MAPREDUCE-6776. yarn.app.mapreduce.client.job.max-retries should have a more useful default (miklos.szegedi@cloudera.com via rkanter)

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/MRJobConfig.java

@@ -505,7 +505,7 @@ public interface MRJobConfig {
    */
   public static final String MR_CLIENT_JOB_MAX_RETRIES =
       MR_PREFIX + "client.job.max-retries";
-  public static final int DEFAULT_MR_CLIENT_JOB_MAX_RETRIES = 0;
+  public static final int DEFAULT_MR_CLIENT_JOB_MAX_RETRIES = 3;
 
   /**
    * How long to wait between jobclient retries on failure

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/resources/mapred-default.xml

@@ -1505,12 +1505,12 @@
 
 <property>
   <name>yarn.app.mapreduce.client.job.max-retries</name>
-  <value>0</value>
+  <value>3</value>
   <description>The number of retries the client will make for getJob and
-  dependent calls.  The default is 0 as this is generally only needed for
-  non-HDFS DFS where additional, high level retries are required to avoid
-  spurious failures during the getJob call.  30 is a good value for
-  WASB</description>
+    dependent calls.
+    This is needed for non-HDFS DFS where additional, high level
+    retries are required to avoid spurious failures during the getJob call.
+    30 is a good value for WASB</description>
 </property>
 
 <property>

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-jobclient/src/test/java/org/apache/hadoop/mapred/JobClientUnitTest.java

@@ -225,10 +225,10 @@ public class JobClientUnitTest {
 
     //To prevent the test from running for a very long time, lower the retry
     JobConf conf = new JobConf();
-    conf.set(MRJobConfig.MR_CLIENT_JOB_MAX_RETRIES, "3");
+    conf.setInt(MRJobConfig.MR_CLIENT_JOB_MAX_RETRIES, 2);
 
     TestJobClientGetJob client = new TestJobClientGetJob(conf);
-    JobID id = new JobID("ajob",1);
+    JobID id = new JobID("ajob", 1);
     RunningJob rj = mock(RunningJob.class);
     client.setRunningJob(rj);
 
@@ -236,13 +236,35 @@ public class JobClientUnitTest {
     assertNotNull(client.getJob(id));
     assertEquals(client.getLastGetJobRetriesCounter(), 0);
 
-    //3 retry
+    //2 retries
+    client.setGetJobRetries(2);
+    assertNotNull(client.getJob(id));
+    assertEquals(client.getLastGetJobRetriesCounter(), 2);
+
+    //beyond yarn.app.mapreduce.client.job.max-retries, will get null
     client.setGetJobRetries(3);
+    assertNull(client.getJob(id));
+  }
+
+  @Test
+  public void testGetJobRetryDefault() throws Exception {
+
+    //To prevent the test from running for a very long time, lower the retry
+    JobConf conf = new JobConf();
+
+    TestJobClientGetJob client = new TestJobClientGetJob(conf);
+    JobID id = new JobID("ajob", 1);
+    RunningJob rj = mock(RunningJob.class);
+    client.setRunningJob(rj);
+
+    //3 retries (default)
+    client.setGetJobRetries(MRJobConfig.DEFAULT_MR_CLIENT_JOB_MAX_RETRIES);
     assertNotNull(client.getJob(id));
-    assertEquals(client.getLastGetJobRetriesCounter(), 3);
+    assertEquals(client.getLastGetJobRetriesCounter(),
+        MRJobConfig.DEFAULT_MR_CLIENT_JOB_MAX_RETRIES);
 
-    //beyond MAPREDUCE_JOBCLIENT_GETJOB_MAX_RETRY_KEY, will get null
-    client.setGetJobRetries(5);
+    //beyond yarn.app.mapreduce.client.job.max-retries, will get null
+    client.setGetJobRetries(MRJobConfig.DEFAULT_MR_CLIENT_JOB_MAX_RETRIES + 1);
     assertNull(client.getJob(id));
   }