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.Exception"/>
public virtual void TestSchedulingDelay()
{
// Add one node
string host = "127.0.0.1";
RMNode node1 = MockNodes.NewNodeInfo(1, Resources.CreateResource(4096, 4), 1, host
);
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node1);
scheduler.Handle(nodeEvent1);
NodeUpdateSchedulerEvent nodeUpdateEvent = new NodeUpdateSchedulerEvent(node1);
scheduler.Handle(nodeUpdateEvent);
// Create one application and submit one each of node-local, rack-local
// and ANY requests
ApplicationAttemptId appAttemptId = CreateAppAttemptId(this.AppId++, this.AttemptId
++);
CreateMockRMApp(appAttemptId);
scheduler.AddApplication(appAttemptId.GetApplicationId(), "queue11", "user11", false
);
scheduler.AddApplicationAttempt(appAttemptId, false, false);
IList <ResourceRequest> ask = new AList <ResourceRequest>();
ask.AddItem(CreateResourceRequest(1024, 1, ResourceRequest.Any, 1, 1, true));
scheduler.Allocate(appAttemptId, ask, new AList <ContainerId>(), null, null);
FSAppAttempt app = scheduler.GetSchedulerApp(appAttemptId);
// Advance time and let continuous scheduling kick in
mockClock.Tick(1);
while (1024 != app.GetCurrentConsumption().GetMemory())
{
Sharpen.Thread.Sleep(100);
}
NUnit.Framework.Assert.AreEqual(1024, app.GetCurrentConsumption().GetMemory());
}
private void NodeUpdate(NodeManager nm1)
{
RMNode node = resourceManager.GetRMContext().GetRMNodes()[nm1.GetNodeId()];
// Send a heartbeat to kick the tires on the Scheduler
NodeUpdateSchedulerEvent nodeUpdate = new NodeUpdateSchedulerEvent(node);
resourceManager.GetResourceScheduler().Handle(nodeUpdate);
}
/// <exception cref="System.Exception"/>
public virtual void Test()
{
conf.Set(FairSchedulerConfiguration.AllocationFile, AllocFile);
PrintWriter @out = new PrintWriter(new FileWriter(AllocFile));
@out.WriteLine("<?xml version=\"1.0\"?>");
@out.WriteLine("<allocations>");
@out.WriteLine("<queue name=\"queueA\">");
@out.WriteLine("<minResources>2048mb,0vcores</minResources>");
@out.WriteLine("</queue>");
@out.WriteLine("<queue name=\"queueB\">");
@out.WriteLine("<minResources>2048mb,0vcores</minResources>");
@out.WriteLine("</queue>");
@out.WriteLine("</allocations>");
@out.Close();
resourceManager = new MockRM(conf);
resourceManager.Start();
scheduler = (FairScheduler)resourceManager.GetResourceScheduler();
// Add one big node (only care about aggregate capacity)
RMNode node1 = MockNodes.NewNodeInfo(1, Resources.CreateResource(4 * 1024, 4), 1,
"127.0.0.1");
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node1);
scheduler.Handle(nodeEvent1);
scheduler.Update();
// Queue A wants 3 * 1024. Node update gives this all to A
CreateSchedulingRequest(3 * 1024, "queueA", "user1");
scheduler.Update();
NodeUpdateSchedulerEvent nodeEvent2 = new NodeUpdateSchedulerEvent(node1);
scheduler.Handle(nodeEvent2);
// Queue B arrives and wants 1 * 1024
CreateSchedulingRequest(1 * 1024, "queueB", "user1");
scheduler.Update();
ICollection <FSLeafQueue> queues = scheduler.GetQueueManager().GetLeafQueues();
NUnit.Framework.Assert.AreEqual(3, queues.Count);
// Queue A should be above min share, B below.
