Step 1 - Download Hadoop binary package
Select download mirror link
Go to download page of the official website:
Apache Download Mirrors - Hadoop 3.2.1
And then choose one of the mirror link. The page lists the mirrors closest to you
based on your location. For me, I am choosing the following mirror link:
http://apache.mirror.digitalpacific.com.au/hadoop/common/hadoop-3.2.1/hadoop-
3.2.1.tar.gz
Download the package
I am installing Hadoop in folder big-data of my F drive (F:\big-data). If you prefer to
install on another drive, please remember to change the path accordingly in the
following command lines. This directory is also called destination directory in the
following sections.
Open PowerShell and then run the following command lines one by one:
$dest_dir="F:\big-data"
$url = "http://apache.mirror.digitalpacific.com.au/hadoop/common/hadoop-3.2.1/hadoop-
3.2.1.tar.gz"
$client = new-object System.Net.WebClient
$client.DownloadFile($url,$dest_dir+"\hadoop-3.2.1.tar.gz")
It may take a few minutes to download.
Once the download completes, you can verify it:
PS F:\big-data> cd $dest_dir
PS F:\big-data> ls
Directory: F:\big-data
Mode LastWriteTime Length Name
---- ------------- ------ ----
-a---- 18/01/2020 11:01 AM 359196911 hadoop-3.2.1.tar.gz
PS F:\big-data>
You can also directly download the package through your web browser and save it to
the destination directory.
Step 2 - Unpack the package
Now we need to unpack the downloaded package using GUI tool (like 7 Zip) or
command line. For me, I will use git bash to unpack it.
Open git bash and change the directory to the destination folder:
cd F:/big-data
And then run the following command to unzip:
tar -xvzf hadoop-3.2.1.tar.gz
The command will take quite a few minutes as there are numerous files included and
the latest version introduced many new features.
After the unzip command is completed, a new folder hadoop-3.2.1 is created under
the destination folder.
When running the command you will experience errors like the following:
tar: hadoop-3.2.1/lib/native/libhadoop.so: Cannot create symlink to ‘libhadoop.so.1.0.0’: No such
file or directory
Please ignore it for now as those native libraries are for Linux/UNIX and we will
create Windows native IO libraries in the following steps.
Step 3 - Install Hadoop native IO binary
Hadoop on Linux includes optional Native IO support. However Native IO is
mandatory on Windows and without it you will not be able to get your installation
working. The Windows native IO libraries are not included as part of Apache Hadoop
release. Thus we need to build and install it.
I also published another article with very detailed steps about how to compile and
build native Hadoop on Windows: Compile and Build Hadoop 3.2.1 on Windows 10
Guide.
The build may take about one hourand to save our time, we can just download the
binary package from github.
infoThe following repository already pre-built Hadoop Windows native libraries
for us:
https://github.com/cdarlint/winutils
warning These libraries are not signed and there is no guarantee that it is
100% safe. We use if purely for test&learn purpose.
Download all the files in the following location and save them to the bin folder under
Hadoop folder. For my environment, the full path is: F:\big-data\hadoop-3.2.1\bin.
Remember to change it to your own path accordingly.
https://github.com/cdarlint/winutils/tree/master/hadoop-3.2.1/bin
Alternatively, you can run the following commands in the previous PowerShell
window to download:
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
hadoop.dll",$dest_dir+"\hadoop-3.2.1\bin\"+"hadoop.dll")
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
hadoop.exp",$dest_dir+"\hadoop-3.2.1\bin\"+"hadoop.exp")
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
hadoop.lib",$dest_dir+"\hadoop-3.2.1\bin\"+"hadoop.lib")
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
hadoop.pdb",$dest_dir+"\hadoop-3.2.1\bin\"+"hadoop.pdb")
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
libwinutils.lib",$dest_dir+"\hadoop-3.2.1\bin\"+"libwinutils.lib")
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
winutils.exe",$dest_dir+"\hadoop-3.2.1\bin\"+"winutils.exe")
$client.DownloadFile("https://github.com/cdarlint/winutils/raw/master/hadoop-3.2.1/bin/
winutils.pdb",$dest_dir+"\hadoop-3.2.1\bin\"+"winutils.pdb")
After this, the bin folder looks like the following:
Step 4 - (Optional) Java JDK installation
Java JDK is required to run Hadoop. If you have not installed Java JDK please install
it.
You can install JDK 8 from the following page:
https://www.oracle.com/technetwork/java/javase/downloads/jdk8-downloads-
2133151.html
Once you complete the installation, please run the following command in PowerShell
or Git Bash to verify:
$ java -version
java version "1.8.0_161"
Java(TM) SE Runtime Environment (build 1.8.0_161-b12)
Java HotSpot(TM) 64-Bit Server VM (build 25.161-b12, mixed mode)
If you got error about 'cannot find java command or executable'. Don't worry we will
resolve this in the following step.
