How To Install the Big-Data Friendly Apache Accumulo NoSQL Database on Ubuntu 14.04

You will need the following:

• Ubuntu 14.04 server (preferably 32-bit)
• A sudo user
• At least 2 GB of swap space

Accumulo, HDFS, and ZooKeeper are all written in Java and need a JVM (Java Virtual Machine) to run. So, let’s start by installing the JDK.

Update the package list index.

sudo apt-get update

Install OpenJDK 7 using apt-get.

sudo apt-get install openjdk-7-jdk

Use nano to edit your shell environment file, .bashrc.

nano ~/.bashrc

Add JAVA_HOME as an environment variable at the end of the file.

export JAVA_HOME=/usr/lib/jvm/java-7-openjdk-i386

Note: The value of JAVA_HOME can be different depending on your server’s architecture. For example, on a 64-bit server, the value would be /usr/lib/jvm/java-7-openjdk-amd64. You can see the exact path by listing the contents of the /usr/lib/jvm/ directory. If your path is different from what’s shown here, make sure you make the appropriate changes here and elsewhere.

Save the file and exit nano. Update the environment variables of the current session by typing:

. ~/.bashrc

Edit the JVM’s java.security configuration file using nano.

sudo nano $JAVA_HOME/jre/lib/security/java.security

Search for the parameter securerandom.source and change the line so that it looks like this:

securerandom.source=file:/dev/./urandom

Save the file and exit nano. This change is necessary to decrease the JVM’s startup time. Not making this change can lead to very long startup times on most virtual servers.

Hadoop needs SSH and Rsync to manage its daemons. Install them using the following command:

sudo apt-get install ssh rsync

Hadoop should be able to connect to your server over SSH without being prompted for a password.

Generate an RSA key using ssh-keygen.

ssh-keygen -P ''

Press ENTER when prompted, to choose the default values.

Add the generated key to the authorized_keys file.

cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys

The values localhost and 0.0.0.0 should be added to the list of known hosts. The easiest way to do this is by running the ssh command.

Let us add localhost first.

ssh localhost

You will be prompted by a message that looks like this:

The authenticity of host 'localhost (127.0.0.1)' can't be established.
ECDSA key fingerprint is bf:01:63:5b:91:aa:35:db:ee:f4:7e:2d:36:e7:de:42.
Are you sure you want to continue connecting (yes/no)?

Type in yes and press ENTER.

Once the login is complete, exit the child SSH session by typing in:

exit

Let us add 0.0.0.0 now.

ssh 0.0.0.0

Type in yes and press ENTER when prompted.

Once again, exit the child SSH session by typing in:

exit

SSH setup is now complete.

You will be downloading a couple of files for this tutorial. Though not really necessary, it is a good idea to store all your downloads in a separate directory.

mkdir -p ~/Downloads

Enter the directory.

cd ~/Downloads

At the time of writing, the latest stable version of Hadoop is 2.6.0. Download it using wget.

wget "http://www.eu.apache.org/dist/hadoop/common/stable/hadoop-2.6.0.tar.gz"

The latest stable version of ZooKeeper is 3.4.6. Download it using wget.

wget "http://www.eu.apache.org/dist/zookeeper/stable/zookeeper-3.4.6.tar.gz"

The latest stable version of Accumulo is 1.6.1. Download it using wget.

wget "http://www.eu.apache.org/dist/accumulo/1.6.1/accumulo-1.6.1-bin.tar.gz"

Create a directory to store all of our Accumulo-related installations.

mkdir -p ~/Installs

Enter the directory.

cd ~/Installs

Use the tar command to extract the contents of hadoop-2.6.0-src.tar.gz.

tar -xvzf ~/Downloads/hadoop-2.6.0.tar.gz

Note: If you installed a different version of any of this software, please use the appropriate version in your file name.

Use nano to open hadoop-env.sh.

nano ~/Installs/hadoop-2.6.0/etc/hadoop/hadoop-env.sh

Look for the line that starts with export JAVA_HOME and change it to:

export JAVA_HOME=/usr/lib/jvm/java-7-openjdk-i386

Make sure this value is identical to the value you set in .bashrc.

