in AllocateRequest from application attempt: "
+ appAttemptId + ", expect responseId to be " + (lastResponse.GetResponseId() +
1);
throw new InvalidApplicationMasterRequestException(message);
}
}
//filter illegal progress values
float filteredProgress = request.GetProgress();
if (float.IsNaN(filteredProgress) || filteredProgress == float.NegativeInfinity ||
filteredProgress < 0)
{
request.SetProgress(0);
}
else
{
if (filteredProgress > 1 || filteredProgress == float.PositiveInfinity)
{
request.SetProgress(1);
}
}
// Send the status update to the appAttempt.
this.rmContext.GetDispatcher().GetEventHandler().Handle(new RMAppAttemptStatusupdateEvent
(appAttemptId, request.GetProgress()));
IList <ResourceRequest> ask = request.GetAskList();
IList <ContainerId> release = request.GetReleaseList();
ResourceBlacklistRequest blacklistRequest = request.GetResourceBlacklistRequest();
IList <string> blacklistAdditions = (blacklistRequest != null) ? blacklistRequest.
GetBlacklistAdditions() : Sharpen.Collections.EmptyList;
IList <string> blacklistRemovals = (blacklistRequest != null) ? blacklistRequest.GetBlacklistRemovals
() : Sharpen.Collections.EmptyList;
RMApp app = this.rmContext.GetRMApps()[applicationId];
// set label expression for Resource Requests if resourceName=ANY
ApplicationSubmissionContext asc = app.GetApplicationSubmissionContext();
foreach (ResourceRequest req in ask)
{
if (null == req.GetNodeLabelExpression() && ResourceRequest.Any.Equals(req.GetResourceName
()))
{
req.SetNodeLabelExpression(asc.GetNodeLabelExpression());
}
}
// sanity check
try
{
RMServerUtils.NormalizeAndValidateRequests(ask, rScheduler.GetMaximumResourceCapability
(), app.GetQueue(), rScheduler, rmContext);
}
catch (InvalidResourceRequestException e)
{
Log.Warn("Invalid resource ask by application " + appAttemptId, e);
throw;
}
try
{
RMServerUtils.ValidateBlacklistRequest(blacklistRequest);
}
catch (InvalidResourceBlacklistRequestException e)
{
Log.Warn("Invalid blacklist request by application " + appAttemptId, e);
throw;
}
// In the case of work-preserving AM restart, it's possible for the
// AM to release containers from the earlier attempt.
if (!app.GetApplicationSubmissionContext().GetKeepContainersAcrossApplicationAttempts
())
{
try
{
RMServerUtils.ValidateContainerReleaseRequest(release, appAttemptId);
}
catch (InvalidContainerReleaseException e)
{
Log.Warn("Invalid container release by application " + appAttemptId, e);
throw;
}
}
// Send new requests to appAttempt.
Allocation allocation = this.rScheduler.Allocate(appAttemptId, ask, release, blacklistAdditions
, blacklistRemovals);
if (!blacklistAdditions.IsEmpty() || !blacklistRemovals.IsEmpty())
{
Log.Info("blacklist are updated in Scheduler." + "blacklistAdditions: " + blacklistAdditions
+ ", " + "blacklistRemovals: " + blacklistRemovals);
}
RMAppAttempt appAttempt = app.GetRMAppAttempt(appAttemptId);
AllocateResponse allocateResponse = recordFactory.NewRecordInstance <AllocateResponse
>();
if (!allocation.GetContainers().IsEmpty())
{
allocateResponse.SetNMTokens(allocation.GetNMTokens());
}
// update the response with the deltas of node status changes
IList <RMNode> updatedNodes = new AList <RMNode>();
if (app.PullRMNodeUpdates(updatedNodes) > 0)
{
IList <NodeReport> updatedNodeReports = new AList <NodeReport>();
foreach (RMNode rmNode in updatedNodes)
{
SchedulerNodeReport schedulerNodeReport = rScheduler.GetNodeReport(rmNode.GetNodeID
());
Resource used = BuilderUtils.NewResource(0, 0);
int numContainers = 0;
if (schedulerNodeReport != null)
{
used = schedulerNodeReport.GetUsedResource();
numContainers = schedulerNodeReport.GetNumContainers();
}
NodeId nodeId = rmNode.GetNodeID();
NodeReport report = BuilderUtils.NewNodeReport(nodeId, rmNode.GetState(), rmNode.
