YARN-3272. Surface container locality info in RM web UI (Jian He via wangda)

hadoop-yarn-project/CHANGES.txt

@@ -348,6 +348,9 @@ Release 2.7.0 - UNRELEASED
     YARN-3281. Added RMStateStore to StateMachine visualization list.
     (Chengbing Liu via jianhe)
 
+    YARN-3272. Surface container locality info in RM web UI.
+    (Jian He via wangda)
+
   OPTIMIZATIONS
 
     YARN-2990. FairScheduler's delay-scheduling always waits for node-local and 

hadoop-yarn-project/hadoop-yarn/dev-support/findbugs-exclude.xml

@@ -62,6 +62,13 @@
     <Class name="~org\.apache\.hadoop\.yarn\.server\.resourcemanager\.rmapp\.attempt\.RMAppAttemptImpl.*" />
     <Bug pattern="BC_UNCONFIRMED_CAST" />
   </Match>
+  <Match>
+    <Class name="~org\.apache\.hadoop\.yarn\.server\.resourcemanager\.rmapp\.attempt\.RMAppAttemptMetrics" />
+    <Method name="getLocalityStatistics" />
+    <Bug pattern="EI_EXPOSE_REP" />
+    <Method name="incNumAllocatedContainers"/>
+    <Bug pattern="VO_VOLATILE_INCREMENT" />
+  </Match>
   <Match>
     <Class name="org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl$AppRejectedTransition" />
     <Bug pattern="BC_UNCONFIRMED_CAST" />

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/attempt/RMAppAttemptMetrics.java

@@ -32,6 +32,7 @@ import org.apache.hadoop.yarn.api.records.ApplicationResourceUsageReport;
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.NodeType;
 import org.apache.hadoop.yarn.util.resource.Resources;
 
 public class RMAppAttemptMetrics {
@@ -49,6 +50,10 @@ public class RMAppAttemptMetrics {
   private AtomicLong finishedVcoreSeconds = new AtomicLong(0);
   private RMContext rmContext;
 
+  private int[][] localityStatistics =
+      new int[NodeType.values().length][NodeType.values().length];
+  private volatile int totalAllocatedContainers;
+
   public RMAppAttemptMetrics(ApplicationAttemptId attemptId,
       RMContext rmContext) {
     this.attemptId = attemptId;
@@ -57,7 +62,7 @@ public class RMAppAttemptMetrics {
     this.writeLock = lock.writeLock();
     this.rmContext = rmContext;
   }
-  
+
   public void updatePreemptionInfo(Resource resource, RMContainer container) {
     try {
       writeLock.lock();
@@ -126,4 +131,18 @@ public class RMAppAttemptMetrics {
     this.finishedMemorySeconds.addAndGet(finishedMemorySeconds);
     this.finishedVcoreSeconds.addAndGet(finishedVcoreSeconds);
   }
+
+  public void incNumAllocatedContainers(NodeType containerType,
+      NodeType requestType) {
+    localityStatistics[containerType.index][requestType.index]++;
+    totalAllocatedContainers++;
+  }
+
+  public int[][] getLocalityStatistics() {
+    return this.localityStatistics;
+  }
+
+  public int getTotalAllocatedContainers() {
+    return this.totalAllocatedContainers;
+  }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/NodeType.java

@@ -22,7 +22,10 @@ package org.apache.hadoop.yarn.server.resourcemanager.scheduler;
  * Resource classification.
  */
 public enum NodeType {
-  NODE_LOCAL,
-  RACK_LOCAL,
-  OFF_SWITCH
+  NODE_LOCAL(0), RACK_LOCAL(1), OFF_SWITCH(2);
+  public int index;
+
+  private NodeType(int index) {
+    this.index = index;
+  }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java

@@ -46,6 +46,7 @@ import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.yarn.api.records.ResourceRequest;
 import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.AggregateAppResourceUsage;
+import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt;
 import org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptState;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerEvent;
@@ -78,7 +79,7 @@ public class SchedulerApplicationAttempt {
   private long lastVcoreSeconds = 0;
 
   protected final AppSchedulingInfo appSchedulingInfo;
-  
+  protected ApplicationAttemptId attemptId;
   protected Map<ContainerId, RMContainer> liveContainers =
       new HashMap<ContainerId, RMContainer>();
   protected final Map<Priority, Map<NodeId, RMContainer>> reservedContainers = 
@@ -132,6 +133,7 @@ public class SchedulerApplicationAttempt {
             activeUsersManager, rmContext.getEpoch());
     this.queue = queue;
     this.pendingRelease = new HashSet<ContainerId>();
+    this.attemptId = applicationAttemptId;
     if (rmContext.getRMApps() != null &&
         rmContext.getRMApps()
             .containsKey(applicationAttemptId.getApplicationId())) {
@@ -619,4 +621,15 @@ public class SchedulerApplicationAttempt {
     // schedulingOpportunities
     // lastScheduledContainer
   }
+
+  public void incNumAllocatedContainers(NodeType containerType,
+      NodeType requestType) {
+    RMAppAttempt attempt =
+        rmContext.getRMApps().get(attemptId.getApplicationId())
+          .getCurrentAppAttempt();
+    if (attempt != null) {
+      attempt.getRMAppAttemptMetrics().incNumAllocatedContainers(containerType,
+        requestType);
+    }
+  }
 }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java

@@ -32,6 +32,7 @@ import java.util.Set;
 import java.util.TreeSet;
 
 import org.apache.commons.lang.StringUtils;
+import org.apache.commons.lang.mutable.MutableObject;
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.classification.InterfaceAudience.Private;
@@ -1242,15 +1243,25 @@ public class LeafQueue extends AbstractCSQueue {
       RMContainer reservedContainer, boolean needToUnreserve) {
     Resource assigned = Resources.none();
 
+    NodeType requestType = null;
+    MutableObject allocatedContainer = new MutableObject();
     // Data-local
     ResourceRequest nodeLocalResourceRequest =
         application.getResourceRequest(priority, node.getNodeName());
     if (nodeLocalResourceRequest != null) {
-      assigned = 
-          assignNodeLocalContainers(clusterResource, nodeLocalResourceRequest,
-              node, application, priority, reservedContainer, needToUnreserve); 
-      if (Resources.greaterThan(resourceCalculator, clusterResource, 
+      requestType = NodeType.NODE_LOCAL;
+      assigned =
+          assignNodeLocalContainers(clusterResource, nodeLocalResourceRequest, 
+            node, application, priority, reservedContainer, needToUnreserve,
+            allocatedContainer);
+      if (Resources.greaterThan(resourceCalculator, clusterResource,
           assigned, Resources.none())) {
+
+        //update locality statistics
+        if (allocatedContainer.getValue() != null) {
+          application.incNumAllocatedContainers(NodeType.NODE_LOCAL,
+            requestType);
+        }
         return new CSAssignment(assigned, NodeType.NODE_LOCAL);
       }
     }
@@ -1262,12 +1273,23 @@ public class LeafQueue extends AbstractCSQueue {
       if (!rackLocalResourceRequest.getRelaxLocality()) {
         return SKIP_ASSIGNMENT;
       }
-      
-      assigned =
-          assignRackLocalContainers(clusterResource, rackLocalResourceRequest,
-              node, application, priority, reservedContainer, needToUnreserve);
-      if (Resources.greaterThan(resourceCalculator, clusterResource, 
+
+      if (requestType != NodeType.NODE_LOCAL) {
+        requestType = NodeType.RACK_LOCAL;
+      }
+
+      assigned = 
+          assignRackLocalContainers(clusterResource, rackLocalResourceRequest, 
+            node, application, priority, reservedContainer, needToUnreserve,
+            allocatedContainer);
+      if (Resources.greaterThan(resourceCalculator, clusterResource,
           assigned, Resources.none())) {
+
+        //update locality statistics
+        if (allocatedContainer.getValue() != null) {
+          application.incNumAllocatedContainers(NodeType.RACK_LOCAL,
+            requestType);
+        }
         return new CSAssignment(assigned, NodeType.RACK_LOCAL);
       }
     }
@@ -1279,11 +1301,21 @@ public class LeafQueue extends AbstractCSQueue {
       if (!offSwitchResourceRequest.getRelaxLocality()) {
         return SKIP_ASSIGNMENT;
       }
+      if (requestType != NodeType.NODE_LOCAL
+          && requestType != NodeType.RACK_LOCAL) {
+        requestType = NodeType.OFF_SWITCH;
+      }
+
+      assigned =
+          assignOffSwitchContainers(clusterResource, offSwitchResourceRequest,
+            node, application, priority, reservedContainer, needToUnreserve,
+            allocatedContainer);
 
-      return new CSAssignment(assignOffSwitchContainers(clusterResource,
-          offSwitchResourceRequest, node, application, priority,
-          reservedContainer, needToUnreserve),
-          NodeType.OFF_SWITCH);
+      // update locality statistics
+      if (allocatedContainer.getValue() != null) {
+        application.incNumAllocatedContainers(NodeType.OFF_SWITCH, requestType);
+      }
+      return new CSAssignment(assigned, NodeType.OFF_SWITCH);
     }
     
