Learning Forum

[soheil Ahmadi]

Hi fellow researchers,

I am running a 3D ANSYS MAPDL simulation in batch mode using a HPC and keep recieving the attached error. The simulation works OK for smaller models but not so much when the number of elements passes 12 milion or so. I am almost certain that I am doing something wrong since there is so much memory available in the HPC!

I used 8 Nodes, 30 Core per Node (2 CPU per Node) with the following Memory setting:

• -m 358400 -db 32768

 

Any help and tip is highly appreciated and thank you for taking the time and reading this.

best,

Soheil

 

[Mike Rife]

Hi Soheil

The last message in the screen shot you posted is pretty clear.  The OS killed the job due to too many hardware resources being used.  There could be some Linux Cgroup rule (control group) that limits the amount of RAM that can be used on the system, like say 90%, then kills the process taking the most RAM if that rule is broken.

You should show this to the system administrator - it would not be unusual for a Cgroup rule to be accidentally set on one, or more, computer node (mistakes happen).  Or maybe they mean for the rule to be there and you need to work within its defined limit.

Mike  

[soheil Ahmadi]

Hi Mike,

Thank you very much for taking the time and replying. I will further investigate this with our system administrator.

best,

Soheil