Reading notes about DRMAA v2

The DRMAA v2 specification draft is ready to be published, and is in public comment until 31st July this year. I used DRMAA v1 to integrate Jenkins and PBS some time ago, but it was not a very elegant solution.

And in the end integrating other grid computing implementations like SGE would not be very simple.

This post contains my reading notes for DRMAA v2, and a short analysis of how this new specification could be used in a new tentative to integrate Jenkins and several grid computing implementations in a single plug-in.

Notes about the new specification

DRMAA Job States
DRMAA Job States

never report it* Important to include in the plug-in documentation, and also troubleshooting page. As jobs could be considered not found, but simply because of its state * Question: is UNDETERMINED state assigned to a job that does not exist, or only to a job that exists but state cannot be determined?

How a DRMAA v2 Java implementation could be used with Jenkins

  1. The Eclipse Science project Triquetrum is working on a Java implementation that could possibly be re-used in BioUno to integrate DRM systems with Jenkins.
  2. The new monitoring methods - if widely used by implementers - could replace most of the code that was written to parse the output of the response of qsub/qstat or even DRMAA v1 methods
  3. Setting up nodes in Jenkins could be simplified by querying the server capabilities

Time to go back to the blackboard, but in the meantime, will start a development cycle for Jenkins plug-ins, and will work with the EXISTS W3C working group for a while.

What I was listening while writing this blog post: Paola Pequena - Sonata herĂ³ica

Categories: Blog

Tags: Programming, Hpc