YARN-3790. usedResource from rootQueue metrics may get stale data for FS scheduler after recovering the container (Zhihai Xu via rohithsharmaks)

hadoop-yarn-project/CHANGES.txt

@@ -705,6 +705,9 @@ Release 2.7.1 - UNRELEASED
     YARN-3832. Resource Localization fails on a cluster due to existing cache
     directories (Brahma Reddy Battula via jlowe)
 
+    YARN-3790. usedResource from rootQueue metrics may get stale data for FS
+    scheduler after recovering the container (Zhihai Xu via rohithsharmaks)
+
 Release 2.7.0 - 2015-04-20
 
   INCOMPATIBLE CHANGES

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

@@ -45,6 +45,7 @@ import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.exceptions.YarnException;
 import org.apache.hadoop.yarn.exceptions.YarnRuntimeException;
 import org.apache.hadoop.yarn.proto.YarnServiceProtos.SchedulerResourceTypes;
+import org.apache.hadoop.yarn.server.api.protocolrecords.NMContainerStatus;
 import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
 import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore.RMState;
 import org.apache.hadoop.yarn.server.resourcemanager.reservation.ReservationConstants;
@@ -841,11 +842,11 @@ public class FairScheduler extends
         + " with event: " + event);
   }
 
-  private synchronized void addNode(RMNode node) {
+  private synchronized void addNode(List<NMContainerStatus> containerReports,
+      RMNode node) {
     FSSchedulerNode schedulerNode = new FSSchedulerNode(node, usePortForNodeName);
     nodes.put(node.getNodeID(), schedulerNode);
     Resources.addTo(clusterResource, node.getTotalCapability());
-    updateRootQueueMetrics();
     updateMaximumAllocation(schedulerNode, true);
 
     triggerUpdate();
@@ -854,6 +855,9 @@ public class FairScheduler extends
     queueMgr.getRootQueue().recomputeSteadyShares();
     LOG.info("Added node " + node.getNodeAddress() +
         " cluster capacity: " + clusterResource);
+
+    recoverContainersOnNode(containerReports, node);
+    updateRootQueueMetrics();
   }
 
   private synchronized void removeNode(RMNode rmNode) {
@@ -1147,8 +1151,7 @@ public class FairScheduler extends
         throw new RuntimeException("Unexpected event type: " + event);
       }
       NodeAddedSchedulerEvent nodeAddedEvent = (NodeAddedSchedulerEvent)event;
-      addNode(nodeAddedEvent.getAddedRMNode());
-      recoverContainersOnNode(nodeAddedEvent.getContainerReports(),
+      addNode(nodeAddedEvent.getContainerReports(),
           nodeAddedEvent.getAddedRMNode());
       break;
     case NODE_REMOVED: