hadoop/hadoop-mapreduce/INSTALL

To compile  Hadoop Mapreduce next following, do the following:

Step 1) Install dependencies for yarn

See http://svn.apache.org/repos/asf/hadoop/common/branches/MR-279/mapreduce/yarn/README
Make sure protbuf library is in your library path or set: export LD_LIBRARY_PATH=/usr/local/lib

Step 2) Checkout

svn checkout http://svn.apache.org/repos/asf/hadoop/common/branches/MR-279/

Step 3) Build common

Go to common directory
ant veryclean mvn-install 

Step 4) Build HDFS 

Go to hdfs directory
ant veryclean mvn-install -Dresolvers=internal 

Step 5) Build yarn and mapreduce

Go to mapreduce directory
export MAVEN_OPTS=-Xmx512m

mvn clean install assembly:assembly
ant veryclean jar jar-test  -Dresolvers=internal 

In case you want to skip the tests run:

mvn clean install assembly:assembly -DskipTests
ant veryclean jar jar-test  -Dresolvers=internal 

You will see a tarball in
ls target/hadoop-mapreduce-1.0-SNAPSHOT-bin.tar.gz  

Step 6) Untar the tarball in a clean and different directory.
say HADOOP_YARN_INSTALL

To run Hadoop Mapreduce next applications :

Step 7) cd $HADOOP_YARN_INSTALL

Step 8) export the following variables:

HADOOP_MAPRED_HOME=
HADOOP_COMMON_HOME=
HADOOP_HDFS_HOME=
YARN_HOME=directory where you untarred yarn
HADOOP_CONF_DIR=
YARN_CONF_DIR=$HADOOP_CONF_DIR

Step 9) bin/yarn-daemon.sh start resourcemanager

Step 10) bin/yarn-daemon.sh start nodemanager

Step 11) bin/yarn-daemon.sh start historyserver

Step 12) Create the following symlinks in hadoop-common/lib 

ln -s $HADOOP_YARN_INSTALL/modules/hadoop-mapreduce-client-app-1.0-SNAPSHOT.jar .	
ln -s $HADOOP_YARN_INSTALL/modules/yarn-api-1.0-SNAPSHOT.jar .
ln -s $HADOOP_YARN_INSTALL/modules/hadoop-mapreduce-client-common-1.0-SNAPSHOT.jar .	
ln -s $HADOOP_YARN_INSTALL/modules/yarn-common-1.0-SNAPSHOT.jar .
ln -s $HADOOP_YARN_INSTALL/modules/hadoop-mapreduce-client-core-1.0-SNAPSHOT.jar .	
ln -s $HADOOP_YARN_INSTALL/modules/yarn-server-common-1.0-SNAPSHOT.jar .
ln -s $HADOOP_YARN_INSTALL/modules/hadoop-mapreduce-client-jobclient-1.0-SNAPSHOT.jar .
ln -s $HADOOP_YARN_INSTALL/lib/protobuf-java-2.4.0a.jar .

Step 13) Yarn daemons are up! But for running mapreduce applications, which now are in user land, you need to setup nodemanager with the following configuration in your yarn-site.xml before you start the nodemanager.
    <property>
      <name>nodemanager.auxiluary.services</name>
      <value>mapreduce.shuffle</value>
    </property>

    <property>
      <name>nodemanager.aux.service.mapreduce.shuffle.class</name>
      <value>org.apache.hadoop.mapred.ShuffleHandler</value>
    </property>

Step 14) You are all set, an example on how to run a mapreduce job is:

cd $HADOOP_MAPRED_HOME
ant examples -Dresolvers=internal
$HADOOP_COMMON_HOME/bin/hadoop jar $HADOOP_MAPRED_HOME/build/hadoop-mapred-examples-0.22.0-SNAPSHOT.jar randomwriter -Dmapreduce.job.user.name=$USER -Dmapreduce.clientfactory.class.name=org.apache.hadoop.mapred.YarnClientFactory -Dmapreduce.randomwriter.bytespermap=10000 -Ddfs.blocksize=536870912 -Ddfs.block.size=536870912 -libjars $HADOOP_YARN_INSTALL/hadoop-mapreduce-1.0-SNAPSHOT/modules/hadoop-mapreduce-client-jobclient-1.0-SNAPSHOT.jar output 

The output on the command line should be almost similar to what you see in the JT/TT setup (Hadoop 0.20/0.21)