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);
}
/// <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);
}
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());
}