     return SKIP_ASSIGNMENT;
@@ -1370,40 +1402,43 @@ public class LeafQueue extends AbstractCSQueue {
   private Resource assignNodeLocalContainers(Resource clusterResource,
       ResourceRequest nodeLocalResourceRequest, FiCaSchedulerNode node,
       FiCaSchedulerApp application, Priority priority,
-      RMContainer reservedContainer, boolean needToUnreserve) {
-    if (canAssign(application, priority, node, NodeType.NODE_LOCAL,
+      RMContainer reservedContainer, boolean needToUnreserve,
+      MutableObject allocatedContainer) {
+    if (canAssign(application, priority, node, NodeType.NODE_LOCAL, 
         reservedContainer)) {
       return assignContainer(clusterResource, node, application, priority,
           nodeLocalResourceRequest, NodeType.NODE_LOCAL, reservedContainer,
-          needToUnreserve);
+          needToUnreserve, allocatedContainer);
     }
     
     return Resources.none();
   }
 
-  private Resource assignRackLocalContainers(Resource clusterResource,
-      ResourceRequest rackLocalResourceRequest, FiCaSchedulerNode node,
-      FiCaSchedulerApp application, Priority priority,
-      RMContainer reservedContainer, boolean needToUnreserve) {
-    if (canAssign(application, priority, node, NodeType.RACK_LOCAL, 
+  private Resource assignRackLocalContainers(
+      Resource clusterResource, ResourceRequest rackLocalResourceRequest,  
+      FiCaSchedulerNode node, FiCaSchedulerApp application, Priority priority,
+      RMContainer reservedContainer, boolean needToUnreserve,
+      MutableObject allocatedContainer) {
+    if (canAssign(application, priority, node, NodeType.RACK_LOCAL,
         reservedContainer)) {
       return assignContainer(clusterResource, node, application, priority,
           rackLocalResourceRequest, NodeType.RACK_LOCAL, reservedContainer,
-          needToUnreserve);
+          needToUnreserve, allocatedContainer);
     }
     
     return Resources.none();
   }
 
-  private Resource assignOffSwitchContainers(Resource clusterResource,
-      ResourceRequest offSwitchResourceRequest, FiCaSchedulerNode node,
-      FiCaSchedulerApp application, Priority priority,
-      RMContainer reservedContainer, boolean needToUnreserve) {
-    if (canAssign(application, priority, node, NodeType.OFF_SWITCH, 
+  private Resource assignOffSwitchContainers(
+      Resource clusterResource, ResourceRequest offSwitchResourceRequest,
+      FiCaSchedulerNode node, FiCaSchedulerApp application, Priority priority, 
+      RMContainer reservedContainer, boolean needToUnreserve,
+      MutableObject allocatedContainer) {
+    if (canAssign(application, priority, node, NodeType.OFF_SWITCH,
         reservedContainer)) {
       return assignContainer(clusterResource, node, application, priority,
           offSwitchResourceRequest, NodeType.OFF_SWITCH, reservedContainer,
-          needToUnreserve);
+          needToUnreserve, allocatedContainer);
     }
     
     return Resources.none();
@@ -1487,7 +1522,7 @@ public class LeafQueue extends AbstractCSQueue {
   private Resource assignContainer(Resource clusterResource, FiCaSchedulerNode node, 
       FiCaSchedulerApp application, Priority priority, 
       ResourceRequest request, NodeType type, RMContainer rmContainer,
-      boolean needToUnreserve) {
+      boolean needToUnreserve, MutableObject createdContainer) {
     if (LOG.isDebugEnabled()) {
       LOG.debug("assignContainers: node=" + node.getNodeName()
         + " application=" + application.getApplicationId()
@@ -1592,7 +1627,7 @@ public class LeafQueue extends AbstractCSQueue {
           " container=" + container + 
           " queue=" + this + 
           " clusterResource=" + clusterResource);
-
+      createdContainer.setValue(allocatedContainer);
       return container.getResource();
     } else {
       // if we are allowed to allocate but this node doesn't have space, reserve it or

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/AppBlock.java

@@ -204,18 +204,55 @@ public class AppBlock extends HtmlBlock {
     table._();
     div._();
 
+    createContainerLocalityTable(html, attemptMetrics);
     createResourceRequestsTable(html, app);
   }
 
