hadoop linux wsl

Install Hadoop 3.2.0 on Windows 10 using Windows Subsystem for Linux (WSL)

1,431 views 12 comments about 4 days ago Raymond Tang

In my previous post, I showed how to configure a single node Hadoop instance on Windows 10. The steps are not too difficult to follow if you have Java programming background. However there is one step that is not very straightforward: native Hadoop executable (winutils.exe) is not included in the official Hadoop distribution and needs to be downloaded separately or built locally.  In Linux or UNIX, you don’t usually need to do that since the native libs are pre-compiled and included in the binary distribution.

In August 2016, Microsoft has published the initial release of Windows Subsystem for Linux (WSL). In Jun this year, WSL 2.0 will also be released with enhanced performance. With WSL, we can run Linux as subsystem in Windows 10. In this post, I am going to show you how to install Hadoop 3.2.0 in WSL.

Prerequisites

Follow the page below to enable WSL and then install one of the Linux systems from Microsoft Store.

Windows Subsystem for Linux Installation Guide for Windows 10

To be specific, enable WSL by running the following PowerShell code as Administrator (or enable it through Control Panel):

Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Windows-Subsystem-Linux

And then install Ubuntu from Microsoft Store.

image

image

Once download is completed, click Launch button to lunch the application. It make take a few minutes to install:

image

During the installation, you need to input a username and password. Once it is done, you are ready to use the Ubuntu terminal:

image

Install Hadoop 3.2.0 in WSL

Install Java JDK

Run the following command to update package index:

sudo apt update

Check whether Java is installed already:

java -version

Command 'java' not found, but can be installed with:

sudo apt install default-jre
sudo apt install openjdk-11-jre-headless
sudo apt install openjdk-8-jre-headless

Install OpenJDK via the following command:

sudo apt-get install openjdk-8-jdk

Check the version installed:

java -version
openjdk version "1.8.0_191"
OpenJDK Runtime Environment (build 1.8.0_191-8u191-b12-2ubuntu0.18.04.1-b12)
OpenJDK 64-Bit Server VM (build 25.191-b12, mixed mode)

*Java 11 is not supported yet by Hadoop as at 2019-05-11.

Download Hadoop binary

Go to release page of Hadoop website to find a download URL for Hadoop 3.2.0:

Hadoop Releases

For me, the close mirror is:

http://mirror.intergrid.com.au/apache/hadoop/common/hadoop-3.2.0/hadoop-3.2.0.tar.gz

Run the following command in Ubuntu terminal to download a binary from the internet:

wget http://mirror.intergrid.com.au/apache/hadoop/common/hadoop-3.2.0/hadoop-3.2.0.tar.gz

Unzip Hadoop binary

Run the following command to create a hadoop folder under home folder:

mkdir ~/hadoop

And then run the following command to unzip the binary package:

tar -xvzf hadoop-3.2.0.tar.gz -C ~/hadoop

Once it is unzipped, change the current directory to the hadoop folder:

cd ~/hadoop/hadoop-3.2.0/

Configure passphraseless ssh

This step is critical and please make sure you follow the steps.

Make sure you can SSH to localhost in Ubuntu:

ssh localhost

If you cannot ssh to localhost without a passphrase, run the following command to initialize your private and public keys:

ssh-keygen -t rsa -P '' -f ~/.ssh/id_rsa
cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys
chmod 0600 ~/.ssh/authorized_keys

If you encounter errors like ‘ssh: connect to host localhost port 22: Connection refused’, run the following commands:

sudo apt-get install ssh

And then restart the service:

sudo service ssh restart

If the above commands still don’t work, try the solution in this comment.

Configure the pseudo-distributed mode (Single-node mode)

Now, we can follow the official guide to configure a single node:

Pseudo-Distributed Operation

The steps are very similar to the ones in my previous post.

Edit etc/hadoop/hadoop-env.sh file:

vi etc/hadoop/hadoop-env.sh

Set a JAVA_HOME environment variable:

export JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64

Edit etc/hadoop/core-site.xml:

vi etc/hadoop/core-site.xml

Add the following configuration:

<configuration>
     <property>
         <name>fs.defaultFS</name>
         <value>hdfs://localhost:9000</value>
     </property>
</configuration>

Edit etc/hadoop/hdfs-site.xml:

vi etc/hadoop/hdfs-site.xml

Add the following configuration:

<configuration>
     <property>
         <name>dfs.replication</name>
         <value>1</value>
     </property>
</configuration>

Edit file etc/hadoop/mapred-site.xml:

vi etc/hadoop/mapred-site.xml

Add the following configuration:

