.. _hadoop:

Apache\* Hadoop\*
#################

This tutorial walks you through the process of installing, configuring, and
running Apache Hadoop on |CL-ATTR|. The Apache Hadoop software library is a
framework for distributed processing of large data sets across clusters of
computers using simple programming models. It is designed to scale up from
single servers to thousands of machines, with each machine offering local
computation and storage.

Prerequisites
*************

Before following this tutorial, you should follow the
:ref:`bare-metal-install-desktop` to ensure you have installed |CL|.

Before you install any new packages, update |CL| with the following command:

.. code-block:: bash

   sudo swupd update

For the purposes of this tutorial, we will install Hadoop in a single machine
running both the master and slave daemons.

Install Apache Hadoop
*********************

Apache Hadoop is included in the `big-data-basic` bundle. To install the
framework, enter the following command:

.. code-block:: bash

   sudo swupd bundle-add big-data-basic

Configure Apache Hadoop
***********************

#. To create the configuration directory, enter the following command:

   .. code-block:: bash

      sudo mkdir /etc/hadoop

#. Copy the defaults from :file:`/usr/share/defaults/hadoop` to
   :file:`/etc/hadoop` with the following command:

   .. code-block:: bash

      $ sudo cp /usr/share/defaults/hadoop/* /etc/hadoop

   .. note:: Since |CL| is a stateless system, you should never modify the
      files under the :file:`/usr/share/defaults` directory. The software
      updater will overwrite those files.

Once all the configuration files are in :file:`/etc/hadoop`, we must edit
them to fit our needs. The `NameNode` server is the master server. It manages
the namespace of the files system and regulates the clients' access to files.
The first file we edit, :file:`/etc/hadoop/core-site.xml`, informs the Hadoop
daemon where `NameNode` is running.

In this tutorial, our `NameNode` runs in our `localhost`. Follow these steps
to set it up correctly:

#. Open the :file:`/etc/hadoop/core-site.xml` file using the editor of your
   choice and modify the file as follows:

   .. code-block:: xml

      <?xml version="1.0" encoding="UTF-8"?>
      <?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
      <configuration>
      <property>
      <name>fs.default.name</name>
      <value>hdfs://localhost:9000</value>
      </property>
      </configuration>

#. Edit the :file:`/etc/hadoop/hdfs-site.xml` file. This file configures the
   :abbr:`HDFS (Hadoop Distributed File System)` daemons. This configuration
   includes the list of permitted and excluded data nodes and the size of
   said blocks. In this example, we are setting the number of block
   replication to 1 from the default of 3 as follows:

   .. code-block:: xml
      :emphasize-lines: 6

      <?xml version="1.0" encoding="UTF-8"?>
      <?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
      <configuration>
      <property>
      <name>dfs.replication</name>
      <value>1</value>
      </property>
      <property>
      <name>dfs.permission</name>
      <value>false</value>
      </property>
      </configuration>

#. Edit the :file:`/etc/hadoop/mapred-site.xml` file. This file configures
   all daemons related to `MapReduce`: `JobTracker` and `TaskTrackers`. With
   `MapReduce`, Hadoop can process big amounts of data in multiple systems. In
   our example, we set :abbr:`YARN (Yet Another Resource Manager)` as our
   runtime framework for executing `MapReduce` jobs as follows:

   .. code-block:: xml
      :emphasize-lines: 5,6

      <?xml version="1.0" encoding="UTF-8"?>
      <?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
      <configuration>
      <property>
      <name>mapreduce.framework.name</name>
      <value>yarn</value>
      </property>
      </configuration>

#. Edit the :file:`/etc/hadoop/yarn-site.xml` file. This file configures all
   daemons related to `YARN`: `ResourceManager` and `NodeManager`. In our
   example, we implement the `mapreduce_shuffle` service, which is the
   default as follows:

   .. code-block:: xml
      :emphasize-lines: 4,5,8,9

      <?xml version="1.0"?>
      <configuration>
      <property>
      <name>yarn.nodemanager.aux-services</name>
      <value>mapreduce_shuffle</value>
      </property>
      <property>
      <name>yarn.nodemanager.auxservices.mapreduce.shuffle.class</name>
      <value>org.apache.hadoop.mapred.ShuffleHandler</value>
      </property>
      </configuration>

Configure your SSH key
**********************

#. Create a SSH key. If you already have one, skip this step.

   .. code-block:: bash

      sudo ssh-keygen -t rsa


#. Copy the key to your authorized keys.

   .. code-block:: bash

      sudo cat /root/.ssh/id_rsa.pub | sudo tee -a /root/.ssh/authorized_keys

#. Log into the localhost. If no password prompt appears, you are ready to
   run the Hadoop daemons.

   .. code-block:: bash

      sudo ssh localhost

Run the Hadoop daemons
**********************

With all the configuration files properly edited, we are ready to start the
daemons.

When we format the `NameNode` server, it formats the meta-data related to
data nodes. Thus, all the information on the data nodes is lost and the nodes
can be reused for new data.

#. Format the `NameNode` server with the following command:

   .. code-block:: bash

      sudo hdfs namenode -format

#. Start the DFS in `NameNode` and `DataNodes` with the following command:

   .. code-block:: bash

      sudo start-dfs.sh

#. The console output should be similar to:

   .. code-block:: console

      Starting namenodes on [localhost]
      The authenticity of host 'localhost (::1)' can't be established.
      ECDSA key fingerprint is
      SHA256:97e+7TnomsS9W7GjFPjzY75HGBp+f1y6sA+ZFcOPIPU.
      Are you sure you want to continue connecting (yes/no)?

   Enter `yes` to continue.

#. Start the `YARN` daemons `ResourceManager` and `NodeManager` with the
   following command:

   .. code-block:: bash

      sudo start-yarn.sh

#. Ensure everything is running as expected with the following command:

   .. code-block:: bash

      sudo jps

#. The console output should be similar to:

   .. code-block:: console

      22674 DataNode
      26228 Jps
      22533 NameNode
      23046 ResourceManager
      22854 SecondaryNameNode
      23150 NodeManager

Run the MapReduce wordcount example
***********************************

#. Create the input directory.

   .. code-block:: bash

      sudo hdfs dfs -mkdir -p /user/root/input

#. Copy a file from the local file system to the HDFS.

   .. code-block:: bash

      sudo hdfs dfs -copyFromLocal local-file /user/root/input

#. Run the `wordcount` example.

   .. code-block:: bash

      sudo hadoop jar /usr/share/hadoop/mapreduce/hadoop-mapreduce-examples-2.8.0.jar wordcount input output

#. Read output file "part-r-00000". This file contains the number of times
   each word appears in the file.

   .. code-block:: bash

      sudo hdfs dfs -cat /user/root/output/part-r-00000

**Congratulations!**

You successfully installed and setup a single node Hadoop cluster.
Additionally, you ran a simple wordcount example.

Your single node Hadoop cluster is up and running!