GetHttpAddress(), rmNode.GetRackName(), used, rmNode.GetTotalCapability(), numContainers
, rmNode.GetHealthReport(), rmNode.GetLastHealthReportTime(), rmNode.GetNodeLabels
());
updatedNodeReports.AddItem(report);
}
allocateResponse.SetUpdatedNodes(updatedNodeReports);
}
allocateResponse.SetAllocatedContainers(allocation.GetContainers());
allocateResponse.SetCompletedContainersStatuses(appAttempt.PullJustFinishedContainers
());
allocateResponse.SetResponseId(lastResponse.GetResponseId() + 1);
allocateResponse.SetAvailableResources(allocation.GetResourceLimit());
allocateResponse.SetNumClusterNodes(this.rScheduler.GetNumClusterNodes());
// add preemption to the allocateResponse message (if any)
allocateResponse.SetPreemptionMessage(GeneratePreemptionMessage(allocation));
// update AMRMToken if the token is rolled-up
MasterKeyData nextMasterKey = this.rmContext.GetAMRMTokenSecretManager().GetNextMasterKeyData
();
if (nextMasterKey != null && nextMasterKey.GetMasterKey().GetKeyId() != amrmTokenIdentifier
.GetKeyId())
{
RMAppAttemptImpl appAttemptImpl = (RMAppAttemptImpl)appAttempt;
Org.Apache.Hadoop.Security.Token.Token <AMRMTokenIdentifier> amrmToken = appAttempt
.GetAMRMToken();
if (nextMasterKey.GetMasterKey().GetKeyId() != appAttemptImpl.GetAMRMTokenKeyId())
{
Log.Info("The AMRMToken has been rolled-over. Send new AMRMToken back" + " to application: "
+ applicationId);
amrmToken = rmContext.GetAMRMTokenSecretManager().CreateAndGetAMRMToken(appAttemptId
);
appAttemptImpl.SetAMRMToken(amrmToken);
}
allocateResponse.SetAMRMToken(Org.Apache.Hadoop.Yarn.Api.Records.Token.NewInstance
(amrmToken.GetIdentifier(), amrmToken.GetKind().ToString(), amrmToken.GetPassword
(), amrmToken.GetService().ToString()));
}
/*
* As we are updating the response inside the lock object so we don't
* need to worry about unregister call occurring in between (which
* removes the lock object).
*/
Lock.SetAllocateResponse(allocateResponse);
return(allocateResponse);
}
}
/// <exception cref="Org.Apache.Hadoop.Yarn.Exceptions.YarnException"/>
/// <exception cref="System.IO.IOException"/>
public virtual RegisterApplicationMasterResponse RegisterApplicationMaster(RegisterApplicationMasterRequest
request)
{
AMRMTokenIdentifier amrmTokenIdentifier = AuthorizeRequest();
ApplicationAttemptId applicationAttemptId = amrmTokenIdentifier.GetApplicationAttemptId
();
ApplicationId appID = applicationAttemptId.GetApplicationId();
ApplicationMasterService.AllocateResponseLock Lock = responseMap[applicationAttemptId
];
if (Lock == null)
{
RMAuditLogger.LogFailure(this.rmContext.GetRMApps()[appID].GetUser(), RMAuditLogger.AuditConstants
.RegisterAm, "Application doesn't exist in cache " + applicationAttemptId, "ApplicationMasterService"
, "Error in registering application master", appID, applicationAttemptId);
ThrowApplicationDoesNotExistInCacheException(applicationAttemptId);
}
// Allow only one thread in AM to do registerApp at a time.
lock (Lock)
{
AllocateResponse lastResponse = Lock.GetAllocateResponse();
if (HasApplicationMasterRegistered(applicationAttemptId))
{
string message = "Application Master is already registered : " + appID;
Log.Warn(message);
RMAuditLogger.LogFailure(this.rmContext.GetRMApps()[appID].GetUser(), RMAuditLogger.AuditConstants
.RegisterAm, string.Empty, "ApplicationMasterService", message, appID, applicationAttemptId
);
throw new InvalidApplicationMasterRequestException(message);
}
this.amLivelinessMonitor.ReceivedPing(applicationAttemptId);
RMApp app = this.rmContext.GetRMApps()[appID];
// Setting the response id to 0 to identify if the
// application master is register for the respective attemptid
lastResponse.SetResponseId(0);
Lock.SetAllocateResponse(lastResponse);
Log.Info("AM registration " + applicationAttemptId);
this.rmContext.GetDispatcher().GetEventHandler().Handle(new RMAppAttemptRegistrationEvent
(applicationAttemptId, request.GetHost(), request.GetRpcPort(), request.GetTrackingUrl
()));
RMAuditLogger.LogSuccess(app.GetUser(), RMAuditLogger.AuditConstants.RegisterAm,
"ApplicationMasterService", appID, applicationAttemptId);
// Pick up min/max resource from scheduler...