Step 5 - Configure environment variables
Now we've downloaded and unpacked all the artefacts we need to configure two
important environment variables.
Configure JAVA_HOME environment variable
As mentioned earlier, Hadoop requires Java and we need to
configure JAVA_HOME environment variable (though it is not mandatory but I
recommend it).
First, we need to find out the location of Java SDK. In my system, the path is: D:\
Java\jdk1.8.0_161.
Your location can be different depends on where you install your JDK.
And then run the following command in the previous PowerShell window:
SETX JAVA_HOME "D:\Java\jdk1.8.0_161"
Remember to quote the path especially if you have spaces in your JDK path.
You can setup evironment variable at system by adding option /M however just in
case you don't have access to change system variables, you can just set it up at
user level.
The output looks like the following:
Configure HADOOP_HOME environment variable
Similarly we need to create a new environment variable for HADOOP_HOME using
the following command. The path should be your extracted Hadoop folder. For my
environment it is: F:\big-data\hadoop-3.2.1.
If you used PowerShell to download and if the window is still open, you can simply
run the following command:
SETX HADOOP_HOME $dest_dir+"/hadoop-3.2.1"
The output looks like the following screenshot:
Alternatively, you can specify the full path:
SETX HADOOP_HOME "F:\big-data\hadoop-3.2.1"
Now you can also verify the two environment variables in the system:
Configure PATH environment variable
Once we finish setting up the above two environment variables, we need to add
the bin folders to the PATH environment variable.
If PATH environment exists in your system, you can also manually add the following
two paths to it:
%JAVA_HOME%/bin
%HADOOP_HOME%/bin
Alternatively, you can run the following command to add them:
setx PATH "$env:PATH;$env:JAVA_HOME/bin;$env:HADOOP_HOME/bin"
If you don't have other user variables setup in the system, you can also directly add
a Path environment variable that references others to make it short:
Close PowerShell window and open a new one and type winutils.exe directly to verify that
our above steps are completed successfully:
You should also be able to run the following command:
hadoop -version
java version "1.8.0_161"
Java(TM) SE Runtime Environment (build 1.8.0_161-b12)
Java HotSpot(TM) 64-Bit Server VM (build 25.161-b12, mixed mode)
Step 6 - Configure Hadoop
Now we are ready to configure the most important part - Hadoop configurations
which involves Core, YARN, MapReduce, HDFS configurations.
Configure core site
Edit file core-site.xml in %HADOOP_HOME%\etc\hadoop folder. For my
environment, the actual path is F:\big-data\hadoop-3.2.1\etc\hadoop.
Replace configuration element with the following:
<configuration>
<property>
<name>fs.default.name</name>
<value>hdfs://0.0.0.0:19000</value>
</property>
</configuration>
Configure HDFS
Edit file hdfs-site.xml in %HADOOP_HOME%\etc\hadoop folder.
Before editing, please correct two folders in your system: one for namenode
directory and another for data directory. For my system, I created the following two
sub folders:
F:\big-data\data\dfs\namespace_logs
F:\big-data\data\dfs\data
Replace configuration element with the following (remember to replace
the highlighted paths accordingly):
<configuration>
<property>
<name>dfs.replication</name>
<value>1</value>
</property>
<property>
<name>dfs.namenode.name.dir</name>
<value>file:///F:/big-data/data/dfs/namespace_logs</value>
</property>
<property>
<name>dfs.datanode.data.dir</name>
<value>file:///F:/big-data/data/dfs/data</value>
</property>
</configuration>
In Hadoop 3, the property names are slightly different from previous version. Refer to
the following official documentation to learn more about the configuration properties:
Hadoop 3.2.1 hdfs_default.xml
For DFS replication we configure it as one as we are configuring just one single
node. By default the value is 3.
The directory configuration are not mandatory and by default it will use Hadoop
temporary folder. For our tutorial purpose, I would recommend customise the
values.
Configure MapReduce and YARN site
Edit file mapred-site.xml in %HADOOP_HOME%\etc\hadoop folder.
Replace configuration element with the following:
<configuration>
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
<property>
<name>mapreduce.application.classpath</name>
<value>%HADOOP_HOME%/share/hadoop/mapreduce/*,%HADOOP_HOME%/share/
hadoop/mapreduce/lib/*,%HADOOP_HOME%/share/hadoop/common/*,%HADOOP_HOME%/
share/hadoop/common/lib/*,%HADOOP_HOME%/share/hadoop/yarn/*,%HADOOP_HOME%/
share/hadoop/yarn/lib/*,%HADOOP_HOME%/share/hadoop/hdfs/*,%HADOOP_HOME%/
share/hadoop/hdfs/lib/*</value>
</property>
</configuration>
Edit file yarn-site.xml in %HADOOP_HOME%\etc\hadoop folder.
