Grid_environment

How to Install PBS Based on DRBL

• 1. Introduction
  
  
  • TORQUE is an open source resource manager providing control over batch jobs and distributed compute nodes based on OpenPBS. Torque developed from OpenPBS, so the commands are the same.

• 2. Configure and Deploy Torque on the DRBL server and clients
  
  
  • At first, you can download ​torque-2.3.3.tar.gz and install and configure it by consulting ​Appendix L: TORQUE Quick Start Guide or​Globus Toolkit Tutorial_Deploying torque.
    
    
  • If you want to deploy Torque based on DRBL, you should re-configure data management on the compute nodes. PBS copies back output files at the end of the job.
    
    
    • By default, "copy back" is done using the pbs_rcp command which might not be supported or might not want to be supported on most clusters. If your home directories are mounted across cluster nodes, use the $usecp parameter to manage copy back.

      Edit /var/spool/torque/mom_priv/config and add one line.
      
            $usecp *:/home /home
      

  • Now you should deploy Torque to clients if you have already install and configure server successfully.
    
    
    • Default shared directory on each machine is Torque installation directory (Ex: /opt/pbs).
      
      
    • Each machine has its own directory: $PBS_HOME (/var/spool/torque). This directory stores the configuration files and logs for PBS daemons (Ex: pbs_server, pbs_sched, and pbs_mom) on each machine.

      On the DRBL_server:
      
            cp /var/spool/torque /tftpboot/nodes/[client_ip]/etc/ -rf
      

References

• ​TORQUE Resource Manager
• ​Appendix L: TORQUE Quick Start Guide
• ​Globus Toolkit Tutorial_Deploying torque