public virtual void TestNodeUpdateBeforeAppAttemptInit()
{
FifoScheduler scheduler = new FifoScheduler();
MockRM rm = new MockRM(conf);
scheduler.SetRMContext(rm.GetRMContext());
scheduler.Init(conf);
scheduler.Start();
scheduler.Reinitialize(conf, rm.GetRMContext());
RMNode node = MockNodes.NewNodeInfo(1, Resources.CreateResource(1024, 4), 1, "127.0.0.1"
);
scheduler.Handle(new NodeAddedSchedulerEvent(node));
ApplicationId appId = ApplicationId.NewInstance(0, 1);
scheduler.AddApplication(appId, "queue1", "user1", false);
NodeUpdateSchedulerEvent updateEvent = new NodeUpdateSchedulerEvent(node);
try
{
scheduler.Handle(updateEvent);
}
catch (ArgumentNullException)
{
NUnit.Framework.Assert.Fail();
}
ApplicationAttemptId attId = ApplicationAttemptId.NewInstance(appId, 1);
scheduler.AddApplicationAttempt(attId, false, false);
rm.Stop();
}
/// <exception cref="System.IO.IOException"/>
protected internal virtual void StartRMs(MockRM rm1, Configuration confForRM1, MockRM
rm2, Configuration confForRM2)
{
rm1.Init(confForRM1);
rm1.Start();
NUnit.Framework.Assert.IsTrue(rm1.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Standby);
rm2.Init(confForRM2);
rm2.Start();
NUnit.Framework.Assert.IsTrue(rm2.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Standby);
rm1.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.IsTrue(rm1.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Active);
}
public virtual void TestRMDispatcherForHA()
{
string errorMessageForEventHandler = "Expect to get the same number of handlers";
string errorMessageForService = "Expect to get the same number of services";
configuration.SetBoolean(YarnConfiguration.AutoFailoverEnabled, false);
Configuration conf = new YarnConfiguration(configuration);
rm = new _MockRM_313(conf);
rm.Init(conf);
int expectedEventHandlerCount = ((TestRMHA.MyCountingDispatcher)rm.GetRMContext()
.GetDispatcher()).GetEventHandlerCount();
int expectedServiceCount = rm.GetServices().Count;
NUnit.Framework.Assert.IsTrue(expectedEventHandlerCount != 0);
HAServiceProtocol.StateChangeRequestInfo requestInfo = new HAServiceProtocol.StateChangeRequestInfo
(HAServiceProtocol.RequestSource.RequestByUser);
NUnit.Framework.Assert.AreEqual(StateErr, HAServiceProtocol.HAServiceState.Initializing
, rm.adminService.GetServiceStatus().GetState());
NUnit.Framework.Assert.IsFalse("RM is ready to become active before being started"
, rm.adminService.GetServiceStatus().IsReadyToBecomeActive());
rm.Start();
//call transitions to standby and active a couple of times
rm.adminService.TransitionToStandby(requestInfo);
rm.adminService.TransitionToActive(requestInfo);
rm.adminService.TransitionToStandby(requestInfo);
rm.adminService.TransitionToActive(requestInfo);
rm.adminService.TransitionToStandby(requestInfo);
TestRMHA.MyCountingDispatcher dispatcher = (TestRMHA.MyCountingDispatcher)rm.GetRMContext
().GetDispatcher();
NUnit.Framework.Assert.IsTrue(!dispatcher.IsStopped());
rm.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.AreEqual(errorMessageForEventHandler, expectedEventHandlerCount
, ((TestRMHA.MyCountingDispatcher)rm.GetRMContext().GetDispatcher()).GetEventHandlerCount
());
NUnit.Framework.Assert.AreEqual(errorMessageForService, expectedServiceCount, rm.
GetServices().Count);
// Keep the dispatcher reference before transitioning to standby
dispatcher = (TestRMHA.MyCountingDispatcher)rm.GetRMContext().GetDispatcher();
rm.adminService.TransitionToStandby(requestInfo);
NUnit.Framework.Assert.AreEqual(errorMessageForEventHandler, expectedEventHandlerCount
, ((TestRMHA.MyCountingDispatcher)rm.GetRMContext().GetDispatcher()).GetEventHandlerCount
());
NUnit.Framework.Assert.AreEqual(errorMessageForService, expectedServiceCount, rm.
GetServices().Count);
NUnit.Framework.Assert.IsTrue(dispatcher.IsStopped());
rm.Stop();
}
public virtual void TestHandleContainerStatusInvalidCompletions()
{
rm = new MockRM(new YarnConfiguration());
rm.Start();
EventHandler handler = Org.Mockito.Mockito.Spy(rm.GetRMContext().GetDispatcher().
GetEventHandler());
// Case 1: Unmanaged AM
RMApp app = rm.SubmitApp(1024, true);
// Case 1.1: AppAttemptId is null
NMContainerStatus report = NMContainerStatus.NewInstance(ContainerId.NewContainerId
(ApplicationAttemptId.NewInstance(app.GetApplicationId(), 2), 1), ContainerState
.Complete, Resource.NewInstance(1024, 1), "Dummy Completed", 0, Priority.NewInstance
(10), 1234);
rm.GetResourceTrackerService().HandleNMContainerStatus(report, null);
Org.Mockito.Mockito.Verify(handler, Org.Mockito.Mockito.Never()).Handle((Org.Apache.Hadoop.Yarn.Event.Event
)Matchers.Any());
// Case 1.2: Master container is null
RMAppAttemptImpl currentAttempt = (RMAppAttemptImpl)app.GetCurrentAppAttempt();
currentAttempt.SetMasterContainer(null);
report = NMContainerStatus.NewInstance(ContainerId.NewContainerId(currentAttempt.
GetAppAttemptId(), 0), ContainerState.Complete, Resource.NewInstance(1024, 1), "Dummy Completed"
, 0, Priority.NewInstance(10), 1234);
rm.GetResourceTrackerService().HandleNMContainerStatus(report, null);
Org.Mockito.Mockito.Verify(handler, Org.Mockito.Mockito.Never()).Handle((Org.Apache.Hadoop.Yarn.Event.Event
)Matchers.Any());
// Case 2: Managed AM
app = rm.SubmitApp(1024);
// Case 2.1: AppAttemptId is null
report = NMContainerStatus.NewInstance(ContainerId.NewContainerId(ApplicationAttemptId
.NewInstance(app.GetApplicationId(), 2), 1), ContainerState.Complete, Resource.NewInstance
(1024, 1), "Dummy Completed", 0, Priority.NewInstance(10), 1234);
try
{
rm.GetResourceTrackerService().HandleNMContainerStatus(report, null);
}
catch (Exception)
{
}
// expected - ignore
Org.Mockito.Mockito.Verify(handler, Org.Mockito.Mockito.Never()).Handle((Org.Apache.Hadoop.Yarn.Event.Event
)Matchers.Any());
// Case 2.2: Master container is null
currentAttempt = (RMAppAttemptImpl)app.GetCurrentAppAttempt();
currentAttempt.SetMasterContainer(null);
report = NMContainerStatus.NewInstance(ContainerId.NewContainerId(currentAttempt.