+  private void createContainerLocalityTable(Block html,
+      RMAppAttemptMetrics attemptMetrics) {
+    if (attemptMetrics == null) {
+      return;
+    }
+
+    DIV<Hamlet> div = html.div(_INFO_WRAP);
+    TABLE<DIV<Hamlet>> table =
+        div.h3(
+          "Total Allocated Containers: "
+              + attemptMetrics.getTotalAllocatedContainers()).h3("Each table cell"
+            + " represents the number of NodeLocal/RackLocal/OffSwitch containers"
+            + " satisfied by NodeLocal/RackLocal/OffSwitch resource requests.").table(
+          "#containerLocality");
+    table.
+      tr().
+        th(_TH, "").
+        th(_TH, "Node Local Request").
+        th(_TH, "Rack Local Request").
+        th(_TH, "Off Switch Request").
+      _();
+
+    String[] containersType =
+        { "Num Node Local Containers (satisfied by)", "Num Rack Local Containers (satisfied by)",
+            "Num Off Switch Containers (satisfied by)" };
+    boolean odd = false;
+    for (int i = 0; i < attemptMetrics.getLocalityStatistics().length; i++) {
+      table.tr((odd = !odd) ? _ODD : _EVEN).td(containersType[i])
+        .td(String.valueOf(attemptMetrics.getLocalityStatistics()[i][0]))
+        .td(i == 0 ? "" : String.valueOf(attemptMetrics.getLocalityStatistics()[i][1]))
+        .td(i <= 1 ? "" : String.valueOf(attemptMetrics.getLocalityStatistics()[i][2]))._();
+    }
+    table._();
+    div._();
+  }
+
   private void createResourceRequestsTable(Block html, AppInfo app) {
     TBODY<TABLE<Hamlet>> tbody =
         html.table("#ResourceRequests").thead().tr()
           .th(".priority", "Priority")
-          .th(".resourceName", "ResourceName")
+          .th(".resourceName", "Resource Name")
           .th(".totalResource", "Capability")
-          .th(".numContainers", "NumContainers")
-          .th(".relaxLocality", "RelaxLocality")
-          .th(".nodeLabelExpression", "NodeLabelExpression")._()._().tbody();
+          .th(".numContainers", "Num Containers")
+          .th(".relaxLocality", "Relax Locality")
+          .th(".nodeLabelExpression", "Node Label Expression")._()._().tbody();
 
     Resource totalResource = Resource.newInstance(0, 0);
     if (app.getResourceRequests() != null) {

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/TestReservations.java

@@ -52,6 +52,7 @@ import org.apache.hadoop.yarn.nodelabels.CommonNodeLabelsManager;
 import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
 import org.apache.hadoop.yarn.server.resourcemanager.ahs.RMApplicationHistoryWriter;
 import org.apache.hadoop.yarn.server.resourcemanager.metrics.SystemMetricsPublisher;
+import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.ContainerAllocationExpirer;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
 import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerEventType;
@@ -218,6 +219,7 @@ public class TestReservations {
         .getMockApplicationAttemptId(0, 0);
     FiCaSchedulerApp app_0 = new FiCaSchedulerApp(appAttemptId_0, user_0, a,
         mock(ActiveUsersManager.class), spyRMContext);
+    rmContext.getRMApps().put(app_0.getApplicationId(), mock(RMApp.class));
 
     a.submitApplicationAttempt(app_0, user_0); 
 
@@ -373,6 +375,7 @@ public class TestReservations {
         .getMockApplicationAttemptId(0, 0);
     FiCaSchedulerApp app_0 = new FiCaSchedulerApp(appAttemptId_0, user_0, a,
         mock(ActiveUsersManager.class), spyRMContext);
+    rmContext.getRMApps().put(app_0.getApplicationId(), mock(RMApp.class));
 
     a.submitApplicationAttempt(app_0, user_0); 
 
@@ -524,6 +527,7 @@ public class TestReservations {
         .getMockApplicationAttemptId(0, 0);
     FiCaSchedulerApp app_0 = new FiCaSchedulerApp(appAttemptId_0, user_0, a,
         mock(ActiveUsersManager.class), spyRMContext);
+    rmContext.getRMApps().put(app_0.getApplicationId(), mock(RMApp.class));
 
     a.submitApplicationAttempt(app_0, user_0); 
 
@@ -765,6 +769,7 @@ public class TestReservations {
         .getMockApplicationAttemptId(0, 0);
     FiCaSchedulerApp app_0 = new FiCaSchedulerApp(appAttemptId_0, user_0, a,
         mock(ActiveUsersManager.class), spyRMContext);
+    rmContext.getRMApps().put(app_0.getApplicationId(), mock(RMApp.class));
 
     a.submitApplicationAttempt(app_0, user_0); 
 
@@ -943,7 +948,7 @@ public class TestReservations {
         .getMockApplicationAttemptId(0, 0);
     FiCaSchedulerApp app_0 = new FiCaSchedulerApp(appAttemptId_0, user_0, a,
         mock(ActiveUsersManager.class), spyRMContext);
-
+    rmContext.getRMApps().put(app_0.getApplicationId(), mock(RMApp.class));
     a.submitApplicationAttempt(app_0, user_0); 
 
     final ApplicationAttemptId appAttemptId_1 = TestUtils
@@ -1073,6 +1078,7 @@ public class TestReservations {
         .getMockApplicationAttemptId(0, 0);
     FiCaSchedulerApp app_0 = new FiCaSchedulerApp(appAttemptId_0, user_0, a,
         mock(ActiveUsersManager.class), spyRMContext);
+    rmContext.getRMApps().put(app_0.getApplicationId(), mock(RMApp.class));
 
     a.submitApplicationAttempt(app_0, user_0);