FSLeafQueue queueA = scheduler.GetQueueManager().GetLeafQueue("queueA", false);
FSLeafQueue queueB = scheduler.GetQueueManager().GetLeafQueue("queueB", false);
NUnit.Framework.Assert.IsFalse(queueA.IsStarvedForMinShare());
NUnit.Framework.Assert.IsTrue(queueB.IsStarvedForMinShare());
// Node checks in again, should allocate for B
scheduler.Handle(nodeEvent2);
// Now B should have min share ( = demand here)
NUnit.Framework.Assert.IsFalse(queueB.IsStarvedForMinShare());
}
private void RegisterNodeAndSubmitApp(int memory, int vcores, int appContainers,
int appMemory)
{
RMNode node1 = MockNodes.NewNodeInfo(1, Resources.CreateResource(memory, vcores),
1, "node1");
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node1);
scheduler.Handle(nodeEvent1);
NUnit.Framework.Assert.AreEqual("Incorrect amount of resources in the cluster", memory
, scheduler.rootMetrics.GetAvailableMB());
NUnit.Framework.Assert.AreEqual("Incorrect amount of resources in the cluster", vcores
, scheduler.rootMetrics.GetAvailableVirtualCores());
CreateSchedulingRequest(appMemory, "queueA", "user1", appContainers);
scheduler.Update();
// Sufficient node check-ins to fully schedule containers
for (int i = 0; i < 3; i++)
{
NodeUpdateSchedulerEvent nodeUpdate1 = new NodeUpdateSchedulerEvent(node1);
scheduler.Handle(nodeUpdate1);
}
NUnit.Framework.Assert.AreEqual("app1's request is not met", memory - appContainers
* appMemory, scheduler.rootMetrics.GetAvailableMB());
}
public override void Handle(SchedulerEvent @event)
{
switch (@event.GetType())
{
case SchedulerEventType.NodeAdded:
{
NodeAddedSchedulerEvent nodeAddedEvent = (NodeAddedSchedulerEvent)@event;
AddNode(nodeAddedEvent.GetAddedRMNode());
RecoverContainersOnNode(nodeAddedEvent.GetContainerReports(), nodeAddedEvent.GetAddedRMNode
());
break;
}
case SchedulerEventType.NodeRemoved:
{
NodeRemovedSchedulerEvent nodeRemovedEvent = (NodeRemovedSchedulerEvent)@event;
RemoveNode(nodeRemovedEvent.GetRemovedRMNode());
break;
}
case SchedulerEventType.NodeResourceUpdate:
{
NodeResourceUpdateSchedulerEvent nodeResourceUpdatedEvent = (NodeResourceUpdateSchedulerEvent
)@event;
UpdateNodeResource(nodeResourceUpdatedEvent.GetRMNode(), nodeResourceUpdatedEvent
.GetResourceOption());
break;
}
case SchedulerEventType.NodeUpdate:
{
NodeUpdateSchedulerEvent nodeUpdatedEvent = (NodeUpdateSchedulerEvent)@event;
NodeUpdate(nodeUpdatedEvent.GetRMNode());
break;
}
case SchedulerEventType.AppAdded:
{
AppAddedSchedulerEvent appAddedEvent = (AppAddedSchedulerEvent)@event;
AddApplication(appAddedEvent.GetApplicationId(), appAddedEvent.GetQueue(), appAddedEvent
.GetUser(), appAddedEvent.GetIsAppRecovering());
break;
}
case SchedulerEventType.AppRemoved:
{
AppRemovedSchedulerEvent appRemovedEvent = (AppRemovedSchedulerEvent)@event;
DoneApplication(appRemovedEvent.GetApplicationID(), appRemovedEvent.GetFinalState
());
break;
}
case SchedulerEventType.AppAttemptAdded:
{
AppAttemptAddedSchedulerEvent appAttemptAddedEvent = (AppAttemptAddedSchedulerEvent
)@event;
AddApplicationAttempt(appAttemptAddedEvent.GetApplicationAttemptId(), appAttemptAddedEvent
.GetTransferStateFromPreviousAttempt(), appAttemptAddedEvent.GetIsAttemptRecovering
());
break;
}
case SchedulerEventType.AppAttemptRemoved:
{
AppAttemptRemovedSchedulerEvent appAttemptRemovedEvent = (AppAttemptRemovedSchedulerEvent
)@event;
try
{
DoneApplicationAttempt(appAttemptRemovedEvent.GetApplicationAttemptID(), appAttemptRemovedEvent
.GetFinalAttemptState(), appAttemptRemovedEvent.GetKeepContainersAcrossAppAttempts
());
}
catch (IOException ie)
{
Log.Error("Unable to remove application " + appAttemptRemovedEvent.GetApplicationAttemptID
(), ie);
}
break;
}
case SchedulerEventType.ContainerExpired:
