urgent computing topline

 

NCSA Mercury


Cluster details

This cluster runs Torque along with Moab as the default job manager/scheduler combination. You can run urgent jobs on these resources using either the 'spruce_sub' or the Globus extensions.

Policy

NCSA Mercury policy map is as follows -

red - 'next-to-run'
orange - 'next-to-run'
yellow - 'next-to-run'

Initial Setup

  • 1) Login to the machine- ssh -A username@tg-login.ncsa.teragrid.org or ssh -A username@login-hg.ncsa.teragrid.org
  • 2) Check if the variable $TG_COMMUNITY is set. This will be the base of the SPRUCE community space you would need to access. If not set, please set it to /usr/projects or contact us for further instructions.
  • 3) Compile your codes and have them handy in your home directory tree.

spruce_sub

Invoke the spruce_sub command with the urgency and job script specified. The script can be found here -

$TG_COMMUNITY/tg-community/spruce/bin

More about running jobs using spruce_sub can be found here.

Globus submissions

SPRUCE jobs can be submitted as any other Globus job. An additional urgency parameter and the right contact string are all that needed. The urgency level can take three values -
(urgency=yellow/orange/red)
and
the contact string is -
grid-hg.ncsa.teragrid.org/jobmanager-spruce

More about running jobs using Globus interface can be found here.

Additional Notes

On NCSA Mercury the SPRUCE filter checks are implemented in Moab as opposed to Torque, so there is an additional step needed to find out the status of your jobs. Once you submit the job, an jobID is returned. Inorder to see if the job was submitted to the scheduler or not, do -

checkjob -v jobID

There might be a delay sometimes for these details to show up, where you might see an 'cannot locate job ID' message, but keep trying. We are trying to fix this as soon as possible to make the results more intuitive.