GetAppAttemptId(), 0), ContainerState.Complete, Resource.NewInstance(1024, 1), "Dummy Completed"
, 0, Priority.NewInstance(10), 1234);
try
{
rm.GetResourceTrackerService().HandleNMContainerStatus(report, null);
}
catch (Exception)
{
}
// expected - ignore
Org.Mockito.Mockito.Verify(handler, Org.Mockito.Mockito.Never()).Handle((Org.Apache.Hadoop.Yarn.Event.Event
)Matchers.Any());
}
/// <exception cref="System.Exception"/>
public virtual void TestTransitionedToStandbyShouldNotHang()
{
configuration.SetBoolean(YarnConfiguration.AutoFailoverEnabled, false);
Configuration conf = new YarnConfiguration(configuration);
MemoryRMStateStore memStore = new _MemoryRMStateStore_464();
memStore.Init(conf);
rm = new _MockRM_472(conf, memStore);
rm.Init(conf);
HAServiceProtocol.StateChangeRequestInfo requestInfo = new HAServiceProtocol.StateChangeRequestInfo
(HAServiceProtocol.RequestSource.RequestByUser);
NUnit.Framework.Assert.AreEqual(StateErr, HAServiceProtocol.HAServiceState.Initializing
, rm.adminService.GetServiceStatus().GetState());
NUnit.Framework.Assert.IsFalse("RM is ready to become active before being started"
, rm.adminService.GetServiceStatus().IsReadyToBecomeActive());
CheckMonitorHealth();
rm.Start();
CheckMonitorHealth();
CheckStandbyRMFunctionality();
// 2. Transition to Active.
rm.adminService.TransitionToActive(requestInfo);
// 3. Try Transition to standby
Sharpen.Thread t = new Sharpen.Thread(new _Runnable_498(this));
// TODO Auto-generated catch block
t.Start();
rm.GetRMContext().GetStateStore().UpdateApplicationState(null);
t.Join();
// wait for thread to finish
rm.adminService.TransitionToStandby(requestInfo);
CheckStandbyRMFunctionality();
rm.Stop();
}
public virtual void TestAdminRefreshQueuesWithFileSystemBasedConfigurationProvider
()
{
configuration.Set(YarnConfiguration.RmConfigurationProviderClass, "org.apache.hadoop.yarn.FileSystemBasedConfigurationProvider"
);
//upload default configurations
UploadDefaultConfiguration();
try
{
rm = new MockRM(configuration);
rm.Init(configuration);
rm.Start();
}
catch (Exception)
{
NUnit.Framework.Assert.Fail("Should not get any exceptions");
}
CapacityScheduler cs = (CapacityScheduler)rm.GetRMContext().GetScheduler();
int maxAppsBefore = cs.GetConfiguration().GetMaximumSystemApplications();
CapacitySchedulerConfiguration csConf = new CapacitySchedulerConfiguration();
csConf.Set("yarn.scheduler.capacity.maximum-applications", "5000");
UploadConfiguration(csConf, "capacity-scheduler.xml");
rm.adminService.RefreshQueues(RefreshQueuesRequest.NewInstance());
int maxAppsAfter = cs.GetConfiguration().GetMaximumSystemApplications();
NUnit.Framework.Assert.AreEqual(maxAppsAfter, 5000);
NUnit.Framework.Assert.IsTrue(maxAppsAfter != maxAppsBefore);
}
/// <exception cref="System.Exception"/>
private void CheckUnealthyNMCount(MockRM rm, MockNM nm1, bool health, int count)
{
int waitCount = 0;
while ((rm.GetRMContext().GetRMNodes()[nm1.GetNodeId()].GetState() != NodeState.Unhealthy
) == health && waitCount++ < 20)
{
lock (this)
{
Sharpen.Runtime.Wait(this, 100);
}
}
NUnit.Framework.Assert.IsFalse((rm.GetRMContext().GetRMNodes()[nm1.GetNodeId()].GetState
() != NodeState.Unhealthy) == health);
NUnit.Framework.Assert.AreEqual("Unhealthy metrics not incremented", count, ClusterMetrics
.GetMetrics().GetUnhealthyNMs());
}
public virtual void TestFailoverClearsRMContext()
{
configuration.SetBoolean(YarnConfiguration.AutoFailoverEnabled, false);
configuration.SetBoolean(YarnConfiguration.RecoveryEnabled, true);
Configuration conf = new YarnConfiguration(configuration);
MemoryRMStateStore memStore = new MemoryRMStateStore();
memStore.Init(conf);
// 1. start RM
rm = new MockRM(conf, memStore);
rm.Init(conf);
rm.Start();
HAServiceProtocol.StateChangeRequestInfo requestInfo = new HAServiceProtocol.StateChangeRequestInfo
(HAServiceProtocol.RequestSource.RequestByUser);
CheckMonitorHealth();
CheckStandbyRMFunctionality();
// 2. Transition to active
rm.adminService.TransitionToActive(requestInfo);
CheckMonitorHealth();
CheckActiveRMFunctionality();
VerifyClusterMetrics(1, 1, 1, 1, 2048, 1);
NUnit.Framework.Assert.AreEqual(1, rm.GetRMContext().GetRMNodes().Count);
NUnit.Framework.Assert.AreEqual(1, rm.GetRMContext().GetRMApps().Count);
// 3. Create new RM
rm = new _MockRM_550(conf, memStore);
rm.Init(conf);
rm.Start();
CheckMonitorHealth();
CheckStandbyRMFunctionality();
// 4. Try Transition to active, throw exception
try
{
rm.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.Fail("Transitioned to Active should throw exception.");
}
catch (Exception e)
{
NUnit.Framework.Assert.IsTrue("Error when transitioning to Active mode".Contains(
e.Message));
}
// 5. Clears the metrics
VerifyClusterMetrics(0, 0, 0, 0, 0, 0);
NUnit.Framework.Assert.AreEqual(0, rm.GetRMContext().GetRMNodes().Count);
NUnit.Framework.Assert.AreEqual(0, rm.GetRMContext().GetRMApps().Count);
}
/// <exception cref="System.IO.IOException"/>
protected internal virtual void ExplicitFailover()
{
rm1.adminService.TransitionToStandby(requestInfo);
rm2.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.IsTrue(rm1.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Standby);
NUnit.Framework.Assert.IsTrue(rm2.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Active);
}
// Test even if AM container is allocated with containerId not equal to 1, the
// following allocate requests from AM should be able to retrieve the
// corresponding NM Token.