<configuration>
     <property>
         <name>mapreduce.framework.name</name>
         <value>yarn</value>
     </property>
     <property>
         <name>mapreduce.application.classpath</name>
         <value>$HADOOP_MAPRED_HOME/share/hadoop/mapreduce/*:$HADOOP_MAPRED_HOME/share/hadoop/mapreduce/lib/*</value>
     </property>
</configuration>

Edit file etc/hadoop/yarn-site.xml:

vi etc/hadoop/yarn-site.xml

Add the following configuration:

<configuration>
     <property>
         <name>yarn.nodemanager.aux-services</name>
         <value>mapreduce_shuffle</value>
     </property>
     <property>
         <name>yarn.nodemanager.env-whitelist</name>
         <value>JAVA_HOME,HADOOP_COMMON_HOME,HADOOP_HDFS_HOME,HADOOP_CONF_DIR,CLASSPATH_PREPEND_DISTCACHE,HADOOP_YARN_HOME,HADOOP_MAPRED_HOME</value>
     </property>
</configuration>

Format namenode

Run the following command to format the name node:

bin/hdfs namenode -format

Run DFS daemons

Run the following commands to start NameNode and DataNode daemons:

sbin/start-dfs.sh

tangr@Raymond-Alienware:~/hadoop/hadoop-3.2.0$ sbin/start-dfs.sh
Starting namenodes on [localhost]
Starting datanodes
Starting secondary namenodes [Raymond-Alienware]
Raymond-Alienware: Warning: Permanently added 'raymond-alienware' (ECDSA) to the list of known hosts.

You can view the name node through the following URL:

http://localhost:9870/dfshealth.html#tab-overview

The web UI looks like the following:

image

Run YARN daemon

Run the following command to start YARN daemon:

sbin/start-yarn.sh

tangr@Raymond-Alienware:~/hadoop/hadoop-3.2.0$ sbin/start-yarn.sh
Starting resourcemanager
Starting nodemanagers

Once the services are started, you can view the YARN resource manager web UI through the following URL:

http://localhost:8088/cluster

The web UI looks like the following:

image

Unhealthy nodes

As I am currently run the WLS Ubuntu terminal in C drive and my C drive is almost full (available capacity is lower than 10%); thus the single node is not started successfully.

For more details, refer to my post: Hadoop on Windows - UNHEALTHY Data Nodes Fix.

You can also install WSL Ubuntu in other drive (instead of C drive). 

Refer to the official guide to learn how to manually install WSL in a non-system drive:

 Install Windows Subsystem for Linux on a Non-System Drive

org.apache.hadoop.http.HttpServer2: HttpServer.start() threw a non Bind IOException java.net.SocketException: Permission denied

You may encounter this issue:

INFO org.apache.hadoop.http.HttpServer2: HttpServer.start() threw a non Bind IOException
java.net.SocketException: Permission denied
     at sun.nio.ch.Net.bind0(Native Method)
     at sun.nio.ch.Net.bind(Net.java:433)
     at sun.nio.ch.Net.bind(Net.java:425)
     at sun.nio.ch.ServerSocketChannelImpl.bind(ServerSocketChannelImpl.java:223)

Name node service cannot be started as socket bind cannot be established. As we are not using privileged ports in core-site configuration, I could not find out the root cause for this one yet. However after I restart my Windows computer, this issue is resolved automatically. 

Environment variables

To make it easier to run Hadoop commands, add the following environment variables into .bashrc file in your home folder:

export JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-amd64
export HADOOP_HOME=/home/tangr/hadoop/hadoop-3.2.0
export PATH=$PATH:$HADOOP_HOME/bin

*Remember to change the highlighted part to your own user name in the Linux system.

Summary

Congratulations! Now you have successfully installed a single node Hadoop 3.2.0 cluster in your Ubuntu subsystem of Windows 10. It’s relatively easier as we don’t need to download or compile/build native Hadoop libraries.

BTW, subsystem is not a virtual machine however it provides you almost the same experience as you would have in a native Linux system.

Have fun!

Add comment

Comments (12)

Raym*** about 4 days ago

Thanks.  I also got the issue when I was writing the post. However I was able to resolve it by running the following two commands only:

sudo apt-get install ssh

sudo service ssh restart

Just in case other people cannot resolve the problem, I've updated the post to include the link to your comment so that they can follow that guide to resolve the connection issue.

