[NotOverUnderated]

Hello,

In my system/controlDict file I set:

runTimeModifiable   true;

As expected, this works just fine and I could edit the simulation files (e.g. system/fvSchemes, system/fvSolution, etc.) at runtime and the solver reads the new files at the next time step.


Problem:

Sometimes I am lucky enough that the simulations do not crash when I save the modifications of the files but in most cases, the simulation crashes. I think this happens because while the text editor writes to the disk, the solver tries to re-read the file which leads to the crash of the simulation.

Question

Is there any reliable solution to this issue so I could modify simulation files at runtime without the risk of crashing when I save the file using the text editor?

P.S: I am using OpenFOAM v2212

[Krapf]

In my experience, crashes of this kind have to do with making a mistake yourself (e.g. forgetting a semicolon). Could you please show an error message?

[NotOverUnderated]

Originally Posted by Krapf

In my experience, crashes of this kind have to do with making a mistake yourself (e.g. forgetting a semicolon). Could you please show an error message?

Thank you for your reply.

No. There are no mistakes while editing files. This happened to me hundreds of times. When I re-run the simulation it works fine (this demonstrates that there are no mistakes in the files).

[Yann]

What is your work environment? (OS, running simulations locally or remotely, etc...)

I've edited countless of controlDict/fvSchemes/fvSolution files at runtime for the past 10 years, using multiple OpenFOAM versions on multiple systems, and from my experience it's quite reliable (actually I don't remember facing any crash, except the ones due to typos)

Yann

[NotOverUnderated]

Originally Posted by Yann

What is your work environment? (OS, running simulations locally or remotely, etc...)

I've edited countless of controlDict/fvSchemes/fvSolution files at runtime for the past 10 years, using multiple OpenFOAM versions on multiple systems, and from my experience it's quite reliable (actually I don't remember facing any crash, except the ones due to typos)

Yann

I am running my simulations using a cluster running CentOS 7. For example while running a simulation of flow around a cylinder with 16 processors, I have tried to change the end time from 100 to 200 and then saved using :w in Vim text editor, the simulation crashed with the following error (again there are not typos, and after the crash I can re-run the simulation just fine):

    From virtual bool Foam::regIOobject::readIfModified()
    in file db/regIOobject/regIOobjectRead.C at line 271
        Re-reading object controlDict from file "/home/feelTheFearAndDoItAnyway/scratch/squareCylinder/piso/system/controlDict"
--> FOAM Warning : 
    From bool Foam::IOobject::readHeader(Foam::dictionary&, Foam::Istream&)
    in file db/IOobject/IOobjectReadHeader.C at line 116
    Reading "/home/feelTheFearAndDoItAnyway/scratch/squareCylinder/piso/system/controlDict" at line 1
    First token could not be read or is not 'FoamFile'

Check header is of the form:

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  2112                                  |
|   \\  /    A nd           | Website:  www.openfoam.com                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    arch        "LSB;label=32;scalar=64";
    class       dictionary;
    location    "system";
    object      controlDict;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //



[0] --> FOAM FATAL IO ERROR: (openfoam-2112)
[0] problem while reading header for object controlDict
[0] 
[0] file: system/controlDict at line 1.
[0] 
[0]     From virtual Foam::autoPtr<Foam::ISstream> Foam::fileOperations::uncollatedFileOperation::readStream(Foam::regIOobject&, const Foam::fileName&, const Foam::word&, bool) const
[0]     in file global/fileOperations/uncollatedFileOperation/uncollatedFileOperation.C at line 563.
[0] 
FOAM parallel run exiting
[0] 
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
In: PMI_Abort(1, N/A)
srun: Job step aborted: Waiting up to 32 seconds for job step to finish.
slurmstepd: error: *** STEP 6296119.0 ON ngm552 CANCELLED AT 2023-07-02T12:23:16 ***
srun: error: ngm555: tasks 11-15: Killed
srun: Terminating StepId=6296119.0
srun: error: ngm552: tasks 0-10: Killed
---- END ----

[NotOverUnderated]

I have finally solved the issue!

It turned out that Vim saves the changes directly to the file so the file goes through a period of time where it is in inconsistent state and when OpenFOAM attempts to read the file during this period, it may not find what it expects and hence fail.


The solution is to told Vim to use an "atomic" save method (saves to a temporary file then replaces the old one) by adding this command in ~/.vimrc:

:set backupcopy=yes

Now, the simulations no longer crash during runtime modifications of the files!

[Yann]

Thanks for your feedback, it's definitely good to know!

Yann