Learning Forum

[helen.durand]

Hello,nI have been running ANSYS Structural simulations on a HPC cluster with a UI, and I am only able to complete simulations with 1 core active. If I try to run the simulation with multiple cores and with the distributed option enabled, the simulation hangs up at 10% at the 'creating mathematical model' stage. What might cause this to happen, and how would I troubleshoot this?nThank you!n

[helen.durand]

To add more details to my previous question:nOn the HPC cluster, I can start jobs with many cores but large Transient Structural jobs will only run to completion if I disable the 'Distributed' option and specify 1 core (as in the image below). nPerhaps this has something to do with the 'Solve Process Settings' not being set up correctly for a HPC cluster? (See: https://ansyshelp.ansys.com/account/secured?returnurl=/Views/Secured/corp/v201/en/wb_sim/ds_using_solve_handlers.html)nOr do I need to use the Remote Solve Manager? (See: https://ansyshelp.ansys.com/account/secured?returnurl=/Views/Secured/corp/v211/en/wb2_help/wb2h_usingrsmwb.html)nI would appreciate any feedback. Thank you!n

[helen.durand]

Since I have not gotten a response, should I try reposting this under a different forum category (perhaps 'ANSYS Products')? n

[Mike Rife]

Array Which version are you using? Are there any security programs installed on that computer? Enhanced Permission type of programs? Any flavor of Python?.MIken

[helen.durand]

The HPC cluster uses ANSYS 2020R1, a Slurm job-scheduler, there are no security programs, and python 3.7 is being used. Please let me know if any other information would be helpful. n

[helen.durand]

Hello, does anyone have any idea of what could be causing my issue? nI notice that the User's Guide (the link above) mentions something about a HPC configuration queue. Could creating this queue address my issue?nThank you!n

[Mike Rife]

If solve jobs sent to the cluster will only solve using one core and not using distributed parallel processing, please bring this to the attention of the cluster admin. There could be something misconfigured with the Slurm queue. nIf your solution is hanging when solving locally, check the solution information: is there any information given? If not then most likely some security program has stopped the start up of the solver. Remember that WB Mechanical is a pre and post processor. Mechanical APDL is being used as the solver behind the scenes. So Mechanical passes off the input file and the solve settings (number of cpu cores etc) and a script starts up MAPDL. When the MAPDL start up is interrupted the vast majority of the time some security program is blocking it. This can happen even if it worked in the past: the security programs can get their settings corrupted on an update.nMike n