In this chapter, we describe how to configure Opal so that it can submit jobs to an external scheduler, using the various job managers provided. We also provide a brief introduction on how to write your own job manager, if need be.
The prerequisite to this section is that you have a scheduler such as Condor or SGE installed on your compute cluster. Access to schedulers is provided in a number of ways - in particular, via the DRMAA API and the Globus Toolkit. Starting with release 2.1, we support Condor natively without using DRMAA or Globus, and also the Community Scheduler Framework (CSF4) for scheduling jobs across multiple clusters. In release 2.3, we introduced direct support for TORQUE/PBS without having to use DRMAA or Globus. Please consult the appropriate documentation for the installation of these scheduler tools.