Installing Hadoop on Ubuntu (Linux) - single node - Problems you may face

This is not a new post, it is based on Michael G. Noll blog about Running Hadoop on Ubuntu (Single Node)
I will go through the same steps, but I will point out some exceptions/errors you may face.

Because I am a very new user of Ubuntu, this post is mainly targeting the Windows users and they have very primitive knowledge about Linux. I may write some hints in linux which seems very trivial for linux geeks, but it may be fruitful for Windows users.

Moreover, I am assuming that you have enough knowledge about HDFS architecture. You can read this document for more details.

I have used Ubuntu 11.04 and Hadoop 0.20.2.

Prerequisites:

1. Installing Sun JDK 1.6: Installing JDK is a required step to install Hadoop. You can follow the steps in my previous post.

Update
There is another simpler way to install JDK (for example installing JDK 1.7) using the instructions on this post. 

2. Adding a dedicated Hadoop system user: You will need a user for hadoop system you will install. To create a new user "hduser" in a group called "hadoop", run the following commands in your terminal:

$sudo addgroup hadoop

$sudo adduser --ingroup hadoop hduser

3.Configuring SSH: in Michael Blog, he assumed that the SSH is already installed. But if you didn't install SSH server before, you can run the following command in your terminal: By this command, you will have installed ssh server on your machine, the port is 22 by default.

 $sudo apt-get install openssh-server

We have installed SSH because Hadoop requires access to localhost (in case single node cluster) or    communicates with remote nodes (in case multi-node cluster).

After this step, you will need to generate SSH key for hduser (and the users you need to administer Hadoop if any) by running the following commands, but you need first to switch to hduser:

$su - hduser
$ssh-keygen -t rsa -P ""

To be sure that SSH installation is went well, you can open a new terminal and try to create ssh session using hduser by the following command:

$ssh localhost

4. Disable IPv6: You will need to disable IP version 6 because Ubuntu is using 0.0.0.0 IP for different Hadoop configurations. You will need to run the following commands using a root account:
$sudo gedit /etc/sysctl.conf

This command will open sysctl.conf in text editor, you can copy the following lines at the end of the file:

#disable ipv6

net.ipv6.conf.all.disable_ipv6 = 1

net.ipv6.conf.default.disable_ipv6 = 1

net.ipv6.conf.lo.disable_ipv6 = 1

You can save the file and close it. If you faced a problem telling you don't have permissions, just remember to run the previous commands by your root account.

 These steps required you to reboot your system, but alternatively, you can run the following command to re-initialize the configurations again.

$sudo sysctl -p 

To make sure that IPV6 is disabled, you can run the following command:

$cat /proc/sys/net/ipv6/conf/all/disable_ipv6

The printed value should be 1, which means that is disabled.

Installing Hadoop

Now we can download Hadoop to begin installation. Go to Apache Downloads and download Hadoop version 0.20.2. To overcome the security issues, you can download the tar file in hduser directory, for example, /home/hduser. Check the following snapshot:

Then you need to extract the tar file and rename the extracted folder to 'hadoop'. Open a new terminal and run the following command:

$ cd /home/hduser

$ sudo tar xzf hadoop-0.20.2.tar.gz

$ sudo mv hadoop-0.20.2 hadoop

Please note if you want to grant access for another hadoop admin user (e.g. hduser2), you have to grant read permission to folder /home/hduser using the following command:

sudo chown -R hduser2:hadoop hadoop  

Update $HOME/.bashrc

You will need to update the .bachrc for hduser (and for every user you need to administer Hadoop). To open .bachrc file, you will need to open it as root:

$sudo gedit /home/hduser/.bashrc

Then you will add the following configurations at the end of .bachrc file

# Set Hadoop-# related environment variables

export HADOOP_HOME=/home/hduser/hadoop

# Set JAVA_HOME (we will also configure JAVA_HOME directly for Hadoop later on)

export JAVA_HOME=/usr/lib/jvm/java-6-sun
# or you can write the following command if you used this post to install your java
# export JAVA_HOME=/usr/lib/jvm/jdk1.7.0_71