RegisterApplicationMasterResponse response = recordFactory.NewRecordInstance <RegisterApplicationMasterResponse
>();
response.SetMaximumResourceCapability(rScheduler.GetMaximumResourceCapability(app
.GetQueue()));
response.SetApplicationACLs(app.GetRMAppAttempt(applicationAttemptId).GetSubmissionContext
().GetAMContainerSpec().GetApplicationACLs());
response.SetQueue(app.GetQueue());
if (UserGroupInformation.IsSecurityEnabled())
{
Log.Info("Setting client token master key");
response.SetClientToAMTokenMasterKey(ByteBuffer.Wrap(rmContext.GetClientToAMTokenSecretManager
().GetMasterKey(applicationAttemptId).GetEncoded()));
}
// For work-preserving AM restart, retrieve previous attempts' containers
// and corresponding NM tokens.
if (app.GetApplicationSubmissionContext().GetKeepContainersAcrossApplicationAttempts
())
{
IList <Container> transferredContainers = ((AbstractYarnScheduler)rScheduler).GetTransferredContainers
(applicationAttemptId);
if (!transferredContainers.IsEmpty())
{
response.SetContainersFromPreviousAttempts(transferredContainers);
IList <NMToken> nmTokens = new AList <NMToken>();
foreach (Container container in transferredContainers)
{
try
{
NMToken token = rmContext.GetNMTokenSecretManager().CreateAndGetNMToken(app.GetUser
(), applicationAttemptId, container);
if (null != token)
{
nmTokens.AddItem(token);
}
}
catch (ArgumentException e)
{
// if it's a DNS issue, throw UnknowHostException directly and
// that
// will be automatically retried by RMProxy in RPC layer.
if (e.InnerException is UnknownHostException)
{
throw (UnknownHostException)e.InnerException;
}
}
}
response.SetNMTokensFromPreviousAttempts(nmTokens);
Log.Info("Application " + appID + " retrieved " + transferredContainers.Count + " containers from previous"
+ " attempts and " + nmTokens.Count + " NM tokens.");
}
}
response.SetSchedulerResourceTypes(rScheduler.GetSchedulingResourceTypes());
return(response);
}
}
public virtual void TestReleaseWhileRunning()
{
DrainDispatcher drainDispatcher = new DrainDispatcher();
EventHandler <RMAppAttemptEvent> appAttemptEventHandler = Org.Mockito.Mockito.Mock
<EventHandler>();
EventHandler generic = Org.Mockito.Mockito.Mock <EventHandler>();
drainDispatcher.Register(typeof(RMAppAttemptEventType), appAttemptEventHandler);
drainDispatcher.Register(typeof(RMNodeEventType), generic);
drainDispatcher.Init(new YarnConfiguration());
drainDispatcher.Start();
NodeId nodeId = BuilderUtils.NewNodeId("host", 3425);
ApplicationId appId = BuilderUtils.NewApplicationId(1, 1);
ApplicationAttemptId appAttemptId = BuilderUtils.NewApplicationAttemptId(appId, 1
);
ContainerId containerId = BuilderUtils.NewContainerId(appAttemptId, 1);
ContainerAllocationExpirer expirer = Org.Mockito.Mockito.Mock <ContainerAllocationExpirer
>();
Resource resource = BuilderUtils.NewResource(512, 1);
Priority priority = BuilderUtils.NewPriority(5);
Container container = BuilderUtils.NewContainer(containerId, nodeId, "host:3465",
resource, priority, null);
ConcurrentMap <ApplicationId, RMApp> rmApps = Org.Mockito.Mockito.Spy(new ConcurrentHashMap
<ApplicationId, RMApp>());
RMApp rmApp = Org.Mockito.Mockito.Mock <RMApp>();
Org.Mockito.Mockito.When(rmApp.GetRMAppAttempt((ApplicationAttemptId)Matchers.Any
())).ThenReturn(null);
Org.Mockito.Mockito.DoReturn(rmApp).When(rmApps)[(ApplicationId)Matchers.Any()];
RMApplicationHistoryWriter writer = Org.Mockito.Mockito.Mock <RMApplicationHistoryWriter
>();
SystemMetricsPublisher publisher = Org.Mockito.Mockito.Mock <SystemMetricsPublisher
>();
RMContext rmContext = Org.Mockito.Mockito.Mock <RMContext>();
