[Gearb0x]

Hello everyone,

I am working on a case where I would like to use redistributePar to decompose and recompose my geometry.

Unfortunately, I am encountering several issues. Some of them I could solved. But now I am blocked and would like to sollicitate your support.

My case has the following structure :

 - 0.orig
 - constant
 - system
   - controlDict
   - decomposeParDict
   - ...

The commands I am using to try to solve my case are the following ones :

mpirun -np 16 redistributePar -decompose -parallel 
ls -d processor* | xargs -i rm -rf ./{}/0
ls -d processor* | xargs -i cp -r 0.orig ./{}/0
mpirun -np 16 renumberMesh -parallel -overwrite -constant -decomposeParDict system/decomposeParDict
mpirun -np 16 patchSummary -parallel -decomposeParDict system/decomposeParDict 
mpirun -np 16 simpleFoam -parallel

I first ran into trouble just getting the decomposition going.

Indeed, OpenFOAM was creating the following directories :

processor1 
processor 2
...
processor15

but was not creating the processor0 directory. This was solved by renaming my 0 directory to 0.orig. I don't understand why since tutorials e.g. $FOAM_TUTORIALS/mesh/parallel/cavity which uses redistributePar does not need that ...

Now the issue I have when running the solver (simpleFOAM) is that I get several messages like :

Cannot find patchField entry for procBoundaryXtoY

and I don't understand why. I have tried applying the commands :

mpirun -np 16 renumberMesh -parallel -overwrite -constant -decomposeParDict system/decomposeParDict
mpirun -np 16 patchSummary -parallel -decomposeParDict system/decomposeParDict

but this does not solve the problem.

Any idea ? Thank you very much for your support !

[HPE]

redistributePar went through several iterations to get rid of various bugs.
could you please share us what software environment were you using? e.g. version

[HPE]

if it would not cause any headache for you, I would simply use "decomposePar" in place of "redistributePar -decompose -parallel" (for now).

[piprus]

I'm facing similar problem. Described it here: [v1812] redistributePar -decompose -parallel and missing /processor0 folder

So in case of any updates in this matter I would be glad to learn.

[otaolafr]

hello!
mmm If I understand correctly you are having trouble running the case in parallel. as another user mentioned before, you can use:

decomposePar -force -copyZero

this will create the processor0 folder without any issue, and it runs smoothless.
there might be a specific reason why you want to use redistributePar (if thats the case sorry for not having an answer) but if it is not the case, this is my workflow script that runs smoothly whatever the case:
I am using the runFucntions https://damogranlabs.com/2020/10/run...ck-cheatsheet/ but it is not necessary obviously, but the great benefit that you have with them is speccially in the runParallel that you dont need to specify the number of domains as with mpirun, it recovers it directly from your decomposeParDict.

#!/bin/bash
cd "${0%/*}" || exit                                # Run from this directory
. ${WM_PROJECT_DIR:?}/bin/tools/RunFunctions        # Tutorial run functions
#------------------------------------------------------------------------------
./Allclean
caseName=$(echo "${PWD##*/}")
echo "================================================================================================================================================="
echo "Begining of Meshing of $caseName case
=================================================================================================================================================
"
cp ./constant/triSurface/modelBackup.stl ./constant/triSurface/model.stl
runApplication blockMesh												
runApplication surfaceFeatureExtract									
cp -r 0.orig/. 0
decomposePar -force -copyZero											
runParallel		snappyHexMesh -overwrite								
runParallel 	renumberMesh -overwrite
runParallel 	checkMesh -allTopology -allGeometry -latestTime  		
touch simulation.foam
#------------------------------------------------------------------------------

The above script is only for the meshing part, but after that a:

runParallel 	$(getApplication)

and it will run the solver that is in your controlDict file.

one short extra info that I had a lot of problems before if you want to reconstruct the case for whatever reason (now paraView can analyse the decomposed case so it is not really important to do it), you should run first reconstructParMesh follow up by recontructPar