Installation

Things you should consider before installing the VDT

If you want to do a quick trial installation of the VDT to see how things work, you do not need to worry about these yet. The defaults in the VDT will work fine for small VDT installations.

• Would you like to install as root or non-root?
  • You can install as any user, but we recommend installing as root (as it will simplify installation and post-configuration). However, you can always install as non-root and perform all the necessary subsequent re-configuration as root later.
• Where would you like to install CA files? (details)
• Do you want to install the complete VDT or some subset of it? (details)
• Thing twice about installing on NFS (details)
• Do you wish to use an existing external installation of Condor? (details)
• Would you like an automatic (non-interactive) installation? (details)
• Create a condor user (recommended if using Condor as a batch system)
  • It is common to set up a Condor cluster with a condor user. The VDT installation will not do this, because of concerns some VDT users have had about the VDT installation messing with an existing Condor installation. If you like, you can read about it in the Condor manual
• RLS has been updated and has a new schema. Do you need to update it? If you don't know, then you don't need to update it. (details)

Installation Instructions

1. Install Pacman:
   • Download Pacman This is version 2.116, which is the most recent version tested with VDT 1.1.13. For installing Redhat 9 specific packages you need to install version 2.116 or above. More recent versions of Pacman and additional information about Pacman may be available from the main Pacman web site.
   • Unpack Pacman:

     tar xzf pacman-2.116.tar.gz

   • Setup your environment to use pacman:
     

     cd pacman-2.116
     source setup.csh 

     This will set your PATH environment variable to find Pacman.
2. Prepare to install the VDT:
   • Decide where you want to install it. It doesn't matter what you choose. Common choices are /vdt and /opt/vdt. For the examples that follow, we'll assume that you wish to install into /vdt (this will also be referred to as $VDT_LOCATION)
   • Create the directory and change into it:

     mkdir /vdt
     cd /vdt

3. Install the VDT:
   • If your system is running behind a Web proxy, tell Pacman about it:
     

     pacman -http_proxy:http://someproxy.com:port

     Of course, subsitute the name and port number of your proxy.
   • Tell Pacman where to find the VDT 1.1.13 software cache:

     pacman -cache:http://vdt.cs.wisc.edu/vdt_1113_cache

   • Install VDT:
     

     pacman -get VDT

     Pacman will be asked two questions before anything can begin:

     • Do you want to start a new installation here? This is your chance to back out if you are installing in the wrong directory.
     • Do you want to trust the registered cache [VDT]? Choose "yes"; otherwise you won't be able to install the VDT.

     After you answer the preliminary questions, you will see some output as Pacman fetches the necessary files. Then, before the rest installation of these files begins, Pacman will ask you to agree to some licenses:

     VDT 1.1.13 installs a variety of software, each with its own license.
     In order to continue, you must agree to the licenses.
     You can view the licenses online at:
           
          http://vdt.cs.wisc.edu/licenses/1.1.13
           
     After the installation has completed, you will also be able to 
     view the licenses in the 'licenses' directory.
           
     Do you agree to the licenses? (y/n)

     You must answer y (presumably after reading the licences) in order to continue with the installation.
     
     You will then be asked a series of other questions on how to configure your VDT installation. If you're not sure how to answer a question or if it is not relevant to you, choose skip. You can always configure the relevant item(s) later. The instructions for configuring them will be placed in $VDT_LOCATION/post-install/README. If you are doing hands-free installations and you want to avoid answering these questions during the installation, you can answer them in advance.
   
   
4. See what happened:
   • Do an 'ls' command, you should see a number of directories with Globus, Condor, and the rest of the software installed.
   • Read through /vdt/post-install/README. It will tell you about anything that the VDT installation process was unable to do. Also, the Condor set up is tricky, and VDT installer may have had to make some guesses about how to configure Condor--these will be noted in this same README. You should be able to use a personal Condor or Condor-G with no changes, but if you want to set up a Condor pool, you may need to edit the Condor Configuration. The post-install README will get you started with that process.
   • If you let the VDT installer set up the daemons, it edited /etc/services, either /etc/inetd.conf (RedHat 6.2) or /etc/xinetd.d/ (RedHat 7.2), and /etc/rc.d/init.d. Look at the changes, if you wish.

Things you might want to do after a VDT installation

• source $VDT_LOCATION/setup.sh
  • This will set all the appropriate environment variables used by VDT components
  • You should do this before using any VDT components (Globus, Condor, MonaLisa, etc)
  • Use setup.sh / setup.csh / setup.ksh based on your shell
  • Be careful about modifying these files. Any subsequent Pacman installation will erase your changes
• Setup Globus services to run via xinetd
  • If you installed as root, you were given a chance to setup the Globus gatekeeper, GridFTP, and MDS to run from xinetd, which is the recommended way. If you want to change which of these service are run from xinetd, use the configure_globus script to enable/disable these componetns in xinetd.
• Configure Condor
• The VDT did the best it could to configure Condor and Condor-G for you, but you may wish to change it. You can use the condor_configure script.
• By default, Condor is configured to run as a personal Condor, i.e. a one-node pool which runs jobs regardless of whether the user is using the machine or not. You might wish to change that.
• You might want to configure your Condor installation to join your existing Condor pool. Generally this is done with condor_configure --central-manager=... command. See condor_configure for more info
• Enable submitting to other batch systems (PBS, LSF, FBSng)
  • You can setup Globus to submit jobs to PBS, LSF, and FBSng easily. It is a two-step process. First, make sure that that PBS or LSF command-line utilities are in your path. (You should be able to execute them if you wish, without using a complete pathname.) Next, install the appropriate package:

    pacman -get Globus-PBS-Setup
        or
    pacman -get Globus-LSF-Setup
       or
    pacman -get FBSNG

  • Interface to Condor is set up automatically by default.
  
  
• Get certificates:
  • You generally need a host certificate for your machine and user certificate(s) for all individual user(s) of Grid services.
  • See Requesting Certificates for more information
• Configure Globus Gridmap file
  • Located in /etc/grid-security/grid-mapfile
  • Essentially it is a mapping of the user's GSI certificate subject to a concrete user account existing on your machine. This serves a dual purpose of allowing remote user(s) to use your resource, and determining which accout the remote user's processes will run as.
  • Example of contents of grid-mapfile:

    "/C=US/O=Globus/O=University of Wisconsin/OU=Computer Sciences Department/CN=Carey Kireyev" ckireyev
    "/DC=org/DC=doegrids/OU=People/CN=Alain Roy 424511" roy
    

  • See Globus mapfile documentation for more info.
• Get more information:
  • Globus documentation:
    • User and Administrator Guides for Globus 2.4.3
    • FAQ
    • Common globus errors
  • Condor manual
• Install other software
  • RLS: The RLS client and server are installed by default in the VDT, but none of the additional support software for the server is installed. For your convenience MySQL, IODBC and MyODBC are installed and configured as part of separate package. MySQL server prefers to run as user "mysql". If you do not want pacman to make any changes to /etc/passwd please create user and group for mysql prior to installing the Globus-RLS-Server-Setup-MySQL package.
    

    You can find more documentation on RLS at the Globus web site.

    To install RLS server as user root do:

    pacman -get Globus-RLS-Server-Setup-MySQL

  • Pippy:

    pacman -get UTA:pippy

    Pippy will publish information about software installed via Pacman to your GRIS. At some point this may be part of the standard VDT installation, but first it needs to undergo some testing by the VDT team. Currently it requires you to be root in order to install.