/// <exception cref="System.Exception"/>
public virtual void TestNMTokenSentForNormalContainer()
{
conf.Set(YarnConfiguration.RmScheduler, typeof(CapacityScheduler).GetCanonicalName
());
MockRM rm = new MockRM(conf);
rm.Start();
MockNM nm1 = rm.RegisterNode("h1:1234", 5120);
RMApp app = rm.SubmitApp(2000);
RMAppAttempt attempt = app.GetCurrentAppAttempt();
// Call getNewContainerId to increase container Id so that the AM container
// Id doesn't equal to one.
CapacityScheduler cs = (CapacityScheduler)rm.GetResourceScheduler();
cs.GetApplicationAttempt(attempt.GetAppAttemptId()).GetNewContainerId();
// kick the scheduling
nm1.NodeHeartbeat(true);
MockAM am = MockRM.LaunchAM(app, rm, nm1);
// am container Id not equal to 1.
NUnit.Framework.Assert.IsTrue(attempt.GetMasterContainer().GetId().GetContainerId
() != 1);
// NMSecretManager doesn't record the node on which the am is allocated.
NUnit.Framework.Assert.IsFalse(rm.GetRMContext().GetNMTokenSecretManager().IsApplicationAttemptNMTokenPresent
(attempt.GetAppAttemptId(), nm1.GetNodeId()));
am.RegisterAppAttempt();
rm.WaitForState(app.GetApplicationId(), RMAppState.Running);
int NumContainers = 1;
IList <Container> containers = new AList <Container>();
// nmTokens keeps track of all the nmTokens issued in the allocate call.
IList <NMToken> expectedNMTokens = new AList <NMToken>();
// am1 allocate 1 container on nm1.
while (true)
{
AllocateResponse response = am.Allocate("127.0.0.1", 2000, NumContainers, new AList
<ContainerId>());
nm1.NodeHeartbeat(true);
Sharpen.Collections.AddAll(containers, response.GetAllocatedContainers());
Sharpen.Collections.AddAll(expectedNMTokens, response.GetNMTokens());
if (containers.Count == NumContainers)
{
break;
}
Sharpen.Thread.Sleep(200);
System.Console.Out.WriteLine("Waiting for container to be allocated.");
}
NodeId nodeId = expectedNMTokens[0].GetNodeId();
// NMToken is sent for the allocated container.
NUnit.Framework.Assert.AreEqual(nm1.GetNodeId(), nodeId);
}
/// <exception cref="System.Exception"/>
public virtual void TestMoveSuccessful()
{
MockRM rm1 = new MockRM(conf);
rm1.Start();
RMApp app = rm1.SubmitApp(1024);
ClientRMService clientRMService = rm1.GetClientRMService();
// FIFO scheduler does not support moves
clientRMService.MoveApplicationAcrossQueues(MoveApplicationAcrossQueuesRequest.NewInstance
(app.GetApplicationId(), "newqueue"));
RMApp rmApp = rm1.GetRMContext().GetRMApps()[app.GetApplicationId()];
NUnit.Framework.Assert.AreEqual("newqueue", rmApp.GetQueue());
rm1.Stop();
}
/// <exception cref="System.Exception"/>
private void KillApplication(MockRM rm, ApplicationId appId, ApplicationAttemptId
appAttemptId, RMAppState rmAppState)
{
KillApplicationResponse response = rm.KillApp(appId);
NUnit.Framework.Assert.IsTrue(response.GetIsKillCompleted() == IsFinalState(rmAppState
));
RMApp loadedApp0 = rm.GetRMContext().GetRMApps()[appId];
rm.WaitForState(appId, RMAppState.Killed);
if (appAttemptId != null)
{
rm.WaitForState(appAttemptId, RMAppAttemptState.Killed);
}
// no new attempt is created.
NUnit.Framework.Assert.AreEqual(1, loadedApp0.GetAppAttempts().Count);
}
public virtual void TestAdminRefreshQueuesWithLocalConfigurationProvider()
{
rm = new MockRM(configuration);
rm.Init(configuration);
rm.Start();
CapacityScheduler cs = (CapacityScheduler)rm.GetRMContext().GetScheduler();
int maxAppsBefore = cs.GetConfiguration().GetMaximumSystemApplications();
try
{
rm.adminService.RefreshQueues(RefreshQueuesRequest.NewInstance());
NUnit.Framework.Assert.AreEqual(maxAppsBefore, cs.GetConfiguration().GetMaximumSystemApplications
());
}
catch (Exception)
{
NUnit.Framework.Assert.Fail("Using localConfigurationProvider. Should not get any exception."
);
}
}
/// <exception cref="System.Exception"/>
public virtual void TestTransitionedToActiveRefreshFail()
{
configuration.SetBoolean(YarnConfiguration.AutoFailoverEnabled, false);
YarnConfiguration conf = new YarnConfiguration(configuration);
configuration = new CapacitySchedulerConfiguration(conf);
rm = new _MockRM_588(this, configuration);
rm.Init(configuration);
rm.Start();
HAServiceProtocol.StateChangeRequestInfo requestInfo = new HAServiceProtocol.StateChangeRequestInfo
(HAServiceProtocol.RequestSource.RequestByUser);
configuration.Set("yarn.scheduler.capacity.root.default.capacity", "100");
rm.adminService.TransitionToStandby(requestInfo);
NUnit.Framework.Assert.AreEqual(HAServiceProtocol.HAServiceState.Standby, rm.GetRMContext
().GetHAServiceState());
configuration.Set("yarn.scheduler.capacity.root.default.capacity", "200");
try
{
rm.adminService.TransitionToActive(requestInfo);
}
catch (Exception e)
{
NUnit.Framework.Assert.IsTrue("Error on refreshAll during transistion to Active".