By default Hadoop generates a lot of debug logs. To stop this behavior, look for the line that starts with export HADOOP_OPTS and change it to:

export HADOOP_OPTS="$HADOOP_OPTS -XX:-PrintWarnings -Djava.net.preferIPv4Stack=true"

Save and exit.

Use nano to open core-site.xml.

nano ~/Installs/hadoop-2.6.0/etc/hadoop/core-site.xml

Add a <property> block named fs.defaultFS. Its value should point to the namenode’s hostname and port (in our case, it is localhost and the default port 9000). Ignoring the comments, edit your file so that it looks like this:

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

Save and exit.

Use nano to open hdfs-site.xml.

nano ~/Installs/hadoop-2.6.0/etc/hadoop/hdfs-site.xml

The following properties need to be added to this file:

• dfs.replication: This number specifies how many times a block is replicated by Hadoop. By default, Hadoop creates 3 replicas for each block. In this tutorial, use the value 1, as we are not creating a cluster.

• dfs.name.dir: This points to a location in the filesystem where the namenode can store the name table. You need to change this because Hadoop uses /tmp by default. Let us use hdfs_storage/name to store the name table.

• dfs.data.dir: This points to a location in the filesystem where the datanode should store its blocks. You need to change this because Hadoop uses /tmp by default. Let us use hdfs_storage/data to store the data blocks.

Ignoring the comments, after adding these properties, your file should look like this:

<?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.name.dir</name>
        <value>hdfs_storage/name</value>
    </property>
    <property>
        <name>dfs.data.dir</name>
        <value>hdfs_storage/data</value>
    </property>
</configuration>

Use nano to create a new file named mapred-site.xml.

nano ~/Installs/hadoop-2.6.0/etc/hadoop/mapred-site.xml

Add a property named mapred.job.tracker to this file. This property contains the hostname and port number on which the MapReduce job tracker runs. For our setup, use localhost and the default port 9001.

Add the following content to the file:

<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
     <property>
         <name>mapred.job.tracker</name>
         <value>localhost:9001</value>
     </property>
</configuration>

Enter Hadoop’s base directory (this is important because Hadoop creates the hdfs_storage directory in the current directory).

cd ~/Installs/hadoop-2.6.0/

The NameNode can now be initialized by typing in:

~/Installs/hadoop-2.6.0/bin/hdfs namenode -format

You should see quite a bit of output.

Next, start the NameNode by typing in:

~/Installs/hadoop-2.6.0/sbin/start-dfs.sh

Wait a minute or two for it to start. Once started, you can use a browser to visit http://<your-ip>:50070/ and browse through the web interface of the NameNode.

Troubleshooting

If you are not able to access the web interface, check if the NameNode is active by using the following command:

jps

Your output should contain the following three processes along with the Jps process:

• DataNode
• NameNode
• SecondaryNameNode

If you see that NameNode is not present in the output, perform the following steps. If they don’t execute in a block, you may have to run them separately. Comments are inluded in-line.

cd ~/Installs/hadoop-2.6.0/
~/Installs/hadoop-2.6.0/sbin/stop-dfs.sh # Stop Hadoop's nodes
rm -rf hdfs_storage # Delete the namenode data
rm -rf /tmp/hadoop-* # Delete the temporary directories
~/Installs/hadoop-2.6.0/bin/hdfs namenode -format # Reformat the namenode

Restart Hadoop using start-dfs.sh:

~/Installs/hadoop-2.6.0/sbin/start-dfs.sh

You should be able to access the web interface now.

Enter the Installs directory.

cd ~/Installs

Use tar to extract zookeeper-3.4.6.tar.gz.

tar -xvzf ~/Downloads/zookeeper-3.4.6.tar.gz

Copy the example file zoo_sample.cfg to zoo.cfg.

cp ~/Installs/zookeeper-3.4.6/conf/zoo_sample.cfg ~/Installs/zookeeper-3.4.6/conf/zoo.cfg

Configuration of ZooKeeper is now complete. Start ZooKeeper by typing in:

~/Installs/zookeeper-3.4.6/bin/zkServer.sh start

You should see output that looks like this:

JMX enabled by default
Using config: ~/Installs/zookeeper-3.4.6/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED

Now that all its dependencies are satisfied, it is time to work on the installation of Accumulo itself.

