[chen112p]

Hello all, I have made sure the test cases can run in serial cartesianMesh.

I keeps getting "segmentation fault" when I ran them again with mpi:

mpirun -np 32 cartesianMesh -parallel

The fault usually comes very randomly:

...
Total number of cells 693956
Finished extracting polyMesh
Checking for irregular surface connections
Checking cells connected to surface vertices
Found 40289 boundary faces
Found 0 problematic vertices
Finished checking cells connected to surface vertices
Checking for non-manifold surface edges
--------------------------------------------------------------------------
A process has executed an operation involving a call to the
"fork()" system call to create a child process. Open MPI is currently
operating in a condition that could result in memory corruption or
other system errors; your job may hang, crash, or produce silent
data corruption. The use of fork() (or system() or other calls that
create child processes) is strongly discouraged.

The process that invoked fork was:

Local host: [[17085,1],27] (PID 18387)

If you are *absolutely sure* that your application will successfully
and correctly survive a call to fork(), you may disable this warning
by setting the mpi_warn_on_fork MCA parameter to 0.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpirun noticed that process rank 27 with PID 18387 on node nnode9 exited on signal 11 (Segmentation fault).
--------------------------------------------------------------------------


I looked some threads but no one really reported this issue seems like. Thanks


Junting

[Krao]

Hi Junting Chen,

I hope the following link would help you to run your case in parallel.

How to run CfMesh in parallel?

[chen112p]

Hello Krao,

I am able to run it on one pc with OpenMP. The issue is MPI.

Thanks,

Junting

[Krao]

Did you run preparePar? before running cartesianMesh in parallel

Krao

[chen112p]

yes I did. I ran preparePar so I got 32 files.

Junting

[chen112p]

So my workflow is :
1. put aaa.stl in the folder
2. run 'surfaceGenerateBoundingBox aaa.stl aaa_boundingbox.stl xMin xMax yMin yMax zMin zMax
3. run 'surfaceFeatureEdges aaa_boundingbox.stl aaa_FeatureEdges.fms -angle 5'
4. run 'surfaceFeatureEdges aaa_boundingbox.stl aaa_FeatureEdges.vtk -angle 5'
5. write systme/meshDict:
- surfaceFile "aaa_FeatureEdges.fms;
- edgeFile "aaa_FeatureEdges.vtk";
6. write decomposeParDict:
- numberOfSubdomains 32;
7. run 'preparePar'
8. run 'mpirun -np 32 cartesianMesh -parallel > 01.cfMesh.log'

output log says following at the end:

Extracting edges
Starting topological adjustment of patches
No topological adjustment was needed
Starting geometrical adjustment of patches
Found 8 corners at the surface of the volume mesh
Found 649 edge points at the surface of the volume mesh
12 edge groups found!
--> FOAM Warning :
From function void edgeExtractor::extractEdges()
in file utilities/surfaceTools/edgeExtraction/edgeExtractor/edgeExtractor.C at line 2115
Found 0 points with inverted surface normals. Getting rid of them...
Starting untangling the surface of the volume mesh
Number of inverted boundary faces is 14
[nnode7:23282] *** An error occurred in MPI_Recv
[nnode7:23282] *** reported by process [1091829761,12]
[nnode7:23282] *** on communicator MPI_COMM_WORLD
[nnode7:23282] *** MPI_ERR_TRUNCATE: message truncated
[nnode7:23282] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
[nnode7:23282] *** and potentially your MPI job)

Thanks a lot for your patience!

Junting Chen

[Krao]

Hi Junting Chen,

I have followed something like what you mentioned and was able to run the case with the following steps, can you try it out once.

1. put aaa.stl in the folder
2. run 'surfaceGenerateBoundingBox aaa.stl aaa_boundingbox.stl xMin xMax yMin yMax zMin zMax
3. run 'surfaceFeatureEdges aaa_boundingbox.stl aaa_FeatureEdges.fms -angle 5'
4. run 'FMSToSurface -exportFeatureEdges aaa_FeatureEdges.fms new_aaa_FeatureEdges.fms'

The 4th step mentioned above will generate a .vtk file along with new .fms file. Write this .vtk file under edge refinement of system/meshDict

5. write systme/meshDict: under edge refinement
-outputof4thstep.vtk

6. write decomposeParDict:
- numberOfSubdomains 32;
7. run 'preparePar'
8. run 'mpirun -n 32 cartesianMesh -parallel > 01.cfMesh.log'

This workflow, which I followed on OpenFOAM6 and got no errors, please let me know if you have any issues with the above steps.

Krao