Installing PXF

The Greenplum Platform Extension Framework (PXF) for Greenplum Database 5.21.2+ and 6.x for CentOS 7.x and RHEL 7.x platforms is available from a separately-downloadable package on Pivotal Network.

The PXF download package is an .rpm file that installs PXF libraries, executables, and script files on a Greenplum Database host.

When you install PXF, you will:

  1. Satisfy the prerequisites.
  2. Download the PXF package.
  3. Install the PXF .rpm on every host in your Greenplum Database cluster.
  4. Check out Next Steps for post-install topics.

Prerequisites

The recommended deployment model is to install PXF on all Greenplum Database hosts. Before you install PXF, ensure that you meet the following prerequisites:

  • Greenplum version 5.21.2 or later or 6.x is installed in the cluster.
  • You have access to all hosts (master, standby master, and segment hosts) in your Greenplum Database cluster.
  • You must be an operating system superuser, or have sudo privileges, to install the PXF package. You can install the package in the default location /usr/local/pxf-gp<greenplum-major-version>. Or, you can choose to install the package into a custom location by specifying the --prefix option to the install command.
  • You have installed Java 8 or 11 on all Greenplum Database hosts as described in Installing Java for PXF.
  • You can identify the operating system user that will own the PXF installation. This user must be the same user that owns the Greenplum Database installation, or a user that has write privileges to the Greenplum Database installation directory.
  • If you have previously configured and are using PXF in your Greenplum installation:
    1. Identify and note the current PXF version number.
    2. Stop PXF as described in Stopping PXF.
If this is your first installation of a PXF package, and the $GPHOME/pxf directory exists in your Greenplum installation, you may choose to remove the directory on all Greenplum hosts after you confirm that you have installed and configured PXF correctly and that it is working as expected.

If you choose to remove this directory, you may encounter warning: <pxf-filename>: remove failed: No such file or directory messages when you upgrade Greenplum. You can ignore these warnings for PXF files.

Downloading the PXF Package

PXF is available as a separate download for Greenplum Database 5.x or 6.x for CentOS 7.x and RHEL 7.x platforms from Pivotal Network:

  1. Download the package by navigating to Pivotal Network and locating and selecting the Release Download directory named Greenplum Platform Extension Framework.

    The format of the PXF download filename is pxf-gp<greenplum-major-version>-<pxf-version>-<pkg-version>.<platform>.<file_type>. For example:

    pxf-gp6-5.15.0-2.el7.x86_64.rpm
    
  2. Make note of the directory to which the file was downloaded.

Installing the PXF Package

You must install the PXF package on the Greenplum Database master and standby master hosts, and on each segment host.

If you installed an older version of the PXF package on your hosts, installing a newer package removes the existing PXF installation, and installs the new version.

The install procedure follows:

  1. Locate the installer file that you downloaded from Pivotal Network.

  2. Create a text file that lists your Greenplum Database standby master host and segment hosts, one host name per line. For example, a file named gphostfile may include:

    gpmaster
    mstandby
    seghost1
    seghost2
    seghost3
    
  3. Copy the PXF .rpm file to all hosts in your Greenplum cluster. For example, to copy the files to the /tmp directory on each host:

    gphost$ gpscp -f gphostfile pxf-gp6-5.15.0-2.el7.x86_64.rpm =:/tmp/
    
  4. Install the package on each Greenplum Database host using your package management utility. If a previous installation of PXF exists for the same Greenplum version, the files and runtime directories from the older version are removed before the current package is installed.

    1. To install PXF into the default location on all Greenplum hosts:

      gphost$ gpssh -e -v -f gphostfile "sudo rpm -Uvh /tmp/pxf-gp6-5.15.0-2.el7.x86_64.rpm"
      

      When you install the PXF .rpm package to the default location, PXF is installed to /usr/local/pxf-gp<greenplum-major-version>.

    2. To install PXF into a custom location on all Greenplum hosts:

      gpadmin@gphost$ gpssh -e -v -f gphostfile "sudo rpm -Uvh --prefix <install-location> pxf-gp6-5.15.0-2.el7.x86_64.rpm"
      
  5. Set the ownership and permissions of the PXF installation files to enable access by the gpadmin user. For example, if you installed PXF to the default location:

    gphost$ gpssh -e -v -f gphostfile "sudo chown -R gpadmin:gpadmin /usr/local/pxf-gp*"
    

    If you installed PXF to a custom <install-location>, specify that location in the command.

  6. (Optional) Add the PXF bin directory to the PXF owner’s $PATH. For example, if you installed PXF for Greenplum 6 in the default location, you could add the following text to the .bashrc shell initialization script for the gpadmin user:

    export PATH=$PATH:/usr/local/pxf-gp6/bin
    

    Be sure to remove any previously-added $PATH entries for PXF in $GPHOME/pxf/bin.

  7. Remove the PXF .rpm file that you copied to each system. For example, to remove the file from /tmp:

    gpadmin@gphost$ gpssh -e -v -f gphostfile "rm -f /tmp/pxf-gp6-5.15.0-2.el7.x86_64.rpm"
    

Next Steps

PXF is not active after installation. You must explicitly initialize and start the PXF server before you can use PXF.

  • See About the PXF Installation and Configuration Directories for a list and description of important PXF files and directories.
  • If this is your first time using PXF, review Configuring PXF for a description of the initialization and configuration procedures that you must perform before you can use PXF.
  • If you installed the PXF .rpm as part of a Greenplum Database upgrade procedure, return to those upgrade instructions.
  • If you installed the PXF .rpm into a Greenplum cluster in which you had already configured and were using PXF, you may be required to perform some upgrade actions. Recall the original version of PXF (before you installed the .rpm), and perform Step 2 of the PXF upgrade procedure.