CFD Online Discussion Forums - [snappyHexMesh] SnappyHexMesh multioprocessor run causes segmentation fault

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- OpenFOAM Meshing & Mesh Conversion (https://www.cfd-online.com/Forums/openfoam-meshing/)

- - [snappyHexMesh] SnappyHexMesh multioprocessor run causes segmentation fault (https://www.cfd-online.com/Forums/openfoam-meshing/150582-snappyhexmesh-multioprocessor-run-causes-segmentation-fault.html)

SnappyHexMesh multioprocessor run causes segmentation fault

Hello experts,

I am new to the subject and I am quite confident that i messed up a lot in the snappyhexmeshDict and preprocessing, however I fail to learn anything from the error message that the mesher throws.
When I run

mpirun -np 4 snappyHexMesh -overwrite -parallel

it finally ends up here (snippet of output)

Code:

Splitting mesh at surface intersections

---------------------------------------



Introducing baffles for 0 faces that are intersected by the surface.



Edge intersection testing:

    Number of edges             : 4156674

    Number of edges to retest   : 0

    Number of intersected edges : 0

Created baffles in = 1.02 s





After introducing baffles : cells:1315931  faces:4156674  points:1544788

Cells per refinement level:

    0    990057

    1    60188

    2    139014

    3    126672



Introducing baffles to block off problem cells

----------------------------------------------



[0] #0  Foam::error::printStack(Foam::Ostream&)[1] #0  Foam::error::printStack(Foam::Ostream&)[3] #0  Foam::error::printStack(Foam::Ostream&)[2] #0  Foam::error::printStack(Foam::Ostream&) at ??:?

[3] #1  Foam::sigSegv::sigHandler(int) at ??:?

[1] #1  Foam::sigSegv::sigHandler(int) at ??:?

the complete log is attached. Can anyone point me to the source of the problem?
I use OpenFoam 2.3.0 on Ubuntu Linux.

All the best,
Marcus

Hello Marcus,

It appears that no faces from the initial mesh intersect the STL surface according to your log file.It may appear that your surface is outside your initial mesh. It's may be dividing by 0, thus the segmentation error.

On line 309 in your log file:
" Introducing baffles for 0 faces that are intersected by the surface."

Try doing an inspection on Paraview of your STL surface and initial mesh.

Thanks,

Daniel

And try to run in serial, before you switch to parallel.

Thank you very much for your answers.

The volume I have to create the mesh for is all inside my basic block mesh, but it has some common surface with it which may cause a problem and is easily changed.
I will also skip the parallel option for my first try, so I have something to fiddle with for the weekend.

Thank you!
Marcus

Changed settings to have the geometry well embedded in the background mesh, also changed to serial processing, but had the same error again.

However I reorganised the case and changed things in the snappyHexMeshDict and now it happily meshes to the end.

The annoying thing is that I cannot change it back to reproduce the error now, what surprises me a bit. So it I don't know what is exactly happening. Someone here raised the same error, but it had not been sorted out then as well

http://www.cfd-online.com/Forums/ope...erminal-2.html

So next time I will take more care for more traceability, sorry.

Again, thanks very much for your help.

Just giving this a bump as I can reproduce the issue now.

It has nothing to do with multiprocessor mode. I can raise it when I refer to a non existing geometry in the refinementSurfaces section of the snappy hex mesh dict file, by a spelling error or typo for instance.

SHM will raise a warning very early but continue with the work, then it will mention the 0 faces that are intersected by the surface, shortly afterwards it will run into the segmentation fault.

Simply correcting names fixes this for me.

Regards,
Marcus