{
ContainerExpiredSchedulerEvent containerExpiredEvent = (ContainerExpiredSchedulerEvent
)@event;
ContainerId containerid = containerExpiredEvent.GetContainerId();
CompletedContainer(GetRMContainer(containerid), SchedulerUtils.CreateAbnormalContainerStatus
(containerid, SchedulerUtils.ExpiredContainer), RMContainerEventType.Expire);
break;
}
case SchedulerEventType.ContainerRescheduled:
{
ContainerRescheduledEvent containerRescheduledEvent = (ContainerRescheduledEvent)
@event;
RMContainer container = containerRescheduledEvent.GetContainer();
RecoverResourceRequestForContainer(container);
break;
}
default:
{
Log.Error("Invalid eventtype " + @event.GetType() + ". Ignoring!");
break;
}
}
}
/// <exception cref="System.Exception"/>
public virtual void TestIsStarvedForFairShare()
{
conf.Set(FairSchedulerConfiguration.AllocationFile, AllocFile);
PrintWriter @out = new PrintWriter(new FileWriter(AllocFile));
@out.WriteLine("<?xml version=\"1.0\"?>");
@out.WriteLine("<allocations>");
@out.WriteLine("<queue name=\"queueA\">");
@out.WriteLine("<weight>.2</weight>");
@out.WriteLine("</queue>");
@out.WriteLine("<queue name=\"queueB\">");
@out.WriteLine("<weight>.8</weight>");
@out.WriteLine("<fairSharePreemptionThreshold>.4</fairSharePreemptionThreshold>");
@out.WriteLine("<queue name=\"queueB1\">");
@out.WriteLine("</queue>");
@out.WriteLine("<queue name=\"queueB2\">");
@out.WriteLine("<fairSharePreemptionThreshold>.6</fairSharePreemptionThreshold>");
@out.WriteLine("</queue>");
@out.WriteLine("</queue>");
@out.WriteLine("<defaultFairSharePreemptionThreshold>.5</defaultFairSharePreemptionThreshold>"
);
@out.WriteLine("</allocations>");
@out.Close();
resourceManager = new MockRM(conf);
resourceManager.Start();
scheduler = (FairScheduler)resourceManager.GetResourceScheduler();
// Add one big node (only care about aggregate capacity)
RMNode node1 = MockNodes.NewNodeInfo(1, Resources.CreateResource(10 * 1024, 10),
1, "127.0.0.1");
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node1);
scheduler.Handle(nodeEvent1);
scheduler.Update();
// Queue A wants 4 * 1024. Node update gives this all to A
CreateSchedulingRequest(1 * 1024, "queueA", "user1", 4);
scheduler.Update();
NodeUpdateSchedulerEvent nodeEvent2 = new NodeUpdateSchedulerEvent(node1);
for (int i = 0; i < 4; i++)
{
scheduler.Handle(nodeEvent2);
}
QueueManager queueMgr = scheduler.GetQueueManager();
FSLeafQueue queueA = queueMgr.GetLeafQueue("queueA", false);
NUnit.Framework.Assert.AreEqual(4 * 1024, queueA.GetResourceUsage().GetMemory());
// Both queue B1 and queue B2 want 3 * 1024
CreateSchedulingRequest(1 * 1024, "queueB.queueB1", "user1", 3);
CreateSchedulingRequest(1 * 1024, "queueB.queueB2", "user1", 3);
scheduler.Update();
for (int i_1 = 0; i_1 < 4; i_1++)
{
scheduler.Handle(nodeEvent2);
}
FSLeafQueue queueB1 = queueMgr.GetLeafQueue("queueB.queueB1", false);
FSLeafQueue queueB2 = queueMgr.GetLeafQueue("queueB.queueB2", false);
NUnit.Framework.Assert.AreEqual(2 * 1024, queueB1.GetResourceUsage().GetMemory());
NUnit.Framework.Assert.AreEqual(2 * 1024, queueB2.GetResourceUsage().GetMemory());
// For queue B1, the fairSharePreemptionThreshold is 0.4, and the fair share
// threshold is 1.6 * 1024
NUnit.Framework.Assert.IsFalse(queueB1.IsStarvedForFairShare());
// For queue B2, the fairSharePreemptionThreshold is 0.6, and the fair share
// threshold is 2.4 * 1024
NUnit.Framework.Assert.IsTrue(queueB2.IsStarvedForFairShare());
// Node checks in again
scheduler.Handle(nodeEvent2);
scheduler.Handle(nodeEvent2);
NUnit.Framework.Assert.AreEqual(3 * 1024, queueB1.GetResourceUsage().GetMemory());
NUnit.Framework.Assert.AreEqual(3 * 1024, queueB2.GetResourceUsage().GetMemory());
