Exporting 670M Element Tet Mesh

Coreform Cubit Importing/Exporting
#1

Coreform Cubit Version : 2024.3
Platform : Red Hat Enterprise Linux 8 on HPC Cluster

Issue
Cubit is exporting a corrupt mesh?

I’m trying to export a 670 M element TET4 mesh for use in MOOSE. Cubit is able to generate the mesh with no errors, and there are no errors output during export process. However when I attempt to read the mesh into MOOSE I seg fault, and if I try and read it into Cubit I get
“ERROR: Encountered an uncaught exception ‘vector::_M_default_append’.”

I am exporting using exodus with netcdf4 turned on. I have tried using single precision exodus output as well as trying out the other exodus output options such as “no_ids”. Neither of these have any impact.

I have successfully generated a ~400 M mesh of the same geometry and it doesn’t suffer from the same issue. I’ve ran an interactive HPC session to monitor memory usage while reading the mesh into MOOSE, and I don’t get anywhere near max memory usage on a 512GB node. The only other thing I can think of is that Cubit is running out of memory while exporting, but the log file suggests the mesh was exported successfully.

any help would be much appreciate with this as I’m running out of ideas in terms of what to try:)

Thanks,
Bill

#2

Hi Bill,
could share your Mesh so that i can take a look at it?

#3

Hi Norbert, would the meshed cub5 file be good enough? Given that I don’t believe the exodus will work at all.

#4

Yes the cub should be fine. You can upload it via https://transfer.coreform.com/

#5

Finally got it uploaded, it’s a biggie:)
https://transfer.coreform.com/YCUbyWnFu4/test_mesh.cub5

#6

@bill.ellis – I think if you were to ask the MOOSE developers (here’s their forum btw) they might suggest that you start with a (much) coarser mesh and then use their Adaptivity (e.g., Initial Adaptivity Parameters) to refine “in-memory”.

#7

@gvernon thank you for the help, I will give it a go:) I’ve avoided MOOSE’s mesh refinement so far as this isn’t the mesh I actually plan to use for the majority of my simulations, it’s just for a convergence study. But if I can get a similar mesh using MOOSE’s refinement that would get me the results I need