public ClusterMetricsInfo(ResourceManager rm)
{
// JAXB needs this
ResourceScheduler rs = rm.GetResourceScheduler();
QueueMetrics metrics = rs.GetRootQueueMetrics();
ClusterMetrics clusterMetrics = ClusterMetrics.GetMetrics();
this.appsSubmitted = metrics.GetAppsSubmitted();
this.appsCompleted = metrics.GetAppsCompleted();
this.appsPending = metrics.GetAppsPending();
this.appsRunning = metrics.GetAppsRunning();
this.appsFailed = metrics.GetAppsFailed();
this.appsKilled = metrics.GetAppsKilled();
this.reservedMB = metrics.GetReservedMB();
this.availableMB = metrics.GetAvailableMB();
this.allocatedMB = metrics.GetAllocatedMB();
this.reservedVirtualCores = metrics.GetReservedVirtualCores();
this.availableVirtualCores = metrics.GetAvailableVirtualCores();
this.allocatedVirtualCores = metrics.GetAllocatedVirtualCores();
this.containersAllocated = metrics.GetAllocatedContainers();
this.containersPending = metrics.GetPendingContainers();
this.containersReserved = metrics.GetReservedContainers();
this.totalMB = availableMB + allocatedMB;
this.totalVirtualCores = availableVirtualCores + allocatedVirtualCores;
this.activeNodes = clusterMetrics.GetNumActiveNMs();
this.lostNodes = clusterMetrics.GetNumLostNMs();
this.unhealthyNodes = clusterMetrics.GetUnhealthyNMs();
this.decommissionedNodes = clusterMetrics.GetNumDecommisionedNMs();
this.rebootedNodes = clusterMetrics.GetNumRebootedNMs();
this.totalNodes = activeNodes + lostNodes + decommissionedNodes + rebootedNodes +
unhealthyNodes;
}
/// <exception cref="System.Exception"/>
private void VerifyClusterMetrics(int activeNodes, int appsSubmitted, int appsPending
, int containersPending, int availableMB, int activeApplications)
{
int timeoutSecs = 0;
QueueMetrics metrics = rm.GetResourceScheduler().GetRootQueueMetrics();
ClusterMetrics clusterMetrics = ClusterMetrics.GetMetrics();
bool isAllMetricAssertionDone = false;
string message = null;
while (timeoutSecs++ < 5)
{
try
{
// verify queue metrics
AssertMetric("appsSubmitted", appsSubmitted, metrics.GetAppsSubmitted());
AssertMetric("appsPending", appsPending, metrics.GetAppsPending());
AssertMetric("containersPending", containersPending, metrics.GetPendingContainers
());
AssertMetric("availableMB", availableMB, metrics.GetAvailableMB());
AssertMetric("activeApplications", activeApplications, metrics.GetActiveApps());
// verify node metric
AssertMetric("activeNodes", activeNodes, clusterMetrics.GetNumActiveNMs());
isAllMetricAssertionDone = true;
break;
}
catch (Exception e)
{
message = e.Message;
System.Console.Out.WriteLine("Waiting for metrics assertion to complete");
Sharpen.Thread.Sleep(1000);
}
}
NUnit.Framework.Assert.IsTrue(message, isAllMetricAssertionDone);
}
public UserMetricsInfo(ResourceManager rm, string user)
{
// JAXB needs this
ResourceScheduler rs = rm.GetResourceScheduler();
QueueMetrics metrics = rs.GetRootQueueMetrics();
QueueMetrics userMetrics = metrics.GetUserMetrics(user);
this.userMetricsAvailable = false;
if (userMetrics != null)
{
this.userMetricsAvailable = true;
this.appsSubmitted = userMetrics.GetAppsSubmitted();
this.appsCompleted = userMetrics.GetAppsCompleted();
this.appsPending = userMetrics.GetAppsPending();
this.appsRunning = userMetrics.GetAppsRunning();
this.appsFailed = userMetrics.GetAppsFailed();
this.appsKilled = userMetrics.GetAppsKilled();
this.runningContainers = userMetrics.GetAllocatedContainers();