# Some convenient aliases and functions for running Hadoop-related commands

unalias fs &> /dev/null

alias fs="hadoop fs"

unalias hls &> /dev/null

alias hls="fs -ls"

# If you have LZO compression enabled in your Hadoop cluster and

# compress job outputs with LZOP (not covered in this tutorial):

# Conveniently inspect an LZOP compressed file from the command

# line; run via:

#

# $ lzohead /hdfs/path/to/lzop/compressed/file.lzo

#

# Requires installed 'lzop' command.

#

lzohead () {

    hadoop fs -cat $1 | lzop -dc | head -1000 | less

}

# Add Hadoop bin/ directory to PATH

export PATH=$PATH:$HADOOP_HOME/bin

Hadoop Configuration

 

Now, we need to configure Hadoop framework on Ubuntu machine. The following are configuration files we can use to do the proper configuration. To know more about hadoop configurations, you can visit this site

hadoop-env.sh

We need only to update the JAVA_HOME variable in this file. Simply you will open this file using a text editor using the following command:

$sudo gedit /home/hduser/hadoop/conf/hadoop-env.sh

Then you will need to change the following line

# export JAVA_HOME=/usr/lib/j2sdk1.5-sun

To 

export JAVA_HOME=/usr/lib/jvm/java-6-sun

or you can write the following command if you used this post to install your java
# export JAVA_HOME=/usr/lib/jvm/jdk1.7.0_71

Note: if you faced "Error: JAVA_HOME is not set" Error while starting the services, then you seems that you forgot toe uncomment the previous line (just remove #).

core-site.xml

First, we need to create a temp directory for Hadoop framework. If you need this environment for testing or a quick prototype (e.g. develop simple hadoop programs for your personal test ...), I suggest to create this folder under /home/hduser/ directory, otherwise, you should create this folder in a shared place under shared folder (like /usr/local ...) but you may face some security issues. But to overcome the exceptions that may caused by security (like java.io.IOException), I have created the tmp folder under hduser space.

To create this folder, type the following command:

$ sudo mkdir  /home/hduser/tmp

Please note that if you want to make another admin user (e.g. hduser2 in hadoop group), you should grant him a read and write permission on this folder using the following commands:

$ sudo chown hduser2:hadoop /home/hduser/tmp

$ sudo chmod 755 /home/hduser/tmp

Now, we can open hadoop/conf/core-site.xml to edit the hadoop.tmp.dir entry.

We can open the core-site.xml using text editor:

$sudo gedit /home/hduser/hadoop/conf/core-site.xml

Then add the following configurations between <configuration> .. </configuration> xml elements:

<!-- In: conf/core-site.xml -->

<property>

  <name>hadoop.tmp.dir</name>

  <value>/home/hduser/tmp</value>

  <description>A base for other temporary directories.</description>

</property>

<property>

  <name>fs.default.name</name>

  <value>hdfs://localhost:54310</value>

  <description>The name of the default file system.  A URI whose

  scheme and authority determine the FileSystem implementation.  The

  uri's scheme determines the config property (fs.SCHEME.impl) naming

  the FileSystem implementation class.  The uri's authority is used to

  determine the host, port, etc. for a filesystem.</description>

</property>

mapred-site.xml

We will open the hadoop/conf/mapred-site.xml using a text editor and add the following configuration values (like core-site.xml)

<!-- In: conf/mapred-site.xml -->

<property>

  <name>mapred.job.tracker</name>

  <value>localhost:54311</value>

  <description>The host and port that the MapReduce job tracker runs

  at.  If "local", then jobs are run in-process as a single map

  and reduce task.

  </description>

</property>

hdfs-site.xml

Open hadoop/conf/hdfs-site.xml using a text editor and add the following configurations:

<!-- In: conf/hdfs-site.xml -->

<property>

  <name>dfs.replication</name>

  <value>1</value>

  <description>Default block replication.

  The actual number of replications can be specified when the file is created.

  The default is used if replication is not specified in create time.

  </description>

</property>

Formatting NameNode

You should format the NameNode in your HDFS. You should not do this step when the system is running. It is usually done once at first time of your installation.

