EDEM Segmentation fault

Peter Roth
Peter Roth Altair Community Member
edited July 2022 in Community Q&A

Hi,
my name is peter and I'm am a research associate at the Institute of Agricultural Sciences in Hohenheim, Germany.
I use EDEM in my research. EDEM is installed on a high performance cluster running RedHat. Lately my simulations crash with the following error:

Is it possible to get more information why the simulation crashes and how to prevent simulations from crashing in the future ?
Thanks for the answers
Regards
Peter

Tagged:

Answers

  • Stephen Cole
    Stephen Cole
    Altair Employee
    edited August 2021

    Hi Peter,

     

    Its possible the simulation is running out of memory or there is some problem writing to the drive.  Does this happen during processing or when starting the run and does it involve any custom API models?

    You can start a simulation with --debug-logger flag (edem --debug-logger \file-location\filename.txt which writes information to the specified file however if it is a generic OS memory message this may not capture the information but worth checking.

    If it is running out of memory you could try running with a larger grid size (Rmin value) which reduces the computational efficiency but uses up less memory if running CPU.


    Regards

    Stephen

     

  • Peter Roth
    Peter Roth Altair Community Member
    edited August 2021

    Hi Stephen,
    thank you very much for your helpful reply. The problem occurs randomly during processing. I am using only the Hertz-Mindlin (no slip) model implemented in EDEM.
    This happened with the multi GPU solver with 2 Nvidia V100 GPUs. I was able to restart the crashed simulations without errors.
    Maybe there were some issues with the HPC.
    MY current simulations where started with the debugger flag. If I get any errors there I will post them here.

    Regards
    Peter

  • selena
    selena New Altair Community Member
    edited March 2022
    Peter Roth said:

    Hi Stephen,
    thank you very much for your helpful reply. The problem occurs randomly during processing. I am using only the Hertz-Mindlin (no slip) model implemented in EDEM.
    This happened with the multi GPU solver with 2 Nvidia V100 GPUs. I was able to restart the crashed simulations without errors.
    Maybe there were some issues with the HPC.
    MY current simulations where started with the debugger flag. If I get any errors there I will post them here.

    Regards
    Peter

    Hi Peter,

     

    May I ask how did you solve the problem? I faced same problem when running the EDEM which is installed on the high performance cluster. The error is same in the line 54. 

     

    Regards,

    Juan

  • Thanh Le_22011
    Thanh Le_22011 New Altair Community Member
    edited July 2022

    Hi Peter,

    We face the same issue on GPU on HPC.

    May I know if you already had solutions for this?

    image

     

  • Stephen Cole
    Stephen Cole
    Altair Employee
    edited July 2022
    Thanh Le_22011 said:

    Hi Peter,

    We face the same issue on GPU on HPC.

    May I know if you already had solutions for this?

    image

     

    Hi, the segmentation fault is a very general error however in this case it looks like there isn't a display associated with the remote machine.

     

    EDEM is typing to open the user interface but there isn't any display options.  There is a section in the EDEM Help > Appendix regarding using EDEM on remote machines.  You could use something such as VNC which is an alternative to Remote Desktop on Windows. 

    Regards

    Stephen

Categories