From 31fff48ec4a7bf04a2078921316c58917c96f610 Mon Sep 17 00:00:00 2001
From: Chris Nauroth <cnauroth@apache.org>
Date: Mon, 8 Jul 2013 23:40:43 +0000
Subject: [PATCH] YARN-894. NodeHealthScriptRunner timeout checking is
 inaccurate on Windows. Contributed by Chuan Liu.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1501016 13f79535-47bb-0310-9956-ffa450edef68
---
 hadoop-yarn-project/CHANGES.txt                             | 3 +++
 .../yarn/server/nodemanager/NodeHealthScriptRunner.java     | 6 ++++++
 .../yarn/server/nodemanager/TestNodeHealthService.java      | 6 ++++--
 3 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/hadoop-yarn-project/CHANGES.txt b/hadoop-yarn-project/CHANGES.txt
index 266b32c989c..69f367f2f6d 100644
--- a/hadoop-yarn-project/CHANGES.txt
+++ b/hadoop-yarn-project/CHANGES.txt
@@ -727,6 +727,9 @@ Release 2.1.0-beta - 2013-07-02
     YARN-852. TestAggregatedLogFormat.testContainerLogsFileAccess fails on
     Windows. (Chuan Liu via cnauroth)
 
+    YARN-894. NodeHealthScriptRunner timeout checking is inaccurate on Windows.
+    (Chuan Liu via cnauroth)
+
     YARN-795. Fair scheduler queue metrics should subtract allocated vCores from 
     available vCores. (ywskycn via tucu) 
 
diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/NodeHealthScriptRunner.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/NodeHealthScriptRunner.java
index 2461d1ccf4f..e3c9862acad 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/NodeHealthScriptRunner.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/NodeHealthScriptRunner.java
@@ -32,6 +32,7 @@
 import org.apache.hadoop.service.AbstractService;
 import org.apache.hadoop.util.Shell.ExitCodeException;
 import org.apache.hadoop.util.Shell.ShellCommandExecutor;
+import org.apache.hadoop.util.Shell;
 import org.apache.hadoop.util.StringUtils;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 
@@ -110,6 +111,11 @@ public void run() {
       } catch (ExitCodeException e) {
         // ignore the exit code of the script
         status = HealthCheckerExitStatus.FAILED_WITH_EXIT_CODE;
+        // On Windows, we will not hit the Stream closed IOException
+        // thrown by stdout buffered reader for timeout event.
+        if (Shell.WINDOWS && shexec.isTimedOut()) {
+          status = HealthCheckerExitStatus.TIMED_OUT;
+        }
       } catch (Exception e) {
         LOG.warn("Caught exception : " + e.getMessage());
         if (!shexec.isTimedOut()) {
diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/TestNodeHealthService.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/TestNodeHealthService.java
index 0c9b595e493..8103c1643ab 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/TestNodeHealthService.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/TestNodeHealthService.java
@@ -30,6 +30,7 @@
 import org.apache.hadoop.fs.FileContext;
 import org.apache.hadoop.fs.FileUtil;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.util.Shell;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.factories.RecordFactory;
 import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
@@ -51,7 +52,7 @@ public class TestNodeHealthService {
       "modified-mapred-site.xml");
 
   private File nodeHealthscriptFile = new File(testRootDir,
-      "failingscript.sh");
+      Shell.appendScriptExtension("failingscript"));
 
   @Before
   public void setup() {
@@ -123,7 +124,8 @@ public void testNodeHealthScript() throws Exception {
         factory.newRecordInstance(NodeHealthStatus.class);
     String errorScript = "echo ERROR\n echo \"Tracker not healthy\"";
     String normalScript = "echo \"I am all fine\"";
-    String timeOutScript = "sleep 4\n echo\"I am fine\"";
+    String timeOutScript = Shell.WINDOWS ? "@echo off\nping -n 4 127.0.0.1 >nul\necho \"I am fine\""
+        : "sleep 4\necho \"I am fine\"";
     Configuration conf = getConfForNodeHealthScript();
     conf.writeXml(new FileOutputStream(nodeHealthConfigFile));
     conf.addResource(nodeHealthConfigFile.getName());