YARN-4227. Ignore expired containers from removed nodes in FairScheduler. (Wilfred Spiegelenburg via rchiang)

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

@@ -669,27 +669,36 @@ public class FairScheduler extends
       ApplicationId appId =
           container.getId().getApplicationAttemptId().getApplicationId();
       if (application == null) {
-        LOG.info(
-            "Container " + container + " of" + " finished application " + appId
-                + " completed with event " + event);
+        LOG.info("Container " + container + " of finished application " +
+            appId + " completed with event " + event);
         return;
       }
 
       // Get the node on which the container was allocated
-      FSSchedulerNode node = getFSSchedulerNode(container.getNodeId());
-
+      NodeId nodeID = container.getNodeId();
+      FSSchedulerNode node = getFSSchedulerNode(nodeID);
+      // node could be null if the thread was waiting for the lock and the node
+      // was removed in another thread
       if (rmContainer.getState() == RMContainerState.RESERVED) {
-        application.unreserve(rmContainer.getReservedSchedulerKey(), node);
-      } else{
+        if (node != null) {
+          application.unreserve(rmContainer.getReservedSchedulerKey(), node);
+        } else if (LOG.isDebugEnabled()) {
+          LOG.debug("Skipping unreserve on removed node: " + nodeID);
+        }
+      } else {
         application.containerCompleted(rmContainer, containerStatus, event);
-        node.releaseContainer(rmContainer.getContainerId(), false);
+        if (node != null) {
+          node.releaseContainer(rmContainer.getContainerId(), false);
+        } else if (LOG.isDebugEnabled()) {
+          LOG.debug("Skipping container release on removed node: " + nodeID);
+        }
         updateRootQueueMetrics();
       }
 
       if (LOG.isDebugEnabled()) {
         LOG.debug("Application attempt " + application.getApplicationAttemptId()
-            + " released container " + container.getId() + " on node: " + node
-            + " with event: " + event);
+            + " released container " + container.getId() + " on node: " +
+            (node == null ? nodeID : node) + " with event: " + event);
       }
     } finally {
       writeLock.unlock();

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/TestFairScheduler.java

@@ -58,6 +58,7 @@ import org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext;
 import org.apache.hadoop.yarn.api.records.ContainerId;
 import org.apache.hadoop.yarn.api.records.ContainerLaunchContext;
 import org.apache.hadoop.yarn.api.records.FinalApplicationStatus;
+import org.apache.hadoop.yarn.api.records.NodeId;
 import org.apache.hadoop.yarn.api.records.NodeState;
 import org.apache.hadoop.yarn.api.records.QueueInfo;
 import org.apache.hadoop.yarn.api.records.Resource;
@@ -5355,4 +5356,62 @@ public class TestFairScheduler extends FairSchedulerTestBase {
     assertEquals("Unexpected state dump string",
         parentQueueString + ", " + childQueueString, parent.dumpState());
   }
+
+  @Test
+  public void testCompletedContainerOnRemovedNode() throws IOException {
+    scheduler.init(conf);
+    scheduler.start();
+    scheduler.reinitialize(conf, resourceManager.getRMContext());
+
+    // Add a node
+    RMNode node = MockNodes.newNodeInfo(1, Resources.createResource(2048), 2,
+        "127.0.0.2");
+    scheduler.handle(new NodeAddedSchedulerEvent(node));
+
+    // Create application attempt
+    ApplicationAttemptId appAttemptId = createAppAttemptId(1, 1);
+    createMockRMApp(appAttemptId);
+    scheduler.addApplication(appAttemptId.getApplicationId(), "root.queue1",
+        "user1", false);
+    scheduler.addApplicationAttempt(appAttemptId, false, false);
+
+    // Create container request that goes to a specific node.
+    // Without the 2nd and 3rd request we do not get live containers
+    List<ResourceRequest> ask1 = new ArrayList<>();
+    ResourceRequest request1 =
+        createResourceRequest(1024, node.getHostName(), 1, 1, true);
+    ask1.add(request1);
+    ResourceRequest request2 =
+        createResourceRequest(1024, node.getRackName(), 1, 1, false);
+    ask1.add(request2);
+    ResourceRequest request3 =
+        createResourceRequest(1024, ResourceRequest.ANY, 1, 1, false);
+    ask1.add(request3);
+
+    // Perform allocation
+    scheduler.allocate(appAttemptId, ask1, new ArrayList<ContainerId>(), null,
+        null, NULL_UPDATE_REQUESTS);
+    scheduler.update();
+    scheduler.handle(new NodeUpdateSchedulerEvent(node));
+
+    // Get the allocated containers for the application (list can not be null)
+    Collection<RMContainer> clist = scheduler.getSchedulerApp(appAttemptId)
+        .getLiveContainers();
+    Assert.assertEquals(1, clist.size());
+
+    // Make sure that we remove the correct node (should never fail)
+    RMContainer rmc = clist.iterator().next();
+    NodeId containerNodeID = rmc.getAllocatedNode();
+    assertEquals(node.getNodeID(), containerNodeID);
+
+    // Remove node
+    scheduler.handle(new NodeRemovedSchedulerEvent(node));
+
+    // Call completedContainer() should not fail even if the node has been
+    // removed
+    scheduler.completedContainer(rmc,
+        SchedulerUtils.createAbnormalContainerStatus(rmc.getContainerId(),
+            SchedulerUtils.COMPLETED_APPLICATION),
+        RMContainerEventType.EXPIRE);
+  }
 }