this.pendingContainers = userMetrics.GetPendingContainers();
this.reservedContainers = userMetrics.GetReservedContainers();
this.reservedMB = userMetrics.GetReservedMB();
this.pendingMB = userMetrics.GetPendingMB();
this.allocatedMB = userMetrics.GetAllocatedMB();
this.reservedVirtualCores = userMetrics.GetReservedVirtualCores();
this.pendingVirtualCores = userMetrics.GetPendingVirtualCores();
this.allocatedVirtualCores = userMetrics.GetAllocatedVirtualCores();
}
}
/// <summary>Validate killing an application when it is at accepted state.</summary>
/// <exception cref="System.Exception">exception</exception>
public virtual void TestApplicationKillAtAcceptedState()
{
Dispatcher dispatcher = new _AsyncDispatcher_573();
MockRM rm = new _MockRM_596(dispatcher, conf);
// test metrics
QueueMetrics metrics = rm.GetResourceScheduler().GetRootQueueMetrics();
int appsKilled = metrics.GetAppsKilled();
int appsSubmitted = metrics.GetAppsSubmitted();
rm.Start();
MockNM nm1 = new MockNM("127.0.0.1:1234", 15120, rm.GetResourceTrackerService());
nm1.RegisterNode();
// a failed app
RMApp application = rm.SubmitApp(200);
MockAM am = MockRM.LaunchAM(application, rm, nm1);
am.WaitForState(RMAppAttemptState.Launched);
nm1.NodeHeartbeat(am.GetApplicationAttemptId(), 1, ContainerState.Running);
rm.WaitForState(application.GetApplicationId(), RMAppState.Accepted);
// Now kill the application before new attempt is launched, the app report
// returns the invalid AM host and port.
KillApplicationRequest request = KillApplicationRequest.NewInstance(application.GetApplicationId
());
rm.GetClientRMService().ForceKillApplication(request);
// Specific test for YARN-1689 follows
// Now let's say a race causes AM to register now. This should not crash RM.
am.RegisterAppAttempt(false);
// We explicitly intercepted the kill-event to RMAppAttempt, so app should
// still be in KILLING state.
rm.WaitForState(application.GetApplicationId(), RMAppState.Killing);
// AM should now be in running
rm.WaitForState(am.GetApplicationAttemptId(), RMAppAttemptState.Running);
// Simulate that appAttempt is killed.
rm.GetRMContext().GetDispatcher().GetEventHandler().Handle(new RMAppEvent(application
.GetApplicationId(), RMAppEventType.AttemptKilled));
rm.WaitForState(application.GetApplicationId(), RMAppState.Killed);
// test metrics
metrics = rm.GetResourceScheduler().GetRootQueueMetrics();
NUnit.Framework.Assert.AreEqual(appsKilled + 1, metrics.GetAppsKilled());
NUnit.Framework.Assert.AreEqual(appsSubmitted + 1, metrics.GetAppsSubmitted());
}
/// <exception cref="System.Exception"/>
public virtual void TestAppAttemptMetrics()
{
AsyncDispatcher dispatcher = new InlineDispatcher();
FifoScheduler scheduler = new FifoScheduler();
RMApplicationHistoryWriter writer = Org.Mockito.Mockito.Mock <RMApplicationHistoryWriter
>();
RMContext rmContext = new RMContextImpl(dispatcher, null, null, null, null, null,
null, null, null, scheduler);
((RMContextImpl)rmContext).SetSystemMetricsPublisher(Org.Mockito.Mockito.Mock <SystemMetricsPublisher
>());
Configuration conf = new Configuration();
scheduler.SetRMContext(rmContext);
scheduler.Init(conf);
scheduler.Start();
scheduler.Reinitialize(conf, rmContext);
QueueMetrics metrics = scheduler.GetRootQueueMetrics();
int beforeAppsSubmitted = metrics.GetAppsSubmitted();
ApplicationId appId = BuilderUtils.NewApplicationId(200, 1);
ApplicationAttemptId appAttemptId = BuilderUtils.NewApplicationAttemptId(appId, 1
);
SchedulerEvent appEvent = new AppAddedSchedulerEvent(appId, "queue", "user");
scheduler.Handle(appEvent);
SchedulerEvent attemptEvent = new AppAttemptAddedSchedulerEvent(appAttemptId, false
);
scheduler.Handle(attemptEvent);
appAttemptId = BuilderUtils.NewApplicationAttemptId(appId, 2);
SchedulerEvent attemptEvent2 = new AppAttemptAddedSchedulerEvent(appAttemptId, false
);
scheduler.Handle(attemptEvent2);
int afterAppsSubmitted = metrics.GetAppsSubmitted();
NUnit.Framework.Assert.AreEqual(1, afterAppsSubmitted - beforeAppsSubmitted);
scheduler.Stop();
}
/// <exception cref="Org.Codehaus.Jettison.Json.JSONException"/>
/// <exception cref="System.Exception"/>
public virtual void VerifyClusterMetrics(int submittedApps, int completedApps, int
reservedMB, int availableMB, int allocMB, int reservedVirtualCores, int availableVirtualCores
, int allocVirtualCores, int totalVirtualCores, int containersAlloc, int totalMB
, int totalNodes, int lostNodes, int unhealthyNodes, int decommissionedNodes, int
rebootedNodes, int activeNodes)
{
ResourceScheduler rs = rm.GetResourceScheduler();
QueueMetrics metrics = rs.GetRootQueueMetrics();
ClusterMetrics clusterMetrics = ClusterMetrics.GetMetrics();
long totalMBExpect = metrics.GetAvailableMB() + metrics.GetAllocatedMB();
long totalVirtualCoresExpect = metrics.GetAvailableVirtualCores() + metrics.GetAllocatedVirtualCores
();
NUnit.Framework.Assert.AreEqual("appsSubmitted doesn't match", metrics.GetAppsSubmitted
(), submittedApps);
NUnit.Framework.Assert.AreEqual("appsCompleted doesn't match", metrics.GetAppsCompleted
(), completedApps);
NUnit.Framework.Assert.AreEqual("reservedMB doesn't match", metrics.GetReservedMB
(), reservedMB);
NUnit.Framework.Assert.AreEqual("availableMB doesn't match", metrics.GetAvailableMB
(), availableMB);
NUnit.Framework.Assert.AreEqual("allocatedMB doesn't match", metrics.GetAllocatedMB
(), allocMB);
NUnit.Framework.Assert.AreEqual("reservedVirtualCores doesn't match", metrics.GetReservedVirtualCores
(), reservedVirtualCores);
NUnit.Framework.Assert.AreEqual("availableVirtualCores doesn't match", metrics.GetAvailableVirtualCores
(), availableVirtualCores);
NUnit.Framework.Assert.AreEqual("allocatedVirtualCores doesn't match", totalVirtualCoresExpect
, allocVirtualCores);
NUnit.Framework.Assert.AreEqual("containersAllocated doesn't match", 0, containersAlloc
);
NUnit.Framework.Assert.AreEqual("totalMB doesn't match", totalMBExpect, totalMB);
NUnit.Framework.Assert.AreEqual("totalNodes doesn't match", clusterMetrics.GetNumActiveNMs
() + clusterMetrics.GetNumLostNMs() + clusterMetrics.GetNumDecommisionedNMs() +
clusterMetrics.GetNumRebootedNMs() + clusterMetrics.GetUnhealthyNMs(), totalNodes
);
NUnit.Framework.Assert.AreEqual("lostNodes doesn't match", clusterMetrics.GetNumLostNMs
(), lostNodes);
NUnit.Framework.Assert.AreEqual("unhealthyNodes doesn't match", clusterMetrics.GetUnhealthyNMs
(), unhealthyNodes);
NUnit.Framework.Assert.AreEqual("decommissionedNodes doesn't match", clusterMetrics
.GetNumDecommisionedNMs(), decommissionedNodes);
NUnit.Framework.Assert.AreEqual("rebootedNodes doesn't match", clusterMetrics.GetNumRebootedNMs
(), rebootedNodes);
NUnit.Framework.Assert.AreEqual("activeNodes doesn't match", clusterMetrics.GetNumActiveNMs
(), activeNodes);
}