OpenFOAM mpirun all messed up and auto reconstructPar

xuegy · May 30, 2017, 19:20

Hello all,

I compiled an own OpenFOAM solver. It was running perfectly. But from yesterday everything goes strange. mpirun just output all messages for 8 times, and they are disordered. But the simulation result looks good. Single-threaded result has no problem at all.

And the most strange thing is it will auto reconstructPar after the mpirun is finished. This is really creepy. I compiled my OpenFOAM again but still the same problem. This even happens in all built-in solvers.

Anyone has met this before? If there's no solution I have reinstall Ubuntu.

Attachment is the log.
log.txt

jherb · May 31, 2017, 07:57

Very likely, you did not use the -parallel option with the solver, e.g.

Code:

mpirun -np 8 simpleFoam -parallel

If you forgot it, you just executed your solver eight times as serial solver. Therefore, you also got the "auto reconstructPar", because (actually) all solvers wrote the same (full) result files in the timestep directories.

May 30, 2017, 19:20	OpenFOAM mpirun all messed up and auto reconstructPar	#1
xuegy Senior Member Join Date: Jun 2016 Posts: 102 Rep Power: 10	Hello all, I compiled an own OpenFOAM solver. It was running perfectly. But from yesterday everything goes strange. mpirun just output all messages for 8 times, and they are disordered. But the simulation result looks good. Single-threaded result has no problem at all. And the most strange thing is it will auto reconstructPar after the mpirun is finished. This is really creepy. I compiled my OpenFOAM again but still the same problem. This even happens in all built-in solvers. Anyone has met this before? If there's no solution I have reinstall Ubuntu. Attachment is the log. log.txt

May 31, 2017, 07:57		#2
jherb Senior Member Joachim Herb Join Date: Sep 2010 Posts: 650 Rep Power: 22	Very likely, you did not use the -parallel option with the solver, e.g. Code: mpirun -np 8 simpleFoam -parallel If you forgot it, you just executed your solver eight times as serial solver. Therefore, you also got the "auto reconstructPar", because (actually) all solvers wrote the same (full) result files in the timestep directories.