Enter the Installs directory.

cd ~/Installs

Extract accumulo-1.6.1-bin.tar.gz using tar:

tar -xvzf ~/Downloads/accumulo-1.6.1-bin.tar.gz

Accumulo comes with sample configurations for servers with various memory sizes: 512 MB, 1 GB, 2 GB and 3 GB. I’m going to use the configuration for 512 MB in this tutorial. You can pick another configurations if your server has more memory.

Copy the 512 MB configuration files to the conf directory.

cp ~/Installs/accumulo-1.6.1/conf/examples/512MB/standalone/* ~/Installs/accumulo-1.6.1/conf/

Use nano to edit your shell environment again, using the .bashrc file.

nano ~/.bashrc

Add the following environment variables to this file:

• HADOOP_HOME: The path to the Hadoop installation
• ZOOKEEPER_HOME: The path to the ZooKeeper installation

Add the following lines to your file:

export HADOOP_HOME=~/Installs/hadoop-2.6.0/
export ZOOKEEPER_HOME=~/Installs/zookeeper-3.4.6/

Save and exit.

Update the environment so that the variables you added to .bashrc are available in the current session.

. ~/.bashrc

Use nano to edit accumulo-env.sh.

nano ~/Installs/accumulo-1.6.1/conf/accumulo-env.sh

By default, Accumulo’s HTTP monitor binds only to the local network interface. To be able to access it over the Internet, you have to set the value of ACCUMULO_MONITOR_BIND_ALL to true.

Find the line that starts with export ACCUMULO_MONITOR_BIND_ALL and uncomment it. It should look like this:

export ACCUMULO_MONITOR_BIND_ALL="true"

Save and exit.

Use nano to edit accumulo-site.xml.

nano ~/Installs/accumulo-1.6.1/conf/accumulo-site.xml

Accumulo’s worker processes communicate with each other using a secret key. This should be changed to a string which is secure. Search for the property instance.secret and change its value. I’m going to use this string: PASS1234. The XML for the property should look like this:

<property>
    <name>instance.secret</name>
    <value>PASS1234</value>
    <description>A secret unique to a given instance that all servers must know in order to communicate with one another.
      Change it before initialization. To
      change it later use ./bin/accumulo org.apache.accumulo.server.util.ChangeSecret --old [oldpasswd] --new [newpasswd],
      and then update this file.
    </description>
</property>

Next, add a new property named instance.volumes. The value of this property specifies where Accumulo should store its data in the HDFS. Let us store the data in the directory /accumulo.

<property>
    <name>instance.volumes</name>
    <value>hdfs://localhost:9000/accumulo</value>
</property>

Find the property trace.token.property.password, and set its value to something secure. Remember this value as you will need it in the next step. I am going to set this to mypassw.

  <property>
    <name>trace.token.property.password</name>
    <value>mypassw</value>
  </property>

Save and exit.

Initialize Accumulo.

~/Installs/accumulo-1.6.1/bin/accumulo init

You will be prompted to enter an Instance name. Use any name of your choice. I choose DIGITAL_OCEAN.

Next, you will be prompted for a password. Type in the same password that you used for the property trace.token.property.password.

Once the command completes, you can start Accumulo.

~/Installs/accumulo-1.6.1/bin/start-all.sh

You might see a few warnings recommending higher values for certain system parameters. As we are creating a very small instance in this tutorial, you can ignore those warnings.

Once the startup is complete, you can use a browser to visit Accumulo’s web interface at http://<your-server-ip>:50095.

Conclusion

In this tutorial, you have learned how to set up Apache Accumulo and all the other components it depends on. We have created a very simple setup today using HDFS in pseudo-distributed mode that can run on a single small server. For optimal performance in a production scenario, HDFS should be running in the fully-distributed mode.

In production scenarios, it is also highly recommended that these processes be deployed on servers that have at least 8GB of RAM and 4 or more processor cores so that each process can use over 2GB of memory and a separate core. Alternately, you could deploy the components separately on different servers.

For more information, refer to Apache Accumulo’s user manual.