CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   OpenFOAM Native Meshers: snappyHexMesh and Others (https://www.cfd-online.com/Forums/openfoam-meshing-snappyhexmesh/)
-   -   SnappyHexMesh in Parallel problem (https://www.cfd-online.com/Forums/openfoam-meshing-snappyhexmesh/100269-snappyhexmesh-parallel-problem.html)

swifty April 22, 2012 19:00

SnappyHexMesh in Parallel problem
 
2 Attachment(s)
Hello,

I am trying to run snappyHexMesh in parallel using openFOAM 2.1.0 to mesh an aerofoil. I had it working without any problems in serial first, I only have problems now I am trying in parallel. I start by running blockMesh to generate a base mesh, then I decompose the mesh using scotch for 2 processors. Then change the decompose dictionary to ptscotch and run snappyHexMesh. This fails with the following error,

[0] processorPolyPatch::calcGeometry : Writing my 6925 faces to OBJ file "/home/aswift/OpenFOAM/aswift-2.1.0/run/coord_seligFmt/aerofoil/processor0/procBoundary0to1_faces.obj"
[1] processorPolyPatch::calcGeometry : Writing my 6925 faces to OBJ file "/home/aswift/OpenFOAM/aswift-2.1.0/run/coord_seligFmt/aerofoil/processor1/procBoundary1to0_faces.obj"
[1] processorPolyPatch::calcGeometry : Dumping cell centre lines between corresponding face centres to OBJ file"/home/aswift/OpenFOAM/aswift-2.1.0/run/coord_seligFmt/aerofoil/processor1/procBoundary1to0_faceCentresConnections.obj"
[0] processorPolyPatch::calcGeometry : Dumping cell centre lines between corresponding face centres to OBJ file"/home/aswift/OpenFOAM/aswift-2.1.0/run/coord_seligFmt/aerofoil/processor0/procBoundary0to1_faceCentresConnections.obj"
[1]
[1]
[1] --> FOAM FATAL ERROR:
[1] face 597 area does not match neighbour by 0.103929% -- possible face ordering problem.
patch:procBoundary1to0 my area:1.33687e-05 neighbour area:1.33549e-05 matching tolerance:6.7218e-10
Mesh face:2108231 vertices:4((0.535348 -0.114329 0.03125) (0.539191 -0.114329 0.03125) (0.539191 -0.110848 0.03125) (0.535348 -0.110851 0.03125))
If you are certain your matching is correct you can increase the 'matchTolerance' setting in the patch dictionary in the boundary file.
Rerun with processor debug flag set for more information.
[1]
[1] From function processorPolyPatch::calcGeometry()
[1] in file meshes/polyMesh/polyPatches/constraint/processor/processorPolyPatch.C at line 239.
[1]
FOAM parallel run exiting
[1]

To try and solve this error I have tried adding a renumberMesh command before snappyHexMesh in my run file, but that doesn't help. I have tried changing the matchTolerance value in the processor0 and 1 folders, but the value appears to be overwritten.

I have attached the snappyHexMesh log and the files I use for the run.

I would be grateful for any help.

Regards

Swifty

kid April 23, 2012 06:12

If snappyHexMesh runs in serial what is need to run it in parallel. use decomposePar and proceed with solver. By running snappyHexMesh in parallel you might be making multiple instances of same mesh on each processor.
If anything you want to run in parallel it should be solver and not snappyHexMesh.

swifty April 23, 2012 19:21

I have gone back to the basics and I am doing the grid generation in serial and then run the solver in parallel. I found a problem with the decomposition of the mesh, to make it work I had to delete ccx ccy ccz cellLevel pointLevel from 0. I also added the keyword

structured yes;

to the decomposeParDict. I can now run simpleFoam in parallel.

haakon June 25, 2012 02:26

I have the exactly same problem with snappyhexMesh as swifty. I try to make a mesh with sHM in parallel, and it fails with the same error message. In serial everything seems to be OK.

If anyone can help me/us with this I would highly appreciate that.

Eloise September 26, 2012 08:37

If you still have this issue, go to have a look at the following thread:
http://www.cfd-online.com/Forums/ope...on-method.html

jojosaxo February 24, 2015 09:31

2 Attachment(s)
Hello everyone,

I'm new to Openfoam and I'm trying to run snappyHexMesh in parallel.
As some people above I got some strange message about facing area no-matching between processor.

FOAM FATAL ERROR:
[1] face 597 area does not match neighbour by 0.103929% -- possible face ordering problem.
patch:procBoundary1to0 my area:1.33687e-05 neighbour area:1.33549e-05 matching tolerance:6.7218e-10

I'm actually doing a mesh around a Naca aifroil.

This issue actually doesn't occur with a 3D case I'm working on.

Does anyone know where this problem is coming from and how to solve it???

Thanks for your help

See Attached my file....

Harak September 17, 2015 06:37

Quote:

Originally Posted by swifty (Post 356516)
I have gone back to the basics and I am doing the grid generation in serial and then run the solver in parallel. I found a problem with the decomposition of the mesh, to make it work I had to delete ccx ccy ccz cellLevel pointLevel from 0. I also added the keyword

structured yes;

to the decomposeParDict. I can now run simpleFoam in parallel.

Hey swifty,

I now it's been a long time but are you able to explain this a little further? I would really appreciate it!
What did you delete and where did you delete it?

Thanks!
:)

Eloise September 17, 2015 11:47

Hi Harak,Try using simple decomposition method instead of scotch or pscotch for snappyHexMesh.
Regards,
Elo´se

Harak September 17, 2015 12:02

Quote:

Originally Posted by Eloise (Post 564465)
Hi Harak,Try using simple decomposition method instead of scotch or pscotch for snappyHexMesh.
Regards,
Elo´se

Hey Eloise,

thanks for your quick reply.

How would you decompose this geometry using simple?

http://s10.postimg.org/6o4t0i6sp/geo.png

I would like to split it in 32 part, because I've 32 cores available. I'm thinking of 4 subdomains to both left and right sides and 2 subdomains to the top...so 4*4*2=32. Is that possible or is there a better way you can provide?

Thanks a lot!
:)

Eloise September 17, 2015 12:26

Tricky indeed :) I guess it's best to avoid having empty partitions, while trying to have equivalent number of cells per partitions... I don't know how many cells you have, but it might not be best to use as many partitions pas possible. Try first (4,4,1), just to see if the decomposition method is working. If it does, you can then vary the decomposition per direction and see what works faster for you.
Regards,
Elo´se

jaydeepKhajure November 6, 2015 05:40

snappyHexMesh parallel run problem
 
Hi all,

I need little help with snappyHexMesh. It's a 3D geometry, placed in the middle of the domain. When I run snappyHexMesh on single processor, it goes well. But, when I run it in parallel, object I placed in refinement box or in domain goes missing. It does not include the patch. I have checked my snappyHexMeshDict, decomposeParDict and blockMeshDict with motorbike case, everything seems fine.

Can anybody look into this ?

Regards,

Jaydeep


All times are GMT -4. The time now is 20:34.