ANSYS: Difference between revisions
No edit summary |
No edit summary |
||
Line 10: | Line 10: | ||
To load MACS into your environment, use the following module command: | To load MACS into your environment, use the following module command: | ||
<pre> | <pre> | ||
module load | module load ANSYS/19.1 | ||
</pre> | </pre> | ||
==Running ANSYS Interactively== | ==Running ANSYS Interactively== | ||
If you are trying to run ANSYS interactively, you can start reserve interactive compute resources with the command " | If you are trying to run ANSYS interactively, you can start reserve interactive compute resources with the command "sinteractive --ntasks=1 --cpus-per-task=1 --mem-per-cpu=4096 --time=04:00:00 --partition=short" once you are logged into the cluster via SSH. Please do not run computations on the head node. You may need to adjust the time and mem-per-cpu (RAM) requirements if the analysis is expected to run more than 4 hours or use more than 4 gigabytes of RAM. After gaining access to the interactive session you can issue the commands "module load ANSYS/19.1" and "runwb2" to run ANSYS interactively. | ||
==Running ANSYS as an SMP Job== | ==Running ANSYS as an SMP Job== | ||
Line 42: | Line 42: | ||
ansys < file.jou >& logfile.txt | ansys < file.jou >& logfile.txt | ||
Note, you will likely need to adjust the | Note, you will likely need to adjust the time= (run time), mem-per-cpu= (RAM) and openmpi 24 (core count) to suite your needs. You should start small and scale up when you are confident in the successful outcome. The [http://148.204.81.206/Ansys/150/Running%20ANSYS%20Fluent%20under%20SGE.pdf Running ANSYS on SGE reference] may provide additional useful information for accomplishing your ANSYS goals. | ||
==Troubleshooting== | |||
If you are getting the following error opening "Geometry" on ANSYS in Cheaha | |||
<pre> | |||
<IBM Platform MPI>: : warning, dlopen of libhwloc.so failed (null)/lib/linux_amd64/libhwloc.so: cannot open shared object file: No such file or directory | |||
****************************************************************************** | |||
*** A fatal error has occurred within AnsysWBU.exe and it must be closed...*** | |||
****************************************************************************** | |||
</pre> | |||
run the following command prior to starting the workbench: | |||
<pre> | |||
unset SLURM_GTIDS | |||
</pre> | |||
If you are getting the following error attempting to start up Ansys: | |||
<pre> | |||
libGL error: No matching fbConfigs or visuals found | |||
libGL error: failed to load driver: swrast | |||
</pre> | |||
Try running ansys within a vnc session. | |||
If you are seeing startup or graphics problems, force ansys to start using software acceleration as follows: | |||
<pre> | |||
runwb2 -oglmesa | |||
</pre> | |||
This may not be required on a GPU node, but for standard compute nodes, this should help. | |||
[[Category:Software]] | [[Category:Software]] |
Latest revision as of 16:24, 30 November 2018
Note: ANSYS on Cheaha is only available for School of Engineering users. The license also cannot be used for commercial purposes!
The license only supports serial and SMP threaded work and doesn't support parallel (MPI) jobs.
Project website: http://www.ansys.com/
Load ANSYS module
To load MACS into your environment, use the following module command:
module load ANSYS/19.1
Running ANSYS Interactively
If you are trying to run ANSYS interactively, you can start reserve interactive compute resources with the command "sinteractive --ntasks=1 --cpus-per-task=1 --mem-per-cpu=4096 --time=04:00:00 --partition=short" once you are logged into the cluster via SSH. Please do not run computations on the head node. You may need to adjust the time and mem-per-cpu (RAM) requirements if the analysis is expected to run more than 4 hours or use more than 4 gigabytes of RAM. After gaining access to the interactive session you can issue the commands "module load ANSYS/19.1" and "runwb2" to run ANSYS interactively.
Running ANSYS as an SMP Job
Running ANSYS tools via batch mode may also be possible. A quick Google turns up these sample job scripts for running fluent via SGE. You may find the non-interactive "Method 2" example could work well on our cluster, since we also offer the module load ansys option. Combining this example with our provided template could lead to the following script that may work for you:
#$ -S /bin/bash #$ -cwd # #$ -N ANSYS #$ -pe smp 8 #$ -l h_rt=00:10:00,s_rt=0:08:00,vf=1G #$ -j y # #$ -M YOUR_EMAIL_ADDRESS #$ -m eas # # Load the appropriate module files module load openmpi/openmpi-gnu module load ansys/ansys-14.5 #$ -V # turn off shell variable "noclobber" which prevents data files from being # overwritten. Fluent jobs are typically setup to overwrite data files. unset noclobber # Run the ansys job; ansys < file.jou >& logfile.txt
Note, you will likely need to adjust the time= (run time), mem-per-cpu= (RAM) and openmpi 24 (core count) to suite your needs. You should start small and scale up when you are confident in the successful outcome. The Running ANSYS on SGE reference may provide additional useful information for accomplishing your ANSYS goals.
Troubleshooting
If you are getting the following error opening "Geometry" on ANSYS in Cheaha
<IBM Platform MPI>: : warning, dlopen of libhwloc.so failed (null)/lib/linux_amd64/libhwloc.so: cannot open shared object file: No such file or directory ****************************************************************************** *** A fatal error has occurred within AnsysWBU.exe and it must be closed...*** ******************************************************************************
run the following command prior to starting the workbench:
unset SLURM_GTIDS
If you are getting the following error attempting to start up Ansys:
libGL error: No matching fbConfigs or visuals found libGL error: failed to load driver: swrast
Try running ansys within a vnc session.
If you are seeing startup or graphics problems, force ansys to start using software acceleration as follows:
runwb2 -oglmesa
This may not be required on a GPU node, but for standard compute nodes, this should help.