Contains(e.Message));
}
TestRMHA.FailFastDispatcher dispatcher = ((TestRMHA.FailFastDispatcher)rm.rmContext
.GetDispatcher());
dispatcher.Await();
NUnit.Framework.Assert.AreEqual(1, dispatcher.GetEventCount());
// Making correct conf and check the state
configuration.Set("yarn.scheduler.capacity.root.default.capacity", "100");
rm.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.AreEqual(HAServiceProtocol.HAServiceState.Active, rm.GetRMContext
().GetHAServiceState());
rm.adminService.TransitionToStandby(requestInfo);
NUnit.Framework.Assert.AreEqual(HAServiceProtocol.HAServiceState.Standby, rm.GetRMContext
().GetHAServiceState());
}
/// <exception cref="System.Exception"/>
public virtual void TestNMToken()
{
MockRM rm = new MockRM(conf);
try
{
rm.Start();
MockNM nm1 = rm.RegisterNode("h1:1234", 10000);
NMTokenSecretManagerInRM nmTokenSecretManager = rm.GetRMContext().GetNMTokenSecretManager
();
// submitting new application
RMApp app = rm.SubmitApp(1000);
// start scheduling.
nm1.NodeHeartbeat(true);
// Starting application attempt and launching
// It should get registered with NMTokenSecretManager.
RMAppAttempt attempt = app.GetCurrentAppAttempt();
MockAM am = rm.SendAMLaunched(attempt.GetAppAttemptId());
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptRegistered
(attempt.GetAppAttemptId()));
// This will register application master.
am.RegisterAppAttempt();
AList <Container> containersReceivedForNM1 = new AList <Container>();
IList <ContainerId> releaseContainerList = new AList <ContainerId>();
Dictionary <string, Token> nmTokens = new Dictionary <string, Token>();
// initially requesting 2 containers.
AllocateResponse response = am.Allocate("h1", 1000, 2, releaseContainerList);
NUnit.Framework.Assert.AreEqual(0, response.GetAllocatedContainers().Count);
AllocateContainersAndValidateNMTokens(am, containersReceivedForNM1, 2, nmTokens,
nm1);
NUnit.Framework.Assert.AreEqual(1, nmTokens.Count);
// requesting 2 more containers.
response = am.Allocate("h1", 1000, 2, releaseContainerList);
NUnit.Framework.Assert.AreEqual(0, response.GetAllocatedContainers().Count);
AllocateContainersAndValidateNMTokens(am, containersReceivedForNM1, 4, nmTokens,
nm1);
NUnit.Framework.Assert.AreEqual(1, nmTokens.Count);
// We will be simulating NM restart so restarting newly added h2:1234
// NM 2 now registers.
MockNM nm2 = rm.RegisterNode("h2:1234", 10000);
nm2.NodeHeartbeat(true);
AList <Container> containersReceivedForNM2 = new AList <Container>();
response = am.Allocate("h2", 1000, 2, releaseContainerList);
NUnit.Framework.Assert.AreEqual(0, response.GetAllocatedContainers().Count);
AllocateContainersAndValidateNMTokens(am, containersReceivedForNM2, 2, nmTokens,
nm2);
NUnit.Framework.Assert.AreEqual(2, nmTokens.Count);
// Simulating NM-2 restart.
nm2 = rm.RegisterNode("h2:1234", 10000);
// Wait for reconnect to make it through the RM and create a new RMNode
IDictionary <NodeId, RMNode> nodes = rm.GetRMContext().GetRMNodes();
while (nodes[nm2.GetNodeId()].GetLastNodeHeartBeatResponse().GetResponseId() > 0)
{
Sharpen.Thread.Sleep(WaitSleepMs);
}
int interval = 40;
// Wait for nm Token to be cleared.
while (nmTokenSecretManager.IsApplicationAttemptNMTokenPresent(attempt.GetAppAttemptId
(), nm2.GetNodeId()) && interval-- > 0)
{
Log.Info("waiting for nmToken to be cleared for : " + nm2.GetNodeId());
Sharpen.Thread.Sleep(WaitSleepMs);
}
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptRegistered
(attempt.GetAppAttemptId()));
// removing NMToken for h2:1234
Sharpen.Collections.Remove(nmTokens, nm2.GetNodeId().ToString());
NUnit.Framework.Assert.AreEqual(1, nmTokens.Count);
// We should again receive the NMToken.
response = am.Allocate("h2", 1000, 2, releaseContainerList);
NUnit.Framework.Assert.AreEqual(0, response.GetAllocatedContainers().Count);
AllocateContainersAndValidateNMTokens(am, containersReceivedForNM2, 4, nmTokens,
nm2);
NUnit.Framework.Assert.AreEqual(2, nmTokens.Count);
// Now rolling over NMToken masterKey. it should resend the NMToken in
// next allocate call.
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptNMTokenPresent
(attempt.GetAppAttemptId(), nm1.GetNodeId()));
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptNMTokenPresent
(attempt.GetAppAttemptId(), nm2.GetNodeId()));
nmTokenSecretManager.RollMasterKey();
nmTokenSecretManager.ActivateNextMasterKey();
NUnit.Framework.Assert.IsFalse(nmTokenSecretManager.IsApplicationAttemptNMTokenPresent
(attempt.GetAppAttemptId(), nm1.GetNodeId()));
NUnit.Framework.Assert.IsFalse(nmTokenSecretManager.IsApplicationAttemptNMTokenPresent
(attempt.GetAppAttemptId(), nm2.GetNodeId()));
// It should not remove application attempt entry.
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptRegistered
(attempt.GetAppAttemptId()));
nmTokens.Clear();
NUnit.Framework.Assert.AreEqual(0, nmTokens.Count);
// We should again receive the NMToken.
response = am.Allocate("h2", 1000, 1, releaseContainerList);
NUnit.Framework.Assert.AreEqual(0, response.GetAllocatedContainers().Count);
AllocateContainersAndValidateNMTokens(am, containersReceivedForNM2, 5, nmTokens,
nm2);
NUnit.Framework.Assert.AreEqual(1, nmTokens.Count);
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptNMTokenPresent
(attempt.GetAppAttemptId(), nm2.GetNodeId()));
// After AM is finished making sure that nmtoken entry for app
NUnit.Framework.Assert.IsTrue(nmTokenSecretManager.IsApplicationAttemptRegistered
(attempt.GetAppAttemptId()));
am.UnregisterAppAttempt();
// marking all the containers as finished.
foreach (Container container in containersReceivedForNM1)
{
nm1.NodeHeartbeat(attempt.GetAppAttemptId(), container.GetId().GetContainerId(),
ContainerState.Complete);
}
foreach (Container container_1 in containersReceivedForNM2)
{
nm2.NodeHeartbeat(attempt.GetAppAttemptId(), container_1.GetId().GetContainerId()
, ContainerState.Complete);
}
nm1.NodeHeartbeat(am.GetApplicationAttemptId(), 1, ContainerState.Complete);
am.WaitForState(RMAppAttemptState.Finished);
NUnit.Framework.Assert.IsFalse(nmTokenSecretManager.IsApplicationAttemptRegistered
(attempt.GetAppAttemptId()));
}
finally
{
rm.Stop();
}
}
// The test verifies processing of NMContainerStatuses which are sent during
// NM registration.
