From 16ef8dd3a5e782ee9546ba5a7232f55fc3fe5c0c Mon Sep 17 00:00:00 2001
From: Suresh Srinivas <suresh@apache.org>
Date: Thu, 25 Apr 2013 04:25:58 +0000
Subject: [PATCH] HDFS-4745. TestDataTransferKeepalive#testSlowReader has race
 condition that causes sporadic failure. Contributed by Chris Nauroth.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1475623 13f79535-47bb-0310-9956-ffa450edef68
---
 hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt   |  3 +++
 .../hdfs/TestDataTransferKeepalive.java       | 24 +++++++++++++++----
 2 files changed, 23 insertions(+), 4 deletions(-)

diff --git a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
index 5bc0b87ecbc..a6de677ed73 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
+++ b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
@@ -588,6 +588,9 @@ Release 2.0.5-beta - UNRELEASED
     HDFS-4739. NN can miscalculate the number of extra edit log segments to
     retain. (atm)
 
+    HDFS-4745. TestDataTransferKeepalive#testSlowReader has race condition that
+    causes sporadic failure. (Chris Nauroth via suresh)
+
 Release 2.0.4-alpha - UNRELEASED
 
   INCOMPATIBLE CHANGES
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDataTransferKeepalive.java b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDataTransferKeepalive.java
index 3c9ee25b111..bf4e13bd027 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDataTransferKeepalive.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDataTransferKeepalive.java
@@ -146,7 +146,15 @@ public class TestDataTransferKeepalive {
       stm.read();
       assertXceiverCount(1);
 
-      Thread.sleep(WRITE_TIMEOUT + 1000);
+      // Poll for 0 running xceivers.  Allow up to 5 seconds for some slack.
+      long totalSleepTime = 0;
+      long sleepTime = WRITE_TIMEOUT + 100;
+      while (getXceiverCountWithoutServer() > 0 && totalSleepTime < 5000) {
+        Thread.sleep(sleepTime);
+        totalSleepTime += sleepTime;
+        sleepTime = 100;
+      }
+
       // DN should time out in sendChunks, and this should force
       // the xceiver to exit.
       assertXceiverCount(0);
@@ -190,9 +198,7 @@ public class TestDataTransferKeepalive {
   }
 
   private void assertXceiverCount(int expected) {
-    // Subtract 1, since the DataXceiverServer
-    // counts as one
-    int count = dn.getXceiverCount() - 1;
+    int count = getXceiverCountWithoutServer();
     if (count != expected) {
       ReflectionUtils.printThreadInfo(
           new PrintWriter(System.err),
@@ -201,4 +207,14 @@ public class TestDataTransferKeepalive {
           count);
     }
   }
+
+  /**
+   * Returns the datanode's xceiver count, but subtracts 1, since the
+   * DataXceiverServer counts as one.
+   * 
+   * @return int xceiver count, not including DataXceiverServer
+   */
+  private int getXceiverCountWithoutServer() {
+    return dn.getXceiverCount() - 1;
+  }
 }