3. Option III - Manual Install One Node At A Time

Use the following instructions to deploy Hadoop using HDP:

  1. Complete all the prerequisites provided here.

  2. Download the HDP for Windows MSI from here.

    [Note]Note

    Downloaded MSI includes full source, binary-only MSI, and documentation for all components. Hortonworks recommends using the binary-only MSI for faster downloads.

  3.  Use the instructions provided here to complete the configuration for HDP installer.

  4.  Install from MSI.

    1. Launch the MSI installer with the clusterproperties.txt file created previously.

      [Important]Important

      This MSI must be executed on each and every cluster node and must use the same clusterproperties.txt file.

    2. On each node, run a command prompt in Administrator mode (to use Administrator privileges), and execute the following command:

       msiexec /i "<$MSI_PATH>" /lv "<$PATH_to_Installer_Log_File>" HDP_LAYOUT="<$PATH_to_clusterproperties.txt_File>" HDP_DIR="<$PATH_to_HDP_Install_Dir>" DESTROY_DATA="<Yes_OR_No>"

      Ensure that you provide appropriate values for the following mandatory command line option:

      • HDP_LAYOUT: Mandatory parameter. Provide location of the clusterproperties.txt file (For example, d:\config\clusterproperties.txt).

        [Important]Important

        The path to the clusterproperties.txt file must be absolute. Relative paths will not work.

      Optionally, you can also use the following command line options:

      • HDP_DIR: Install directory for HDP (For example, d:\hdp). Default value is <$Default_Drive>/hdp.

      • DESTROY_DATA: Specifies whether to preserve or delete existing data in target data directories (allowed values are undefined(default), yes, and no).

        The DESTROY_DATA parameter takes care of the following conditions:

        • During installation, when HDP_DATA_DIR has data from previous installation if DESTROY_DATA is set to undefined, installation will fail.

        • During installation, if DESTROY_DATA is set to no, the installer will reuse the existing data and would not format the NameNode.

          [Note]Note

          Installer does not check for the data correctness.

        • During installation, if DESTROY_DATA is set to yes, installation will remove previous data and format the NameNode.

        • During installation, if no data exists in $HDP_DATA_DIR then the HDP_DATA_DIR is created irrespective of the value of DESTROY_DATA and NameNode is formatted.

      The following example illustrates the command to launch the MSI installer:

       msiexec /i "hdp-win-1.1.msi" /lv "hdp.log" HDP_LAYOUT="D:\config\clusterproperties.txt" HDP_DIR="D:\hdp\hadoop" DESTROY_DATA="no"
            
    3. (Optional): Configure compression for HDFS. Download the zlib1.dll from here. Copy the downloaded file to either $HADOOP_HOME\lib\native or to C:\Windows\System32. To use GzipCodec, ensure that you copy the downloaded file to C:\Windows\System32.

  5.  Validate the install.

    1. Use the instructions provided here to start the HDP Services.

    2. On a cluster node, open a command shell and execute the smoke test command script as shown below:

       
      cd %HADOOP_NODE_INSTALL_ROOT%
      Run-SmokeTests
                   

      The smoke tests validate the installed functionality by executing a set of tests for each HDP component.

    [Note]Note

    It is recommended to re-install HDP, if you see installation failures for any HDP component.