// 1. Start the cluster-RM,NM,Submit app with 1024MB,Launch & register AM
// 2. AM sends ResourceRequest for 1 container with memory 2048MB.
// 3. Verify for number of container allocated by RM
// 4. Verify Memory Usage by cluster, it should be 3072. AM memory + requested
// memory. 1024 + 2048=3072
// 5. Re-register NM by sending completed container status
// 6. Verify for Memory Used, it should be 1024
// 7. Send AM heatbeat to RM. Allocated response should contain completed
// container.
/// <exception cref="System.Exception"/>
public virtual void TestProcessingNMContainerStatusesOnNMRestart()
{
conf.SetInt(YarnConfiguration.RmAmMaxAttempts, 1);
MemoryRMStateStore memStore = new MemoryRMStateStore();
memStore.Init(conf);
// 1. Start the cluster-RM,NM,Submit app with 1024MB,Launch & register AM
MockRM rm1 = new MockRM(conf, memStore);
rm1.Start();
int nmMemory = 8192;
int amMemory = 1024;
int containerMemory = 2048;
MockNM nm1 = new MockNM("127.0.0.1:1234", nmMemory, rm1.GetResourceTrackerService
());
nm1.RegisterNode();
RMApp app0 = rm1.SubmitApp(amMemory);
MockAM am0 = MockRM.LaunchAndRegisterAM(app0, rm1, nm1);
// 2. AM sends ResourceRequest for 1 container with memory 2048MB.
int noOfContainers = 1;
IList <Container> allocateContainers = am0.AllocateAndWaitForContainers(noOfContainers
, containerMemory, nm1);
// 3. Verify for number of container allocated by RM
NUnit.Framework.Assert.AreEqual(noOfContainers, allocateContainers.Count);
Container container = allocateContainers[0];
nm1.NodeHeartbeat(am0.GetApplicationAttemptId(), 1, ContainerState.Running);
nm1.NodeHeartbeat(am0.GetApplicationAttemptId(), container.GetId().GetContainerId
(), ContainerState.Running);
rm1.WaitForState(app0.GetApplicationId(), RMAppState.Running);
// 4. Verify Memory Usage by cluster, it should be 3072. AM memory +
// requested memory. 1024 + 2048=3072
ResourceScheduler rs = rm1.GetRMContext().GetScheduler();
int allocatedMB = rs.GetRootQueueMetrics().GetAllocatedMB();
NUnit.Framework.Assert.AreEqual(amMemory + containerMemory, allocatedMB);
// 5. Re-register NM by sending completed container status
IList <NMContainerStatus> nMContainerStatusForApp = CreateNMContainerStatusForApp(
am0);
nm1.RegisterNode(nMContainerStatusForApp, Arrays.AsList(app0.GetApplicationId()));
WaitForClusterMemory(nm1, rs, amMemory);
// 6. Verify for Memory Used, it should be 1024
NUnit.Framework.Assert.AreEqual(amMemory, rs.GetRootQueueMetrics().GetAllocatedMB
());
// 7. Send AM heatbeat to RM. Allocated response should contain completed
// container
AllocateRequest req = AllocateRequest.NewInstance(0, 0F, new AList <ResourceRequest
>(), new AList <ContainerId>(), null);
AllocateResponse allocate = am0.Allocate(req);
IList <ContainerStatus> completedContainersStatuses = allocate.GetCompletedContainersStatuses
();
NUnit.Framework.Assert.AreEqual(noOfContainers, completedContainersStatuses.Count
);
// Application clean up should happen Cluster memory used is 0
nm1.NodeHeartbeat(am0.GetApplicationAttemptId(), 1, ContainerState.Complete);
WaitForClusterMemory(nm1, rs, 0);
rm1.Stop();
}
public virtual void TestReconnectNode()
{
DrainDispatcher dispatcher = new DrainDispatcher();
rm = new _MockRM_567(this, dispatcher);
rm.Start();
MockNM nm1 = rm.RegisterNode("host1:1234", 5120);
MockNM nm2 = rm.RegisterNode("host2:5678", 5120);
nm1.NodeHeartbeat(true);
nm2.NodeHeartbeat(false);
dispatcher.Await();
CheckUnealthyNMCount(rm, nm2, true, 1);
int expectedNMs = ClusterMetrics.GetMetrics().GetNumActiveNMs();
QueueMetrics metrics = rm.GetResourceScheduler().GetRootQueueMetrics();
// TODO Metrics incorrect in case of the FifoScheduler
NUnit.Framework.Assert.AreEqual(5120, metrics.GetAvailableMB());
// reconnect of healthy node
nm1 = rm.RegisterNode("host1:1234", 5120);
NodeHeartbeatResponse response = nm1.NodeHeartbeat(true);
NUnit.Framework.Assert.IsTrue(NodeAction.Normal.Equals(response.GetNodeAction()));
dispatcher.Await();
NUnit.Framework.Assert.AreEqual(expectedNMs, ClusterMetrics.GetMetrics().GetNumActiveNMs
());
CheckUnealthyNMCount(rm, nm2, true, 1);
// reconnect of unhealthy node
nm2 = rm.RegisterNode("host2:5678", 5120);
response = nm2.NodeHeartbeat(false);
NUnit.Framework.Assert.IsTrue(NodeAction.Normal.Equals(response.GetNodeAction()));
dispatcher.Await();
NUnit.Framework.Assert.AreEqual(expectedNMs, ClusterMetrics.GetMetrics().GetNumActiveNMs
());
CheckUnealthyNMCount(rm, nm2, true, 1);
// unhealthy node changed back to healthy
nm2 = rm.RegisterNode("host2:5678", 5120);
dispatcher.Await();
response = nm2.NodeHeartbeat(true);
response = nm2.NodeHeartbeat(true);
dispatcher.Await();
NUnit.Framework.Assert.AreEqual(5120 + 5120, metrics.GetAvailableMB());
// reconnect of node with changed capability
nm1 = rm.RegisterNode("host2:5678", 10240);
dispatcher.Await();
response = nm1.NodeHeartbeat(true);
dispatcher.Await();
NUnit.Framework.Assert.IsTrue(NodeAction.Normal.Equals(response.GetNodeAction()));
NUnit.Framework.Assert.AreEqual(5120 + 10240, metrics.GetAvailableMB());
// reconnect of node with changed capability and running applications
