| schuyler |
February 15, 2019 15:05 |
cfMesh - How to reconstruct after running using mpi
Hello!
I have been using cfMesh for a very long time with tons of success. Recently I have been trying to experiment with deploying it onto my cluster. I am running cfMesh with OpenFOAM-1806.
From the manual, and a few forum entries, I gathered that I should run it like (this test was on two 4-core remote computers):
Code:
preparePar
mpirun -np 8 -hostfile machines -x FOAM_SETTINGS ~/OpenFOAM/OpenFOAM-v1806/bin/foamExec cartesianMesh -parallel
This runs without problem. However, then I don't know how to recombine it! The execution doesn't do it automatically, as I had first suspected it would. I have tried reconstructParMesh -constant and it gives the following error:
Code:
Reading points from "8Core_0.25MeshSize/processor0" for time = constant
Reading points from "8Core_0.25MeshSize/processor1" for time = constant
Reading points from "8Core_0.25MeshSize/processor2" for time = constant
Reading points from "8Core_0.25MeshSize/processor3" for time = constant
Reading points from "8Core_0.25MeshSize/processor4" for time = constant
Reading points from "8Core_0.25MeshSize/processor5" for time = constant
Reading points from "8Core_0.25MeshSize/processor6" for time = constant
Reading points from "8Core_0.25MeshSize/processor7" for time = constant
Overall mesh bounding box : (-0.00257379658 -0.00374847875 -12.8007856) (24.9997 3.00000102 27.2012334)
Relative tolerance : 1e-07
Absolute matching distance : 4.72683587e-06
Merging mesh 0 with 1
--> FOAM FATAL ERROR:
Perfect match specified but number of master and slave faces differ.
master:215 slave:0
From function Foam::faceCoupleInfo::faceCoupleInfo(const Foam::polyMesh&, const labelList&, const Foam::polyMesh&, const labelList&, Foam::scalar, bool, bool, bool)
in file polyMeshAdder/faceCoupleInfo.C at line 2037.
FOAM aborting
am I missing something simple? I operate OpenFOAM in parallel on my cluster all the time, and use cfMesh all the time on my desktop. But this is the first i've tried running cfMesh remotely on the cluster nodes (since I plan to do some relatively large scale optimization).
Thanks in advance!
Schuyler |