Run the following command

$/home/hduser/hadoop/bin/hadoop namenode -format

NameNode Formatting

Starting Hadoop Cluster

You will need to navigate to hadoop/bin directory and run ./start-all.sh script.

Starting Hadoop Services using ./start-all.sh

There is a nice tool called jps. You can use it to ensure that all the services are up.

Using jps tool

Running an Example (Pi Example)

There are many built-in examples. We can run PI estimator example using the following command:

hduser@ubuntu:~/hadoop/bin$ hadoop jar ../hadoop-0.20.2-examples.jar pi 3 10

If you faced "Incompatible namespaceIDs" Exception you can do the following:

1.  Stop all the services (by calling ./stop-all.sh).

2.  Delete /tmp/hadoop/dfs/data/*

3.  Start all the services.

Java Performance Tips 2

Introduction

We talked about Strings operations in the last article, now we will talk about more tips on performance tips, we will talk about the importance of creating stateless methods and class if your logic doesn't depend on the state of the object, also we will talk about object reusing and how will affect the performance by reducing the headache of garbage collection.

Take 'State or Stateless' Decision

Creating and destroying objects issue in java can cause performance issues, for example creating State classes for stateless data; a stateless class means that the data of the created object doesn't depend on the state where there is no fields or attributes that makes an object differs from other.

We can imagine that the Stateless class can say 'Use me only if your code doesn't depend on your object state, All the objects are functionally equivalent to me, you will gain besides that no creation of objects, the CPU will be happy of doing that J'

Code Example

The following StatelessClassWithoutStaticMethods class has a method called execute, where the logic of this method doesn't depend on the state of the object, while it is not static method.

package performancetest.episode2;

public class StatelessClassWithoutStaticMethods {

      public void execute()

      {

            // the logic here doesn't depend on the object state

            for (int i = 0; i < 10000; i++) {

                  System.out.println(i);

            }

      }

}

While the right version as the following:

package performancetest.episode2;

public class StatelessClassWithStaticMethods {

      public static void execute()

      {

            // the logic here doesn't depend on the object state

            for (int i = 0; i < 10000; i++) {

                  System.out.println(i);

            }

      }

}

Performance Analysis

Using TBTB (Eclipse Test & Performance Tools Platform Project), the following results are shown as follows

Performance Analysis

Based on this figure, the static execute method has loss approximately 18% of time from the non-static one.

That doesn't mean that you have to make all the class a static (i.e. a class has static methods only), but if you have a state method in this class depends on the object state, make this method only non-static.

Don't Create A New One If You Can What You Have

'You can use me before I go to the garbage collection, clear my data and reuse me again; we want to lessen the work on our friend garbage collection' Any object says this quote J, Creating new objects is expensive as we know as more objects we create, the garbage collection job will be more time consuming.

So if you have a chance to reuse the object again without creating a new one of the same type, do it immediately, it will improve the performance and will make the garbage collection job be easier.

Code Example

public static void fillVectorDataWithoutRecycling()

      {

            for (int i = 0 ; i < 1000 ; i ++)

            {

                  Vector v = new Vector();

                  v.add("item1");

                  v.add("item2");

                  v.add("item3");

                  System.out.println(v.get(0));

            }

      }

     

      public static void fillVectorDataWithRecycling()

      {

            Vector v = new Vector();

            for (int i = 0 ; i < 1000 ; i ++)

            {

                  v.clear();

                  v.add("item1");

                  v.add("item2");

                  v.add("item3");

                  System.out.println(v.get(0));

            }

      }

We have two methods, fillVectorDataWithoutRecycling method creates a new vector object in the loop, while fillVectorDataWithRecycling method uses one vector object and clears the vector object and reuse it.

Performance Analysis

Performance Analysis With/Without Vector Recycling

From this figure we found that reusing the current object will cause loss approximately 50% of time (of course without taking in the consideration the hardware architecture like CPU, cache and memory).

Java Performance Tips 1 - String Operations

String Operations

There are some performance tips for String manipulation in java, one of them is the concatenation operation.

