Globus Toolkit 3.2: Installation Guide

Overview
Before You Begin
Support Software
Installing GT 3.2
Configuring > WS GRAM <
Testing
Troubleshooting

Configuring WS GRAM - Basic

MMJFS (also known as the JobManager) is a feature of the WS GRAM component of the Globus Toolkit. Although it must be installed with the Globus Toolkit, MMJFS is an option that only needs to be configured if you want to use WS GRAM.

There is only one step you must perform in order to make the WS GRAM service available for testing and other very basic purposes. For more advanced configurations, see the WS GRAM System Administrator's Guide.

Requirements
  • You should be familiar with the Globus Toolkit 3.2 Security Concepts
  • You must already have installed the Globus Toolkit 3.2
  • You must have completed the required configuration which provides the following:
    • host certificate
    • user account with a user certificate
    • trusted CA certificate in /etc/grid-security/certificates
    • setuid globus-grim executable (This is what you get after running setperms.sh as root)
    • A grid-mapfile in /etc/grid-security/grid-mapfile
Testing GRAM via the command line

Start the container:
1

To start the container as the globus user:

globus$ bin/globus-start-container

It will print out a list of services. One of them should look like:

http://127.0.0.1:8080/ogsa/services/base/gram/MasterForkManagedJobFactoryService

Now that the container is running, you can submit a job through the GRAM component.
2

To submit a test GRAM job as a user:

user$ source $GLOBUS_LOCATION/etc/globus-user-env.{sh,csh}
user$ grid-proxy-init
user$ bin/managed-job-globusrun -factory
  http://localhost:8080/ogsa/services/base/gram/MasterForkManagedJobFactoryService
  -file schema/base/gram/examples/test.xml

This test is successful if the output looks something like this:

========== Status Notification ==========
Job Status: StageIn
=========================================
========== Status Notification ==========
Job Status: active
=========================================
========== Status Notification ==========
Job Status: done
=========================================
DESTROYING SERVICE
If you have a scheduler...
3

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

$ gpt-build scheduler-[name]-3.2-src_bundle.tar.gz gcc32dbg
$ gpt-postinstall
If you need to customize the jobmanager for your scheduler type...
4

Edit the following file to affect how jobs are submitted.

$GLOBUS_LOCATION/lib/perl/Globus/GRAM/JobManager/[scheduler].pm

For example, some PBS clusters use rsh, and some use ssh. You may need to perform this step for pbs.pm to change the remote shell between rsh and ssh.

For more information, see the WS GRAM System Administrator's Guide.