4.11. Using Pre- and Post- exec commands

In some cases, applications (and thus CUs) need more complex and customizable setup routines than can be expressed via environment or MPI flags. A frequent example is the use of module load commands on various HPC resources, which are used to prepare application runtime environments in a well defined, system specific way.

RP supports the invocation of such commands via the pre_exec and post_exec keys for the CU descriptions.


Pre- and Post- execution is performed on the resource headnode – abuse of these commands for any compute or I/O heavy loads can lead to serious consequences, and will likely draw the wrath of the system administrators upon you! You have been warned…

The code example below exemplarily demonstrates the same environment setup we have been using in an earlier section, but now rendered via an pre_exec command:

cud = rp.ComputeUnitDescription()

cud.pre_exec    = ['export TEST=jabberwocky']
cud.executable  = '/bin/echo'
cud.arguments   = ['$RP_UNIT_ID greets $TEST']

which again will make the environment variable TEST available during CU execution.

4.11.1. Running the Example

10_pre_and_post_exec.py. uses the above blob to run a bag of echo commands:


4.11.2. What’s Next?

The RP User-Guide concludes with this section. We recommend to check out the RP API documentation next, and use it to write an RP application to run your own workload. Ii is easiest to do so by starting off with the canonical example, and then add bits and pieces from the various user :ref:<chapter_user_guide>` sections as needed.