Org.Mockito.Mockito.When(rmContext.GetDispatcher()).ThenReturn(drainDispatcher);
Org.Mockito.Mockito.When(rmContext.GetContainerAllocationExpirer()).ThenReturn(expirer
);
Org.Mockito.Mockito.When(rmContext.GetRMApplicationHistoryWriter()).ThenReturn(writer
);
Org.Mockito.Mockito.When(rmContext.GetRMApps()).ThenReturn(rmApps);
Org.Mockito.Mockito.When(rmContext.GetSystemMetricsPublisher()).ThenReturn(publisher
);
Org.Mockito.Mockito.When(rmContext.GetYarnConfiguration()).ThenReturn(new YarnConfiguration
());
RMContainer rmContainer = new RMContainerImpl(container, appAttemptId, nodeId, "user"
, rmContext);
NUnit.Framework.Assert.AreEqual(RMContainerState.New, rmContainer.GetState());
NUnit.Framework.Assert.AreEqual(resource, rmContainer.GetAllocatedResource());
NUnit.Framework.Assert.AreEqual(nodeId, rmContainer.GetAllocatedNode());
NUnit.Framework.Assert.AreEqual(priority, rmContainer.GetAllocatedPriority());
Org.Mockito.Mockito.Verify(writer).ContainerStarted(Matchers.Any <RMContainer>());
Org.Mockito.Mockito.Verify(publisher).ContainerCreated(Matchers.Any <RMContainer>(
), Matchers.AnyLong());
rmContainer.Handle(new RMContainerEvent(containerId, RMContainerEventType.Start));
drainDispatcher.Await();
NUnit.Framework.Assert.AreEqual(RMContainerState.Allocated, rmContainer.GetState(
));
rmContainer.Handle(new RMContainerEvent(containerId, RMContainerEventType.Acquired
));
drainDispatcher.Await();
NUnit.Framework.Assert.AreEqual(RMContainerState.Acquired, rmContainer.GetState()
);
rmContainer.Handle(new RMContainerEvent(containerId, RMContainerEventType.Launched
));
drainDispatcher.Await();
NUnit.Framework.Assert.AreEqual(RMContainerState.Running, rmContainer.GetState());
NUnit.Framework.Assert.AreEqual("http://host:3465/node/containerlogs/container_1_0001_01_000001/user"
, rmContainer.GetLogURL());
// In RUNNING state. Verify RELEASED and associated actions.
Org.Mockito.Mockito.Reset(appAttemptEventHandler);
ContainerStatus containerStatus = SchedulerUtils.CreateAbnormalContainerStatus(containerId
, SchedulerUtils.ReleasedContainer);
rmContainer.Handle(new RMContainerFinishedEvent(containerId, containerStatus, RMContainerEventType
.Released));
drainDispatcher.Await();
NUnit.Framework.Assert.AreEqual(RMContainerState.Released, rmContainer.GetState()
);
NUnit.Framework.Assert.AreEqual(SchedulerUtils.ReleasedContainer, rmContainer.GetDiagnosticsInfo
());
NUnit.Framework.Assert.AreEqual(ContainerExitStatus.Aborted, rmContainer.GetContainerExitStatus
());
NUnit.Framework.Assert.AreEqual(ContainerState.Complete, rmContainer.GetContainerState
());
Org.Mockito.Mockito.Verify(writer).ContainerFinished(Matchers.Any <RMContainer>());
Org.Mockito.Mockito.Verify(publisher).ContainerFinished(Matchers.Any <RMContainer>
(), Matchers.AnyLong());
ArgumentCaptor <RMAppAttemptContainerFinishedEvent> captor = ArgumentCaptor.ForClass
<RMAppAttemptContainerFinishedEvent>();
Org.Mockito.Mockito.Verify(appAttemptEventHandler).Handle(captor.Capture());
RMAppAttemptContainerFinishedEvent cfEvent = captor.GetValue();
NUnit.Framework.Assert.AreEqual(appAttemptId, cfEvent.GetApplicationAttemptId());
NUnit.Framework.Assert.AreEqual(containerStatus, cfEvent.GetContainerStatus());
NUnit.Framework.Assert.AreEqual(RMAppAttemptEventType.ContainerFinished, cfEvent.
GetType());
// In RELEASED state. A FINIHSED event may come in.
rmContainer.Handle(new RMContainerFinishedEvent(containerId, SchedulerUtils.CreateAbnormalContainerStatus
(containerId, "FinishedContainer"), RMContainerEventType.Finished));
NUnit.Framework.Assert.AreEqual(RMContainerState.Released, rmContainer.GetState()
);
}
