ABAQUS is a finite element analysis and multiphysics suite for simulation.
To show the software versions:
module avail abaqus
Loading the module will set the path and any necessary
module load abaqus/version
Abaqus is a licensed product and may not be available to all CCR users.
Abaqus has built-in support for running batch jobs across
multiple nodes - an example script for doing this may be found on
the CCR front-end:
Abaqus manuals are also available on the CCR front-end in pdf
format and may be viewed using xpdf. Manuals are located
at the following path, where the * is a placeholder for the
desired version number:
Jobs run on CCR's general-compute nodes are limited to a maximum walltime of 72 hours. Users who anticipate running Abaqus jobs that require longer than 72 hours should make use of the built-in Abaqus restart/recover functionality. The tips and discussion given here may be helpful. Also, details on abaqus restart and recover features may be found in chapter 9 of book 2 (ANALYSIS_2) of the user manuals.
Abaqus supports running parametric studies using python scripts. Such studies can be run in distributed mode on the CCR cluster. An example slurm script and abaqus .psf file for doing this may be found on the CCR front-end:
To customize to your application, copy and then modify the
(1) tutorial.psf --- this is the python code that drives the parametric study.
Two lines are key:
--- Generates the input files required for the study.
--- A driver that processes the input files in distributed mode.
It replaces the execute(ALL) command.
Any other valid python code may be added to the file, just make sure to call
generate and invoke abaqus_parstudy rather than calling execute(ALL).
(2) slurm-abaqus-parametric-study --- This is the slurm script. When doing a
parametric study use this as the example instead of slurm-abaqus.
When a parametric study completes be sure to examine the .var output file to verify that all tasks completed successfully. If not, you'll need to redo the ones that are marked FAILED. This might happen if study tasks use up all of the memory on a node. Such failures can be corrected by asking for more memory (#SBATCH --mem) or fewer tasks per node (#SBATCH --ntasks-per-node).