IList <ApplicationId> runningApps = new AList <ApplicationId>();
runningApps.AddItem(ApplicationId.NewInstance(1, 0));
nm1 = rm.RegisterNode("host2:5678", 15360, 2, runningApps);
dispatcher.Await();
response = nm1.NodeHeartbeat(true);
dispatcher.Await();
NUnit.Framework.Assert.IsTrue(NodeAction.Normal.Equals(response.GetNodeAction()));
NUnit.Framework.Assert.AreEqual(5120 + 15360, metrics.GetAvailableMB());
// reconnect healthy node changing http port
nm1 = new MockNM("host1:1234", 5120, rm.GetResourceTrackerService());
nm1.SetHttpPort(3);
nm1.RegisterNode();
dispatcher.Await();
response = nm1.NodeHeartbeat(true);
response = nm1.NodeHeartbeat(true);
dispatcher.Await();
RMNode rmNode = rm.GetRMContext().GetRMNodes()[nm1.GetNodeId()];
NUnit.Framework.Assert.AreEqual(3, rmNode.GetHttpPort());
NUnit.Framework.Assert.AreEqual(5120, rmNode.GetTotalCapability().GetMemory());
NUnit.Framework.Assert.AreEqual(5120 + 15360, metrics.GetAvailableMB());
}
public virtual void TestRMHAWithFileSystemBasedConfiguration()
{
HAServiceProtocol.StateChangeRequestInfo requestInfo = new HAServiceProtocol.StateChangeRequestInfo
(HAServiceProtocol.RequestSource.RequestByUser);
configuration.Set(YarnConfiguration.RmConfigurationProviderClass, "org.apache.hadoop.yarn.FileSystemBasedConfigurationProvider"
);
configuration.SetBoolean(YarnConfiguration.RmHaEnabled, true);
configuration.SetBoolean(YarnConfiguration.AutoFailoverEnabled, false);
configuration.Set(YarnConfiguration.RmHaIds, "rm1,rm2");
int @base = 100;
foreach (string confKey in YarnConfiguration.GetServiceAddressConfKeys(configuration
))
{
configuration.Set(HAUtil.AddSuffix(confKey, "rm1"), "0.0.0.0:" + (@base + 20));
configuration.Set(HAUtil.AddSuffix(confKey, "rm2"), "0.0.0.0:" + (@base + 40));
@base = @base * 2;
}
Configuration conf1 = new Configuration(configuration);
conf1.Set(YarnConfiguration.RmHaId, "rm1");
Configuration conf2 = new Configuration(configuration);
conf2.Set(YarnConfiguration.RmHaId, "rm2");
// upload default configurations
UploadDefaultConfiguration();
MockRM rm1 = null;
MockRM rm2 = null;
try
{
rm1 = new MockRM(conf1);
rm1.Init(conf1);
rm1.Start();
NUnit.Framework.Assert.IsTrue(rm1.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Standby);
rm2 = new MockRM(conf2);
rm2.Init(conf1);
rm2.Start();
NUnit.Framework.Assert.IsTrue(rm2.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Standby);
rm1.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.IsTrue(rm1.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Active);
CapacitySchedulerConfiguration csConf = new CapacitySchedulerConfiguration();
csConf.Set("yarn.scheduler.capacity.maximum-applications", "5000");
UploadConfiguration(csConf, "capacity-scheduler.xml");
rm1.adminService.RefreshQueues(RefreshQueuesRequest.NewInstance());
int maxApps = ((CapacityScheduler)rm1.GetRMContext().GetScheduler()).GetConfiguration
().GetMaximumSystemApplications();
NUnit.Framework.Assert.AreEqual(maxApps, 5000);
// Before failover happens, the maxApps is
// still the default value on the standby rm : rm2
int maxAppsBeforeFailOver = ((CapacityScheduler)rm2.GetRMContext().GetScheduler()
).GetConfiguration().GetMaximumSystemApplications();
NUnit.Framework.Assert.AreEqual(maxAppsBeforeFailOver, 10000);
// Do the failover
rm1.adminService.TransitionToStandby(requestInfo);
rm2.adminService.TransitionToActive(requestInfo);
NUnit.Framework.Assert.IsTrue(rm1.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Standby);
NUnit.Framework.Assert.IsTrue(rm2.GetRMContext().GetHAServiceState() == HAServiceProtocol.HAServiceState
.Active);
int maxAppsAfter = ((CapacityScheduler)rm2.GetRMContext().GetScheduler()).GetConfiguration
().GetMaximumSystemApplications();
NUnit.Framework.Assert.AreEqual(maxAppsAfter, 5000);
}
finally
{
if (rm1 != null)
{
rm1.Stop();
}
if (rm2 != null)
{
rm2.Stop();
}
}
}
/// <exception cref="System.Exception"/>
public virtual void TestHeadroom()
{
Configuration conf = new Configuration();
conf.SetClass(YarnConfiguration.RmScheduler, typeof(FifoScheduler), typeof(ResourceScheduler
));
MockRM rm = new MockRM(conf);
rm.Start();
FifoScheduler fs = (FifoScheduler)rm.GetResourceScheduler();
// Add a node
RMNode n1 = MockNodes.NewNodeInfo(0, MockNodes.NewResource(4 * Gb), 1, "127.0.0.2"
);
fs.Handle(new NodeAddedSchedulerEvent(n1));
// Add two applications
ApplicationId appId1 = BuilderUtils.NewApplicationId(100, 1);
ApplicationAttemptId appAttemptId1 = BuilderUtils.NewApplicationAttemptId(appId1,
1);
CreateMockRMApp(appAttemptId1, rm.GetRMContext());
SchedulerEvent appEvent = new AppAddedSchedulerEvent(appId1, "queue", "user");
fs.Handle(appEvent);
SchedulerEvent attemptEvent = new AppAttemptAddedSchedulerEvent(appAttemptId1, false
);
fs.Handle(attemptEvent);
ApplicationId appId2 = BuilderUtils.NewApplicationId(200, 2);
ApplicationAttemptId appAttemptId2 = BuilderUtils.NewApplicationAttemptId(appId2,
1);
CreateMockRMApp(appAttemptId2, rm.GetRMContext());
SchedulerEvent appEvent2 = new AppAddedSchedulerEvent(appId2, "queue", "user");
fs.Handle(appEvent2);
SchedulerEvent attemptEvent2 = new AppAttemptAddedSchedulerEvent(appAttemptId2, false
);
fs.Handle(attemptEvent2);
IList <ContainerId> emptyId = new AList <ContainerId>();
IList <ResourceRequest> emptyAsk = new AList <ResourceRequest>();
// Set up resource requests
// Ask for a 1 GB container for app 1
IList <ResourceRequest> ask1 = new AList <ResourceRequest>();
ask1.AddItem(BuilderUtils.NewResourceRequest(BuilderUtils.NewPriority(0), ResourceRequest
.Any, BuilderUtils.NewResource(Gb, 1), 1));
fs.Allocate(appAttemptId1, ask1, emptyId, null, null);
// Ask for a 2 GB container for app 2
IList <ResourceRequest> ask2 = new AList <ResourceRequest>();
ask2.AddItem(BuilderUtils.NewResourceRequest(BuilderUtils.NewPriority(0), ResourceRequest
.Any, BuilderUtils.NewResource(2 * Gb, 1), 1));
fs.Allocate(appAttemptId2, ask2, emptyId, null, null);
// Trigger container assignment
fs.Handle(new NodeUpdateSchedulerEvent(n1));
// Get the allocation for the applications and verify headroom
Allocation allocation1 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("Allocation headroom", 1 * Gb, allocation1.GetResourceLimit
().GetMemory());
Allocation allocation2 = fs.Allocate(appAttemptId2, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("Allocation headroom", 1 * Gb, allocation2.GetResourceLimit
().GetMemory());
rm.Stop();
}
/// <exception cref="System.Exception"/>
public virtual void TestUsageWithMultipleContainersAndRMRestart()
{
// Set max attempts to 1 so that when the first attempt fails, the app
// won't try to start a new one.
conf.SetInt(YarnConfiguration.RmAmMaxAttempts, 1);
conf.SetBoolean(YarnConfiguration.RecoveryEnabled, true);
conf.SetBoolean(YarnConfiguration.RmWorkPreservingRecoveryEnabled, false);
MemoryRMStateStore memStore = new MemoryRMStateStore();
memStore.Init(conf);
MockRM rm0 = new MockRM(conf, memStore);
rm0.Start();
MockNM nm = new MockNM("127.0.0.1:1234", 65536, rm0.GetResourceTrackerService());
nm.RegisterNode();
RMApp app0 = rm0.SubmitApp(200);
rm0.WaitForState(app0.GetApplicationId(), RMAppState.Accepted);
RMAppAttempt attempt0 = app0.GetCurrentAppAttempt();
ApplicationAttemptId attemptId0 = attempt0.GetAppAttemptId();
rm0.WaitForState(attemptId0, RMAppAttemptState.Scheduled);
nm.NodeHeartbeat(true);
rm0.WaitForState(attemptId0, RMAppAttemptState.Allocated);
MockAM am0 = rm0.SendAMLaunched(attempt0.GetAppAttemptId());
am0.RegisterAppAttempt();
int NumContainers = 2;
am0.Allocate("127.0.0.1", 1000, NumContainers, new AList <ContainerId>());
nm.NodeHeartbeat(true);
IList <Container> conts = am0.Allocate(new AList <ResourceRequest>(), new AList <ContainerId
>()).GetAllocatedContainers();
while (conts.Count != NumContainers)
{
nm.NodeHeartbeat(true);
Sharpen.Collections.AddAll(conts, am0.Allocate(new AList <ResourceRequest>(), new
AList <ContainerId>()).GetAllocatedContainers());
Sharpen.Thread.Sleep(500);
}
// launch the 2nd and 3rd containers.
foreach (Container c in conts)
{
nm.NodeHeartbeat(attempt0.GetAppAttemptId(), c.GetId().GetContainerId(), ContainerState
.Running);
rm0.WaitForState(nm, c.GetId(), RMContainerState.Running);
}
// Get the RMContainers for all of the live containers, to be used later
// for metrics calculations and comparisons.
ICollection <RMContainer> rmContainers = rm0.scheduler.GetSchedulerAppInfo(attempt0
.GetAppAttemptId()).GetLiveContainers();
// Allow metrics to accumulate.
int sleepInterval = 1000;
int cumulativeSleepTime = 0;
while (app0.GetRMAppMetrics().GetMemorySeconds() <= 0 && cumulativeSleepTime < 5000
)
{
Sharpen.Thread.Sleep(sleepInterval);
cumulativeSleepTime += sleepInterval;
}
// Stop all non-AM containers
foreach (Container c_1 in conts)
{
if (c_1.GetId().GetContainerId() == 1)
{
continue;
}
nm.NodeHeartbeat(attempt0.GetAppAttemptId(), c_1.GetId().GetContainerId(), ContainerState
.Complete);
rm0.WaitForState(nm, c_1.GetId(), RMContainerState.Completed);
}
// After all other containers have completed, manually complete the master
// container in order to trigger a save to the state store of the resource
// usage metrics. This will cause the attempt to fail, and, since the max
// attempt retries is 1, the app will also fail. This is intentional so
// that all containers will complete prior to saving.
ContainerId cId = ContainerId.NewContainerId(attempt0.GetAppAttemptId(), 1);
nm.NodeHeartbeat(attempt0.GetAppAttemptId(), cId.GetContainerId(), ContainerState
.Complete);
rm0.WaitForState(nm, cId, RMContainerState.Completed);
// Check that the container metrics match those from the app usage report.
long memorySeconds = 0;
long vcoreSeconds = 0;
foreach (RMContainer c_2 in rmContainers)
{
AggregateAppResourceUsage ru = CalculateContainerResourceMetrics(c_2);
memorySeconds += ru.GetMemorySeconds();
vcoreSeconds += ru.GetVcoreSeconds();
}
RMAppMetrics metricsBefore = app0.GetRMAppMetrics();
NUnit.Framework.Assert.AreEqual("Unexcpected MemorySeconds value", memorySeconds,
metricsBefore.GetMemorySeconds());
NUnit.Framework.Assert.AreEqual("Unexpected VcoreSeconds value", vcoreSeconds, metricsBefore
.GetVcoreSeconds());
// create new RM to represent RM restart. Load up the state store.
