This information is for a release that is no longer supported by the Globus Toolkit. The currently supported versions of the Globus Toolkit are 4.2 (recommended) and 4.0.

WS GRAM: System Administrator's Guide

For basic configuration information, see the Installing WS GRAM - Basic topic of the Installation Guide.

This page describes how to configure a WS GRAM service for submitting jobs to a batch scheduler.

Before You Begin

This guide will be in addition to the GT admin guide at:
http://www-unix.globus.org/toolkit/docs/3.2/installation/index.html

  • You should already have a properly configured and working GT 3.2 installation with MMJFS fork already installed. See the Installing WS GRAM - Basic topic of the Installation Guide for more information.
  • You should be able to submit a job to the fork service using managed-job-globusrun.

    For example,
 % managed-job-globusrun -factory <mmjfs fork factory> -file schema/base/gram/examples/test.xml
  • You should already have a properly configured and working scheduler installation.

    For example (using LSF), you should be able to:
    • submit a job using BSUB
    • monitor a job using BJOBS
    • cancel a job using BDEL

Installing a Batch Scheduler

Find the scheduler bundle from the installer you downloaded (look under /scheduler in your installer directory) and install it by running:

% gpt-build scheduler-pbs-3.2-src_bundle.tar.gz gcc32dbg

Configuring a Batch Scheduler (default)

If all the default configuration options are acceptable, then simply run:

% gpt-postinstall

And now test the installation by running:

% managed-job-globusrun -factory http://<your host>:<your port>/ogsa/services/base/gram/Master
<scheduler name>ManagedJobFactoryService> file schema/base/gram/examples/test.xml

Configuring a Batch Scheduler (non-default)

First configure all setup packages for the defaults, run:

gpt-postinstall

Then run the individual setup package(s) by hand, which will replace the default configuration.

A scheduler bundle contains 4 individual packages:

mmjfs_lsf_setup

Creates $GLOBUS_LOCATION/setup/globus/setup-mmjfs-lsf.

Contains the configuration information to add the MasterLsfManagedJobFactoryService to the GT installation.
mjs_lsf_setup

Creates $GLOBUS_LOCATION/setup/globus/setup-mjs-lsf.

Contains the configuration information to add the LsfManagedJobService to the GT installation.
rips_lsf_provider_setup

Creates $GLOBUS_LOCATION/setup/globus/setup-lsf-provider.

Contains the Borne Shell code that queries LSF for all jobs and reports the information to RIPS for processing.
globus_gram_job_manager_setup_lsf

Creates $GLOBUS_LOCATION/setup/globus/setup-globus-job-manager-pbs.

Contains the Perl code that interacts with LSF. Translates the RSL to an LSF job.

In GT 3.2, the job manager setup package is the only one that takes arguments to change the default configuration.

For PBS installations, most people need to set the remote shell to rsh instead of ssh by running:

% cd $GLOBUS_LOCATION/setup/globus
% ./setup-globus-job-manager-pbs --remote-shell=rsh

To see all configuration options, run:

% ./setup-globus-job-manager-pbs -help

Updating a Master Hosting Environment (MHE) and User Hosting Environment (UHE) configuration

After successfully running the gpt-postinstall and/or individual setup packages, your GT installation is now properly configured in order to submit jobs to the MMJFS in the Master Hosting Environment (MHE).

Important: The MHE needs to be restarted in order to have the new configuration take effect.

Additionally, any users that had submitted a job to the MHE prior to the configuration change will have an outdated copy of the configuration in their User Hosting Environment (UHE):

~user/.globus/uhe-<host>

To propogate the new configuration, remove the old UHE directory:

% rm -Rf $HOME/.globus/uhe-*/

Troubleshooting

Information coming soon .