<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_H
OME</value>
</property>
</configuration>
Step 7 - Initialise HDFS & bug fix
Run the following command in Command Prompt
hdfs namenode -format
This command failed with the following error and we need to fix it:
2020-01-18 13:36:03,021 ERROR namenode.NameNode: Failed to start namenode.
java.lang.UnsupportedOperationException
at java.nio.file.Files.setPosixFilePermissions(Files.java:2044)
at
org.apache.hadoop.hdfs.server.common.Storage$StorageDirectory.clearDirectory(Storage.java:
452)
at org.apache.hadoop.hdfs.server.namenode.NNStorage.format(NNStorage.java:591)
at org.apache.hadoop.hdfs.server.namenode.NNStorage.format(NNStorage.java:613)
at org.apache.hadoop.hdfs.server.namenode.FSImage.format(FSImage.java:188)
at org.apache.hadoop.hdfs.server.namenode.NameNode.format(NameNode.java:1206)
at
org.apache.hadoop.hdfs.server.namenode.NameNode.createNameNode(NameNode.java:1649)
at org.apache.hadoop.hdfs.server.namenode.NameNode.main(NameNode.java:1759)
2020-01-18 13:36:03,025 INFO util.ExitUtil: Exiting with status 1:
java.lang.UnsupportedOperationException
Refer to the following sub section (About 3.2.1 HDFS bug on Windows) about
the details of fixing this problem.
Once this is fixed, the format command (hdfs namenode -format) will show
something like the following:
About 3.2.1 HDFS bug on Windows
This is a bug with 3.2.1 release:
https://issues.apache.org/jira/browse/HDFS-14890
It will be resolved in version 3.2.2 and 3.3.0.
We can apply a temporary fix as the following change diff shows:
Code fix for HDFS-14890
I've done the following to get this temporarily fixed before 3.2.2/3.3.0 is released:
Checkout the source code of Hadoop project from GitHub.
Checkout branch 3.2.1
Open pom file of hadoop-hdfs project
Update class StorageDirectory as described in the above code diff screen
shot:
if (permission != null) {
try {
Set<PosixFilePermission> permissions =
PosixFilePermissions.fromString(permission.toString());
Files.setPosixFilePermissions(curDir.toPath(), permissions);
} catch (UnsupportedOperationException uoe) {
// Default to FileUtil for non posix file systems
FileUtil.setPermission(curDir, permission);
}
}
Use Maven to rebuild this project as the following screenshot shows:
Fix bug HDFS-14890
I've uploaded the JAR file into the following location. Please download it from the
following link:
https://github.com/FahaoTang/big-data/blob/master/hadoop-hdfs-3.2.1.jar
And then rename the file name hadoop-hdfs-3.2.1.jar to hadoop-hdfs-3.2.1.bk in
folder %HADOOP_HOME%\share\hadoop\hdfs.
Copy the downloaded hadoop-hdfs-3.2.1.jar to folder %HADOOP_HOME%\share\
hadoop\hdfs.
This is just a temporary fix before the official improvement is published. I publish it
purely for us to complete the whole installation process and there is no guarantee
this temporary fix won't cause any new issue.
Refer to this article for more details about how to build a native Windows
Hadoop: Compile and Build Hadoop 3.2.1 on Windows 10 Guide.
Step 8 - Start HDFS daemons
Run the following command to start HDFS daemons in Command Prompt:
%HADOOP_HOME%\sbin\start-dfs.cmd
Two Command Prompt windows will open: one for datanode and another for
namenode as the following screenshot shows:
Step 9 - Start YARN daemons
You may encounter permission issues if you start YARN daemons using normal
user. To ensure you don't encounter any issues. Please open a Command Prompt
window using Run as administrator.
Alternatively, you can follow this comment on this page which doesn't require
Administrator permission using a local Windows account:
https://kontext.tech/column/hadoop/377/latest-hadoop-321-installation-on-windows-
10-step-by-step-guide#comment314
Run the following command in an elevated Command Prompt window (Run as
administrator) to start YARN daemons:
%HADOOP_HOME%\sbin\start-yarn.cmd
Similarly two Command Prompt windows will open: one for resource manager and
another for node manager as the following screenshot shows:
Step 10 - Useful Web portals exploration
The daemons also host websites that provide useful information about the cluster.
HDFS Namenode information UI
http://localhost:9870/dfshealth.html#tab-overview
The website looks like the following screenshot:
HDFS Datanode information UI
http://localhost:9864/datanode.html
The website looks like the following screenshot:
YARN resource manager UI
http://localhost:8088
The website looks like the following screenshot:
Step 11 - Shutdown YARN & HDFS daemons
You don't need to keep the services running all the time. You can stop them by
running the following commands one by one:
%HADOOP_HOME%\sbin\stop-yarn.cmd
%HADOOP_HOME%\sbin\stop-dfs.cmd
Congratulations! You've successfully completed the installation of Hadoop 3.2.1 on
Windows 10.
Let me know if you encounter any issues. Enjoy with your latest Hadoop on Windows
10.