MockRM rm1 = new MockRM(conf, memStore);
rm1.Start();
RMApp app0After = rm1.GetRMContext().GetRMApps()[app0.GetApplicationId()];
// Compare container resource usage metrics from before and after restart.
RMAppMetrics metricsAfter = app0After.GetRMAppMetrics();
NUnit.Framework.Assert.AreEqual("Vcore seconds were not the same after RM Restart"
, metricsBefore.GetVcoreSeconds(), metricsAfter.GetVcoreSeconds());
NUnit.Framework.Assert.AreEqual("Memory seconds were not the same after RM Restart"
, metricsBefore.GetMemorySeconds(), metricsAfter.GetMemorySeconds());
rm0.Stop();
rm0.Close();
rm1.Stop();
rm1.Close();
}
/// <exception cref="System.Exception"/>
public virtual void TestBlackListNodes()
{
Configuration conf = new Configuration();
conf.SetClass(YarnConfiguration.RmScheduler, typeof(FifoScheduler), typeof(ResourceScheduler
));
MockRM rm = new MockRM(conf);
rm.Start();
FifoScheduler fs = (FifoScheduler)rm.GetResourceScheduler();
int rack_num_0 = 0;
int rack_num_1 = 1;
// Add 4 nodes in 2 racks
// host_0_0 in rack0
string host_0_0 = "127.0.0.1";
RMNode n1 = MockNodes.NewNodeInfo(rack_num_0, MockNodes.NewResource(4 * Gb), 1, host_0_0
);
fs.Handle(new NodeAddedSchedulerEvent(n1));
// host_0_1 in rack0
string host_0_1 = "127.0.0.2";
RMNode n2 = MockNodes.NewNodeInfo(rack_num_0, MockNodes.NewResource(4 * Gb), 1, host_0_1
);
fs.Handle(new NodeAddedSchedulerEvent(n2));
// host_1_0 in rack1
string host_1_0 = "127.0.0.3";
RMNode n3 = MockNodes.NewNodeInfo(rack_num_1, MockNodes.NewResource(4 * Gb), 1, host_1_0
);
fs.Handle(new NodeAddedSchedulerEvent(n3));
// host_1_1 in rack1
string host_1_1 = "127.0.0.4";
RMNode n4 = MockNodes.NewNodeInfo(rack_num_1, MockNodes.NewResource(4 * Gb), 1, host_1_1
);
fs.Handle(new NodeAddedSchedulerEvent(n4));
// Add one application
ApplicationId appId1 = BuilderUtils.NewApplicationId(100, 1);
ApplicationAttemptId appAttemptId1 = BuilderUtils.NewApplicationAttemptId(appId1,
1);
CreateMockRMApp(appAttemptId1, rm.GetRMContext());
SchedulerEvent appEvent = new AppAddedSchedulerEvent(appId1, "queue", "user");
fs.Handle(appEvent);
SchedulerEvent attemptEvent = new AppAttemptAddedSchedulerEvent(appAttemptId1, false
);
fs.Handle(attemptEvent);
IList <ContainerId> emptyId = new AList <ContainerId>();
IList <ResourceRequest> emptyAsk = new AList <ResourceRequest>();
// Allow rack-locality for rack_1, but blacklist host_1_0
// Set up resource requests
// Ask for a 1 GB container for app 1
IList <ResourceRequest> ask1 = new AList <ResourceRequest>();
ask1.AddItem(BuilderUtils.NewResourceRequest(BuilderUtils.NewPriority(0), "rack1"
, BuilderUtils.NewResource(Gb, 1), 1));
ask1.AddItem(BuilderUtils.NewResourceRequest(BuilderUtils.NewPriority(0), ResourceRequest
.Any, BuilderUtils.NewResource(Gb, 1), 1));
fs.Allocate(appAttemptId1, ask1, emptyId, Sharpen.Collections.SingletonList(host_1_0
), null);
// Trigger container assignment
fs.Handle(new NodeUpdateSchedulerEvent(n3));
// Get the allocation for the application and verify no allocation on blacklist node
Allocation allocation1 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("allocation1", 0, allocation1.GetContainers().Count
);
// verify host_1_1 can get allocated as not in blacklist
fs.Handle(new NodeUpdateSchedulerEvent(n4));
Allocation allocation2 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("allocation2", 1, allocation2.GetContainers().Count
);
IList <Container> containerList = allocation2.GetContainers();
foreach (Container container in containerList)
{
NUnit.Framework.Assert.AreEqual("Container is allocated on n4", container.GetNodeId
(), n4.GetNodeID());
}
// Ask for a 1 GB container again for app 1
IList <ResourceRequest> ask2 = new AList <ResourceRequest>();
// this time, rack0 is also in blacklist, so only host_1_1 is available to
// be assigned
ask2.AddItem(BuilderUtils.NewResourceRequest(BuilderUtils.NewPriority(0), ResourceRequest
.Any, BuilderUtils.NewResource(Gb, 1), 1));
fs.Allocate(appAttemptId1, ask2, emptyId, Sharpen.Collections.SingletonList("rack0"
), null);
// verify n1 is not qualified to be allocated
fs.Handle(new NodeUpdateSchedulerEvent(n1));
Allocation allocation3 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("allocation3", 0, allocation3.GetContainers().Count
);
// verify n2 is not qualified to be allocated
fs.Handle(new NodeUpdateSchedulerEvent(n2));
Allocation allocation4 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("allocation4", 0, allocation4.GetContainers().Count
);
// verify n3 is not qualified to be allocated
fs.Handle(new NodeUpdateSchedulerEvent(n3));
Allocation allocation5 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("allocation5", 0, allocation5.GetContainers().Count
);
fs.Handle(new NodeUpdateSchedulerEvent(n4));
Allocation allocation6 = fs.Allocate(appAttemptId1, emptyAsk, emptyId, null, null
);
NUnit.Framework.Assert.AreEqual("allocation6", 1, allocation6.GetContainers().Count
);
containerList = allocation6.GetContainers();
foreach (Container container_1 in containerList)
{
NUnit.Framework.Assert.AreEqual("Container is allocated on n4", container_1.GetNodeId
(), n4.GetNodeID());
}
rm.Stop();
}