Da*** about 5 days ago

This is a really cool blog.  I had the same issue with the ssh not working as others (connection closed), and was able to get past it by following the solution on this SE post: https://superuser.com/questions/1111591/how-can-i-ssh-into-bash-on-ubuntu-on-windows-10

Da*** about 5 days ago

This is a really cool blog.  I had the same issue with the ssh not working as others (connection closed), and was able to get past it by following the solution on this SE post: https://superuser.com/questions/1111591/how-can-i-ssh-into-bash-on-ubuntu-on-windows-10

Raym*** about 11 days ago

Have you tried the steps I mentioned in the post?

sudo apt-get install ssh

sudo service ssh restart

I'm not expert in network and I don't think the following solution will definitely help as they are all local traffics. I cannot reproduce this issue in my environment, so it will be hard to say where it goes wrong in your environment. 

There must be some other reasons that you cannot ssh localhost. For example, is port 22 used by your other programs? 

Can you also please try the same approach to allow ssh connections?

The websites won't start successfully until you resolve the ssh issue. So make sure you can ssh localhost first.

  • Protocol type: TCP
  • Local port: 22
  • Remote port: All Ports
  • Scope: make sure all your local IP addresses are added.
  • Profiles: Private. I'm choosing this one I will only connect to my wSL when connecting to private network. 

*** about 16 days ago

ssh localhost

Connection closed by 127.0.0.1 port 22


http://localhost:9870/dfshealth.html#tab-overview not working

Raym*** about 11 days ago

Have you tried the solution I mentioned in the post? I got the same issue when it is first installed but after the following commands, it work. And also make sure you stop and restart hadoop daemons. 

sudo apt-get install ssh

sudo service ssh restart

I'm not expert in network and I don't think the following solution will definitely help as they are all local traffics. There must be some other reasons that you cannot ssh localhost. For example, is port 22 used by your other programs?  Can you also use IPv4 addresses for localhost instead of the IPv6 one?

Can you try to add firewall rule to allow TCP traffic to ssh port 22?

  • Protocol type: TCP
  • Local port: 22
  • Remote port: All Ports
  • Scope: make sure all your local IP addresses are added.
  • Profiles: Private. I'm choosing this one I will only connect to my wSL when connecting to private network. 

jo*** about 19 days ago

I installed ssh and restarted it. Now 'ssh localhost' just says 'Connection closed by ::1 port 22.'

*** about 16 days ago

ssh localhost

Connection closed by 127.0.0.1 port 22


http://localhost:9870/dfshealth.html#tab-overview not working

jo*** about 19 days ago

I installed ssh and restarted it. Now 'ssh localhost' just says 'Connection closed by ::1 port 22.'

Raym*** about 2 months ago

You are welcome! I’m glad it helped.

Rudy*** about 2 months ago

This was a great and easy to follow post. THANK YOU!

Rudy*** about 2 months ago

This was a great and easy to follow post. THANK YOU!

Raym*** about 2 months ago

Hmm, sudo is actually not required if you install everything in your own folder though the better approach is to install them in other folders using root user (then sudo is required).

Mo*** about 2 months ago

I get Permission Denied when trying to get hadoop binary. after research I found that I need to use sudo in front of command. So need to use 

sudo wget http://mirrors.....


Thanks for great article!


Mo*** about 2 months ago

I get Permission Denied when trying to get hadoop binary. after research I found that I need to use sudo in front of command. So need to use 

sudo wget http://mirrors.....


Thanks for great article!


Raym*** about 2 months ago

Hi,

Most likely it is because ssh is not working. I found that each time when you restart your windows system, you need to re-run the command to restart ssh services:

sudo service ssh restart

Make sure you can ssh localhost successfully without a passphrase.

To make it easy, I have added command ‘sudo service ssh restart’ into my .bashrc file so that each time when I start WSL it will restart ssh to make sure it work.

Please stop all Hadoop services first and then restart the services:

sbin/stop-all.sh

sbin/start-all.sh

In Hadoop home folder there is a ‘logs’ folder created which includes the name node log file too. You can find the detailed the error message there.

Please provide your detailed error logs here if the above suggestions don’t work.

be*** about 2 months ago

In my case, the command: 

sbin/start-dfs.sh 

is executed without errors, but the NameNode is not started and therefore it is not responding on http://localhost:9870.

Executing jps command I can see how running processes are:

1) SecondaryNameNode

2) DataNode

3) Jps

NameNode process is missing from the returned list.


Any idea on what can be going wrong?

I followed all the instructions in this guide to configure my WSL environment.


Thanks 

be*** about 2 months ago

In my case, the command: 

sbin/start-dfs.sh 

is executed without errors, but the NameNode is not started and therefore it is not responding on http://localhost:9870.

Executing jps command I can see how running processes are:

1) SecondaryNameNode

2) DataNode

3) Jps

NameNode process is missing from the returned list.


Any idea on what can be going wrong?

I followed all the instructions in this guide to configure my WSL environment.


Thanks