abaqus
Abaqus is a software suite for finite element analysis and computer aided engineering.
1. How to Access the Software
Load the module and see the version informations:
$> module load TECHNICS # which abaqus versions are available? $> module avail abaqus $> module load abaqus # or load specific version $> module load abaqus/2018 $> abaqus information=release |
2. Example Batch Scripts
2.1 single node job
#!/usr/local_rwth/bin/zsh ### Job name #SBATCH --job-name=abaqus_slurm_job ### File/Path where STDOUT will be written to, %J is the job id #SBATCH --output abaqus-job-log.%J ### Request the time you need for execution. The full format is D-HH:MM:SS ### You must at least specify minutes or days and hours and may add or ### leave out any other parameters #SBATCH --time=5:00 ### Request the memory you need for your job. You can specify this ### in either MB (1024M) or GB (4G). BEWARE: This is a per-cpu limit, ### and will be multiplied with cpus-per-task for the total requested memory #SBATCH --mem-per-cpu=3900M ### Request one host #SBATCH --nodes=1 ### Request number of CPUs/MPI Ranks #SBATCH --ntasks=4 ### Initialization of the software module load TECHNICS module load abaqus ### Set the amount of memory to be passed to Abaqus as a command line argument ### Beware: This HAS to be lower than the value you requested via --mem-per-cpu export ABAQUS_MEM_ARG="3584 mb" ### Change (!) to your desired work directory cd $HOME/path/to/your/dir ### Create ABAQUS environment file for current job, you can set/add your own options (Python syntax) env_file=abaqus_v6.env cat << EOF > ${env_file} #verbose = 3 #ask_delete = OFF #usub_lib_dir=os.getcwd() mp_file_system = (SHARED, LOCAL) mp_host_list = $R_WLM_ABAQUSHOSTLIST EOF unset SLURM_GTIDS ### name your job HERE, name it DIFFERENT from your input file! JOBNAME=name_of_my_job INPUTFILE=input_file.inp unsetopt -o NOMATCH rm -f $JOBNAME.* 2>/dev/null setopt -o NOMATCH ### Execute your application ### Please remember, to adjust the memory, it must be less than requested above abaqus interactive job=$JOBNAME input=$INPUTFILE cpus=$SLURM_NTASKS memory="$ABAQUS_MEM_ARG" |
2.2 multi node job
#!/usr/local_rwth/bin/zsh ### Job name #SBATCH --job-name=abaqus_slurm_job ### File/Path where STDOUT will be written to, %J is the job id #SBATCH --output abaqus-job-log.%J ### Request the time you need for execution. The full format is D-HH:MM:SS ### You must at least specify minutes or days and hours and may add or ### leave out any other parameters #SBATCH --time=5:00 ### Request the memory you need for your job. You can specify this ### in either MB (1024M) or GB (4G). BEWARE: This is a per-cpu limit, ### and will be multiplied with cpus-per-task for the total requested memory #SBATCH --mem-per-cpu=3900M ### Request number of hosts #SBATCH --nodes=2 ### Request number of CPUs/MPI Ranks #SBATCH --ntasks=4 ### Initialization of the software module load TECHNICS module load abaqus ### Set the amount of memory to be passed to Abaqus as a command line argument ### Beware: This HAS to be lower than the value you requested via --mem-per-cpu export ABAQUS_MEM_ARG="3584 mb" ### Change (!) to your desired work directory cd $HOME/path/to/your/dir ### Create ABAQUS environment file for current job, you can set/add your own options (Python syntax) env_file=abaqus_v6.env cat << EOF > ${env_file} #verbose = 3 #ask_delete = OFF #usub_lib_dir=os.getcwd() mp_mpi_implementation = IMPI mp_mpirun_path = {IMPI:'/opt/intel/impi/2018.4.274/compilers_and_libraries/linux/mpi/bin64/mpiexec.hydra'} mp_file_system = (SHARED, LOCAL) mp_host_list = $R_WLM_ABAQUSHOSTLIST EOF unset SLURM_GTIDS ### name your job HERE, name it DIFFERENT from your input file! JOBNAME=name_of_my_job INPUTFILE=input_file.inp unsetopt -o NOMATCH rm -f $JOBNAME.* 2>/dev/null setopt -o NOMATCH ### Execute your application ### Please remember, to adjust the memory, it must be less than requested above abaqus interactive job=$JOBNAME input=$INPUTFILE cpus=$SLURM_NTASKS memory="$ABAQUS_MEM_ARG" |
3. Best Practices for Abaqus Jobs
- Don't name your input file $JOBNAME.inp. It will be deleted by the batch script and your job won't start.
- Use separate directories for each job. The file abaqus_v6.env e.g. could be overwritten by a newly starting job.
- Use job dependencies or job arrays, such that only one or at least a few jobs run concurrently. They might "steal" licenses from each other.
4. Submitting Abaqus Jobs with Fortran User Subroutines
- Pre-compile your subroutine as a shared library with the following command line:
$> module load TECHNICS $> module load abaqus $> abaqus make -library yoursubfile.f |
- Un-comment (remove the '#' character) from the line 'usub_lib_dir=os.getcwd()' in the above example script
- Please make sure that the above entrys (between 'EOF' marks) in don't have leading spaces or tabs
5. General information about abaqus
Look >>>here<<< for more information about abaqus and especially this PDF document about the license model of abaqus.
But in short, the following table shows the token usage for a number of cores for an abaqus job.
number of cores | 1 | 2 | 4 | 8 | 12 | 16 | 24 | 32 | 64 | 128 |
|---|---|---|---|---|---|---|---|---|---|---|
| needed tokens | 5 | 6 | 8 | 12 | 14 | 16 | 19 | 21 | 28 | 38 |
or as picture:
I am not able to see my model in abaqus cae, I only get a "blue" screen.
- FastX ( the preferred way )
- ssh, which is horrably slow
I am getting an error: Abaqus Error: The following file(s) could not be located: ....inp
Don't name your input file $JOBNAME.inp, it got deleted! Also make sure that you did use an existing .inp filename and the right path.
I am getting an error: Abaqus Error: It is required that the local host is in the host list for this run. Local host: ncm0100.hpc.itc.rwth-aachen.de, mp_host_list: (('ncm1234.hpc.itc.rwth-aachen.de', 12),)
Use separate directories for each job, abaqus_v6.env got overwritten by another starting job!
I am getting an error: "..." license request queued for the License Server. Total time in queue: ... seconds.
Too many licenses have been checked out. The Slurm scheduler does not check if free abaqus licenses are available before a job is started. Unfortunately, if no license is available, abaqus does not directly abort but waits until a license becomes available. From Slurm's point of view however, the job is running and will therefore be accounted as usual.
So only let one or at least a few jobs run at one time. This could be done with job dependencies or with array jobs.