// Both queue B1 and queue B2 usages go to 3 * 1024
NUnit.Framework.Assert.IsFalse(queueB1.IsStarvedForFairShare());
NUnit.Framework.Assert.IsFalse(queueB2.IsStarvedForFairShare());
}
/// <exception cref="System.Exception"/>
public virtual void TestUpdateResourceOnNode()
{
AsyncDispatcher dispatcher = new InlineDispatcher();
Configuration conf = new Configuration();
RMContainerTokenSecretManager containerTokenSecretManager = new RMContainerTokenSecretManager
(conf);
containerTokenSecretManager.RollMasterKey();
NMTokenSecretManagerInRM nmTokenSecretManager = new NMTokenSecretManagerInRM(conf
);
nmTokenSecretManager.RollMasterKey();
RMApplicationHistoryWriter writer = Org.Mockito.Mockito.Mock <RMApplicationHistoryWriter
>();
FifoScheduler scheduler = new _FifoScheduler_275(this);
RMContext rmContext = new RMContextImpl(dispatcher, null, null, null, null, null,
containerTokenSecretManager, nmTokenSecretManager, null, scheduler);
rmContext.SetSystemMetricsPublisher(Org.Mockito.Mockito.Mock <SystemMetricsPublisher
>());
rmContext.SetRMApplicationHistoryWriter(Org.Mockito.Mockito.Mock <RMApplicationHistoryWriter
>());
((RMContextImpl)rmContext).SetYarnConfiguration(new YarnConfiguration());
scheduler.SetRMContext(rmContext);
scheduler.Init(conf);
scheduler.Start();
scheduler.Reinitialize(new Configuration(), rmContext);
RMNode node0 = MockNodes.NewNodeInfo(1, Resources.CreateResource(2048, 4), 1, "127.0.0.1"
);
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node0);
scheduler.Handle(nodeEvent1);
MethodInfo method = Sharpen.Runtime.GetDeclaredMethod(scheduler.GetType(), "getNodes"
);
IDictionary <NodeId, FiCaSchedulerNode> schedulerNodes = (IDictionary <NodeId, FiCaSchedulerNode
>)method.Invoke(scheduler);
NUnit.Framework.Assert.AreEqual(schedulerNodes.Values.Count, 1);
Org.Apache.Hadoop.Yarn.Api.Records.Resource newResource = Resources.CreateResource
(1024, 4);
NodeResourceUpdateSchedulerEvent node0ResourceUpdate = new NodeResourceUpdateSchedulerEvent
(node0, ResourceOption.NewInstance(newResource, RMNode.OverCommitTimeoutMillisDefault
));
scheduler.Handle(node0ResourceUpdate);
// SchedulerNode's total resource and available resource are changed.
NUnit.Framework.Assert.AreEqual(schedulerNodes[node0.GetNodeID()].GetTotalResource
().GetMemory(), 1024);
NUnit.Framework.Assert.AreEqual(schedulerNodes[node0.GetNodeID()].GetAvailableResource
().GetMemory(), 1024);
QueueInfo queueInfo = scheduler.GetQueueInfo(null, false, false);
NUnit.Framework.Assert.AreEqual(0.0f, queueInfo.GetCurrentCapacity(), 0.0f);
int _appId = 1;
int _appAttemptId = 1;
ApplicationAttemptId appAttemptId = CreateAppAttemptId(_appId, _appAttemptId);
CreateMockRMApp(appAttemptId, rmContext);
AppAddedSchedulerEvent appEvent = new AppAddedSchedulerEvent(appAttemptId.GetApplicationId
(), "queue1", "user1");
scheduler.Handle(appEvent);
AppAttemptAddedSchedulerEvent attemptEvent = new AppAttemptAddedSchedulerEvent(appAttemptId
, false);
scheduler.Handle(attemptEvent);
int memory = 1024;
int priority = 1;
IList <ResourceRequest> ask = new AList <ResourceRequest>();
ResourceRequest nodeLocal = CreateResourceRequest(memory, node0.GetHostName(), priority
, 1);
ResourceRequest rackLocal = CreateResourceRequest(memory, node0.GetRackName(), priority
, 1);
ResourceRequest any = CreateResourceRequest(memory, ResourceRequest.Any, priority
, 1);
ask.AddItem(nodeLocal);
ask.AddItem(rackLocal);
ask.AddItem(any);
scheduler.Allocate(appAttemptId, ask, new AList <ContainerId>(), null, null);
// Before the node update event, there are one local request
NUnit.Framework.Assert.AreEqual(1, nodeLocal.GetNumContainers());
NodeUpdateSchedulerEvent node0Update = new NodeUpdateSchedulerEvent(node0);
// Now schedule.
scheduler.Handle(node0Update);
// After the node update event, check no local request
NUnit.Framework.Assert.AreEqual(0, nodeLocal.GetNumContainers());
// Also check that one container was scheduled
SchedulerAppReport info = scheduler.GetSchedulerAppInfo(appAttemptId);
NUnit.Framework.Assert.AreEqual(1, info.GetLiveContainers().Count);
// And check the default Queue now is full.
queueInfo = scheduler.GetQueueInfo(null, false, false);
NUnit.Framework.Assert.AreEqual(1.0f, queueInfo.GetCurrentCapacity(), 0.0f);
}
/// <exception cref="System.Exception"/>
public virtual void TestNodeLocalAssignment()
{
AsyncDispatcher dispatcher = new InlineDispatcher();
Configuration conf = new Configuration();
RMContainerTokenSecretManager containerTokenSecretManager = new RMContainerTokenSecretManager
(conf);
containerTokenSecretManager.RollMasterKey();
NMTokenSecretManagerInRM nmTokenSecretManager = new NMTokenSecretManagerInRM(conf
);
nmTokenSecretManager.RollMasterKey();
RMApplicationHistoryWriter writer = Org.Mockito.Mockito.Mock <RMApplicationHistoryWriter
>();
FifoScheduler scheduler = new FifoScheduler();
RMContext rmContext = new RMContextImpl(dispatcher, null, null, null, null, null,
containerTokenSecretManager, nmTokenSecretManager, null, scheduler);
rmContext.SetSystemMetricsPublisher(Org.Mockito.Mockito.Mock <SystemMetricsPublisher
>());
rmContext.SetRMApplicationHistoryWriter(Org.Mockito.Mockito.Mock <RMApplicationHistoryWriter
>());
((RMContextImpl)rmContext).SetYarnConfiguration(new YarnConfiguration());
scheduler.SetRMContext(rmContext);
scheduler.Init(conf);
scheduler.Start();
scheduler.Reinitialize(new Configuration(), rmContext);
RMNode node0 = MockNodes.NewNodeInfo(1, Resources.CreateResource(1024 * 64), 1, "127.0.0.1"
);
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node0);
scheduler.Handle(nodeEvent1);
int _appId = 1;
int _appAttemptId = 1;
ApplicationAttemptId appAttemptId = CreateAppAttemptId(_appId, _appAttemptId);
CreateMockRMApp(appAttemptId, rmContext);
AppAddedSchedulerEvent appEvent = new AppAddedSchedulerEvent(appAttemptId.GetApplicationId
(), "queue1", "user1");
scheduler.Handle(appEvent);
AppAttemptAddedSchedulerEvent attemptEvent = new AppAttemptAddedSchedulerEvent(appAttemptId
, false);
scheduler.Handle(attemptEvent);
int memory = 64;
int nConts = 3;
int priority = 20;
IList <ResourceRequest> ask = new AList <ResourceRequest>();
ResourceRequest nodeLocal = CreateResourceRequest(memory, node0.GetHostName(), priority
, nConts);
ResourceRequest rackLocal = CreateResourceRequest(memory, node0.GetRackName(), priority
, nConts);
ResourceRequest any = CreateResourceRequest(memory, ResourceRequest.Any, priority
, nConts);
ask.AddItem(nodeLocal);
ask.AddItem(rackLocal);
ask.AddItem(any);
scheduler.Allocate(appAttemptId, ask, new AList <ContainerId>(), null, null);
NodeUpdateSchedulerEvent node0Update = new NodeUpdateSchedulerEvent(node0);
// Before the node update event, there are 3 local requests outstanding
NUnit.Framework.Assert.AreEqual(3, nodeLocal.GetNumContainers());
scheduler.Handle(node0Update);
// After the node update event, check that there are no more local requests
// outstanding
NUnit.Framework.Assert.AreEqual(0, nodeLocal.GetNumContainers());
//Also check that the containers were scheduled
SchedulerAppReport info = scheduler.GetSchedulerAppInfo(appAttemptId);
NUnit.Framework.Assert.AreEqual(3, info.GetLiveContainers().Count);
scheduler.Stop();
}
