MAPREDUCE-5773. Provide dedicated MRAppMaster syslog length limit. Contributed by Gera Shegalov

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1573775 13f79535-47bb-0310-9956-ffa450edef68

hadoop-mapreduce-project/CHANGES.txt

@@ -185,6 +185,9 @@ Release 2.4.0 - UNRELEASED
     MAPREDUCE-5766. Moved ping messages from TaskAttempts to be at DEBUG level
     inside the ApplicationMaster log. (Jian He via vinodkv)
 
+    MAPREDUCE-5773. Provide dedicated MRAppMaster syslog length limit (Gera
+    Shegalov via jlowe)
+
   OPTIMIZATIONS
 
   BUG FIXES

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

@@ -414,9 +414,13 @@ public interface MRJobConfig {
     MR_AM_PREFIX+"log.level";
   public static final String DEFAULT_MR_AM_LOG_LEVEL = "INFO";
 
+  public static final String MR_AM_LOG_KB =
+      MR_AM_PREFIX + "container.log.limit.kb";
+  public static final int DEFAULT_MR_AM_LOG_KB = 0; // don't roll
+
   public static final String MR_AM_LOG_BACKUPS =
       MR_AM_PREFIX + "container.log.backups";
-  public static final int DEFAULT_MR_AM_LOG_BACKUPS = 0; // don't roll
+  public static final int DEFAULT_MR_AM_LOG_BACKUPS = 0;
 
   /**The number of splits when reporting progress in MR*/
   public static final String MR_AM_NUM_PROGRESS_SPLITS = 

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

@@ -510,6 +510,14 @@
   </description>
 </property>
 
+<property>
+  <name>yarn.app.mapreduce.am.container.log.limit.kb</name>
+  <value>0</value>
+  <description>The maximum size of the MRAppMaster attempt container logs in KB.
+    0 disables the cap.
+  </description>
+</property>
+
 <property>
   <name>yarn.app.mapreduce.task.container.log.backups</name>
   <value>0</value>

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-jobclient/src/main/java/org/apache/hadoop/mapred/YARNRunner.java

@@ -391,7 +391,8 @@ public class YARNRunner implements ClientProtocol {
     vargs.add(Environment.JAVA_HOME.$() + "/bin/java");
 
     // TODO: why do we use 'conf' some places and 'jobConf' others?
-    long logSize = TaskLog.getTaskLogLength(new JobConf(conf));
+    long logSize = jobConf.getLong(MRJobConfig.MR_AM_LOG_KB,
+        MRJobConfig.DEFAULT_MR_AM_LOG_KB) << 10;
     String logLevel = jobConf.get(
         MRJobConfig.MR_AM_LOG_LEVEL, MRJobConfig.DEFAULT_MR_AM_LOG_LEVEL);
     int numBackups = jobConf.getInt(MRJobConfig.MR_AM_LOG_BACKUPS,

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-jobclient/src/test/java/org/apache/hadoop/mapreduce/v2/TestMRJobs.java

@@ -27,6 +27,7 @@ import java.io.IOException;
 import java.io.StringReader;
 import java.net.URI;
 import java.security.PrivilegedExceptionAction;
+import java.util.Arrays;
 import java.util.EnumSet;
 import java.util.HashMap;
 import java.util.Map;
@@ -442,9 +443,12 @@ public class TestMRJobs {
     final SleepJob sleepJob = new SleepJob();
     final JobConf sleepConf = new JobConf(mrCluster.getConfig());
     sleepConf.set(MRJobConfig.MAP_LOG_LEVEL, Level.ALL.toString());
-    sleepConf.set(MRJobConfig.MR_AM_LOG_LEVEL, Level.ALL.toString());
+    final long userLogKb = 4;
-    sleepConf.setLong(MRJobConfig.TASK_USERLOG_LIMIT, 1);
+    sleepConf.setLong(MRJobConfig.TASK_USERLOG_LIMIT, userLogKb);
     sleepConf.setInt(MRJobConfig.TASK_LOG_BACKUPS, 3);
+    sleepConf.set(MRJobConfig.MR_AM_LOG_LEVEL, Level.ALL.toString());
+    final long amLogKb = 7;
+    sleepConf.setLong(MRJobConfig.MR_AM_LOG_KB, amLogKb);
     sleepConf.setInt(MRJobConfig.MR_AM_LOG_BACKUPS, 7);
     sleepJob.setConf(sleepConf);
 
@@ -503,6 +507,8 @@ public class TestMRJobs {
 
           final FileStatus[] sysSiblings = localFs.globStatus(new Path(
               containerPathComponent, TaskLog.LogName.SYSLOG + "*"));
+          // sort to ensure for i > 0 sysSiblings[i] == "syslog.i"
+          Arrays.sort(sysSiblings);
 
           if (foundAppMaster) {
             numAppMasters++;
@@ -510,11 +516,19 @@ public class TestMRJobs {
             numMapTasks++;
           }
 
-          Assert.assertSame("Number of sylog* files",
+          if (foundAppMaster) {
-              foundAppMaster
+            Assert.assertSame("Unexpected number of AM sylog* files",
-                ? sleepConf.getInt(MRJobConfig.MR_AM_LOG_BACKUPS, 0) + 1
+                sleepConf.getInt(MRJobConfig.MR_AM_LOG_BACKUPS, 0) + 1,
-                : sleepConf.getInt(MRJobConfig.TASK_LOG_BACKUPS, 0) + 1,
                 sysSiblings.length);
+            Assert.assertTrue("AM syslog.1 length kb should be >= " + amLogKb,
+                sysSiblings[1].getLen() >= amLogKb * 1024);
+          } else {
+            Assert.assertSame("Unexpected number of MR task sylog* files",
+                sleepConf.getInt(MRJobConfig.TASK_LOG_BACKUPS, 0) + 1,
+                sysSiblings.length);
+            Assert.assertTrue("MR syslog.1 length kb should be >= " + userLogKb,
+                sysSiblings[1].getLen() >= userLogKb * 1024);
+          }
         }
       }
     }