From 7dd378c274bbbbc11f835a99fc5205f5c48c8faa Mon Sep 17 00:00:00 2001
From: Karthik Kambatla <kasha@apache.org>
Date: Fri, 23 May 2014 23:51:00 +0000
Subject: [PATCH] YARN-2096. Race in TestRMRestart#testQueueMetricsOnRMRestart.
 (Anubhav Dhoot via kasha)

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1597223 13f79535-47bb-0310-9956-ffa450edef68
---
 hadoop-yarn-project/CHANGES.txt                   |  3 +++
 .../server/resourcemanager/TestRMRestart.java     | 15 +++++++++++----
 2 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/hadoop-yarn-project/CHANGES.txt b/hadoop-yarn-project/CHANGES.txt
index ae2b1bab02..c736f0ece0 100644
--- a/hadoop-yarn-project/CHANGES.txt
+++ b/hadoop-yarn-project/CHANGES.txt
@@ -143,6 +143,9 @@ Release 2.5.0 - UNRELEASED
     YARN-2089. FairScheduler: QueuePlacementPolicy and QueuePlacementRule 
     are missing audience annotations. (Zhihai Xu via kasha)
 
+    YARN-2096. Race in TestRMRestart#testQueueMetricsOnRMRestart.
+    (Anubhav Dhoot via kasha)
+
 Release 2.4.1 - UNRELEASED
 
   INCOMPATIBLE CHANGES
diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java
index 49eff8b014..6a0512307a 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java
@@ -1638,10 +1638,16 @@ public void testQueueMetricsOnRMRestart() throws Exception {
 
     // create app that gets launched and does allocate before RM restart
     RMApp app1 = rm1.submitApp(200);
-    assertQueueMetrics(qm1, 1, 1, 0, 0);
-    nm1.nodeHeartbeat(true);
+    // Need to wait first for AppAttempt to be started (RMAppState.ACCEPTED)
+    // and then for it to reach RMAppAttemptState.SCHEDULED
+    // inorder to ensure appsPending metric is incremented
+    rm1.waitForState(app1.getApplicationId(), RMAppState.ACCEPTED);
     RMAppAttempt attempt1 = app1.getCurrentAppAttempt();
     ApplicationAttemptId attemptId1 = attempt1.getAppAttemptId();
+    rm1.waitForState(attemptId1, RMAppAttemptState.SCHEDULED);
+    assertQueueMetrics(qm1, 1, 1, 0, 0);
+
+    nm1.nodeHeartbeat(true);
     rm1.waitForState(attemptId1, RMAppAttemptState.ALLOCATED);
     MockAM am1 = rm1.sendAMLaunched(attempt1.getAppAttemptId());
     am1.registerAppAttempt();
@@ -1660,11 +1666,12 @@ public void testQueueMetricsOnRMRestart() throws Exception {
     // PHASE 2: create new RM and start from old state
     // create new RM to represent restart and recover state
     MockRM rm2 = new MockRM(conf, memStore);
-    rm2.start();
-    nm1.setResourceTrackerService(rm2.getResourceTrackerService());
     QueueMetrics qm2 = rm2.getResourceScheduler().getRootQueueMetrics();
     resetQueueMetrics(qm2);
     assertQueueMetrics(qm2, 0, 0, 0, 0);
+
+    rm2.start();
+    nm1.setResourceTrackerService(rm2.getResourceTrackerService());
     // recover app
     RMApp loadedApp1 = rm2.getRMContext().getRMApps().get(app1.getApplicationId());
     am1.setAMRMProtocol(rm2.getApplicationMasterService());