Concatenation operation can be done by appending the values on the same String object for example:

String str = new String ();
String s = "test";
str+=s;

str+="testString";

The compiler translated this simple line to the following J

str = (new StringBuffer()).append(s).append("testString").toString();

But this method is not preferable (as we will know later), one other method is using StringBuffer .

Using StringBuffer Method

StringBuffer is used to store character strings that will be changed as we know that String class is immutable, so we can concatenate the strings as follows:

StringBuffer sbuffer = new StringBuffer();

sbuffer.append("testString");

StringBuffer vs StringBuilder

Also there is another method to concatenate the String using StringBuilder which is introduced in Java 5, StringBuilder is like the StringBuffer except it is not synchronized, which means that if there are many threads, they can change it in the same time (StringBuilder is not suitable in the multithreading applications).

'Ok, why this stuff for, just for concatenate some strings!' you may ask this question, after running a sample of each and profiling the performance.

Code Example

public class StringOperations {

    public void concatenateUsingString() {

        String str = new String();

        for (int i = 0; i < 10000; i++) {

            str += "testString";

        }

    }

    public void concatenateUsingStringBuffer() {

        StringBuffer sbuffer = new StringBuffer();

        for (int i = 0; i < 10000; i++) {

            sbuffer.append("testString");

        }

    }

    public void concatenateUsingStringBuilder() {

        StringBuilder sbuilder = new StringBuilder();

        for (int i = 0; i < 10000; i++) {

            sbuilder.append("testString");

        }

    }

}

And in the main method, a simple calling to the three methods

public static void main(String[] args) {

        StringOperations soperations = new StringOperations();        

        soperations.concatenateUsingString();

        soperations.concatenateUsingStringBuffer();

        soperations.concatenateUsingStringBuilder();

    }

The result is like the following:
The Results by seconds

The Results by percentage

I have used the Eclipse Test & Performance Tools Platform Project (TPTP) to validate the results, just right click on the project and choose 'Profile As'.

Then choose 'ProfileàProfile Configuration '

Profile Configuration

This result shows the big performance issue of using the String concatenation (Plus operation),

the profiler tells us that the calling of concatenateUsingStringBuffer and concatenateUsingStringBuilder (approximately 0.08%) of time are nothing with respect to concatenateUsingString (99.85% of time).

Now you can go to Java Performance Tips 2 article

References

    http://www.javaworld.com/javaworld/jw-03-2000/jw-0324-javaperf.html
    http://leepoint.net/notes-java/data/strings/23stringbufferetc.html

Installing JDK 1.6 on Ubuntu

• Installing Sun JDK 1.6: You can download oab-java6 zip file from this URL , then you will need to extract this zip file and open the terminal (pressing ctrl+alt+t) to run ./oab-java6.sh. You must be run this command as sudo (using a root account).

 This script is doing the following:

The basic execution steps are:

• Remove, my now disabled, Java PPA 'ppa:flexiondotorg/java'.
• Install the tools required to build the Java packages.
• Create download cache in /var/local/oab/pkg.
• Download the i586 and x64 Java install binaries from Oracle. Yes, both are required.
• Clone the build scripts from https://github.com/rraptorr/sun-java6
• Build the Java packages applicable to your system.
• Create local apt repository in /var/local/oab/deb for the newly built Java Packages.
• Create a GnuPG signing key in /var/local/oab/gpg if none exists.
• Sign the local apt repository using the local GnuPG signing key.

What if this script didn't go well with you ? on of the problems you may face is throwing an error saying "Error: Package sun-java6-jdk is not available, but is referred to by another package"

You can run the following in the terminal :

sudo add-apt-repository ppa:ferramroberto/java
sudo apt-get update

sudo apt-get install sun-java6-jdk

By previous steps, it is expected that you have installed JDK 1.6 on your systems. The only remaining step is to set the class path of Java by the following commands:

export JAVA_HOME=/usr/lib/jvm/java-6-sun

To ensure that the new Java Home is set correctly, you can print (echo) the JAVA_HOME environment variable by the following command:

echo $JAVA_HOME