[Pallav]

Hi,

I run my simulations in OF in parallel on a supercomputer (TACC). I have to submit the simulation job and ask for a defined period of time for the simulation to run.

TACC kills the simulation after the simulation has run for the requested amount of time.

In OF, is there a way to resume the simulation from where it stopped?

Please advise.

Thanks.

-Pallav

[wyldckat]

Greetings Pallav,

What you are looking for is this option in controlDict:

Originally Posted by http://www.openfoam.com/docs/user/controlDict.php

startFrom       latestTime;    //Most recent time step from the set of time directories.

Check the link in the quote for more informations on what you can do in controlDict

Best regards,
Bruno

[Pallav]

Thanks a lot Bruno.

[lovecraft22]

You'll have your simulation to be started with the runTimeModifiable option set to yes on your controlDict or no changes on that file will affect your simulation…

[Bernhard]

Well, if the simulation was killed by the computer, then these changes will still take effect.

[Pallav]

Lore and Bernhard,

Thank you for your inputs.

[aujamal20]

Dear
I am having a problem to resume the openFoam simulation after making manual changes in last time directory of previous run. For example if I run a simulation for 100 s and I have 0,20, 40 ...,100 time directories. Now I reverse the flow direction by changing inlet/outlet BC for 100/U , manually. From now onward I cant run the simulation from latestTime of 100.
While I set my controlDict properly but it shows an error. But If I remove the time directory 100 and want to run from any other directory like 80, which is not changed, then the simulation runs properly.
Please let me how to resolve this problem.

Regards

[wyldckat]

Greetings Jamal,

A few questions:

1. Which OpenFOAM version are you using? It's possible that you're triggering an old bug on an old version of OpenFOAM.
2. What value do you have for "endTime" in "system/controlDict"?
3. Are you using a stopping criteria based on residuals?

Best regards,
Bruno

[aujamal20]

Dear Bruno
1. I am using OF 2.1.0
2. When I start a simulation I set endTime in controlDict to 100 when simulation ends then I edit the last time directory 100/U and 100/T to reverse the flow by switching the BC for inlet/outlet & set endTime to say 200.
3. So I don't use any stopping criteria.

Thanks and regards,
Jamal

[wyldckat]

Hi Jamal,

I can't remember very well... there was at least one bug in 2.1.0 that might be related: http://www.openfoam.org/mantisbt/view.php?id=382
So you might want to update to 2.1.1.

Nonetheless... do you change the "endTime" value in "controlDict" before continuing?
edit: OK, you updated your answer... so you did change...

Additionally, is "startFrom" set to "latestTime" for at least the second run?

Best regards,
Bruno

[hpp10]

Thanks a lot! I did parallel computing with mpi, when there was no space for my computing in my devices. So, I cleared some room, using your suggestion, I set in controlDict ：startFrom latestTime; This didnot work. But with your suggesting website tutorial guide, I got the topic of the problem: the lastestTime cannot be used, when I set in the controlDict:sartFrom startTime; startTime 2140; It works well.(my latest time is 2150, I guess there must be some missing files in 2150, so latestTime didnot work)

[jaydeepKhajure]

Hello everybody,

we can resume stopped openfoam simulation. just need some changes in system/controldict file as follows,

startFrom latestTime;

latestTime 0.xxx;

you will need to check the stopped time in the log file and enter that time as latestTime. it worked for me. if anybody has another better solution, please share.

Regards,

Jaydeep

[sukratu]

Hello Jaydeep,

Are you sure it works without us considering the "write Interval" we
set for our flow fields? Please confirm or modify your post.

Best!

[wyldckat]

Greetings to all,

OK, in order to make things perfectly clear, let's use the tutorial "incompressible/icoFoam/cavity" as a basis for demonstrating this issue.

The default settings in this tutorial are these:

startFrom       startTime;

startTime       0;

stopAt          endTime;

endTime         0.5;

If you want to continue from the latest time, you have at least 2 choices:

1. Either you force the start and stop times:
   

   startFrom       startTime;
   
   startTime       0.5;
   
   stopAt          endTime;
   
   endTime         1.0;

2. Or you use the "latestTime" setting for "startFrom":
   

   startFrom       latestTime;
   
   startTime       0;
   
   stopAt          endTime;
   
   endTime         1.0;

And if you don't believe be, I advise people to study again the very first tutorial case that is explained in the OpenFOAM User Guide, section 2.1. More specifically:

1. First choice, see sub-section: "2.1.5.4 Control adjustments"
2. Second choice, see sub-section: "2.1.7.1 Pre-processing"

And for those who complain about this being a "needle in a haystack"... Well, this is OpenFOAM you're dealing with, you better train yourself to find these kinds of needles in your cases . One wrong decimal separator is all that it takes for you to be simulating a pipe as large as a river or as the Earth .

Best regards,
Bruno

[ss32]

Are there any special steps to resume a parallel case?

[aee]

Parallel cases will resume the same way (using the same mpirun command). However, I find my self reconstructing and then deleting the decomposed directories between stages (to see results in case I want to change /constant files), so I have to re-decompose before resuming. I always have "startFrom latestTime;" in controlDict.

[srikrishnan_balasubramani]

Hi
I have some problem restarting the simulation from the stopped time ( stopped due to system time out). The thing is my write intervals are quite high. When I try to reconstruct the files, I am not able to. I think so because there is only 0 and constant folders in each processor. So when I use

---> reconstructPar -latestTime
I get error message "no time selected"


Is there any way to restart the simulation from the stopped time ?

[mitti]

As far as I know: No

[mkjmalik]

Hi all,

I want to know that how to resume the simulation with log file..
I lose my log file starting data if i run my simulation from latest time.. the new log file is written from latest time.

[cdegroot]

Originally Posted by mkjmalik

Hi all,

I want to know that how to resume the simulation with log file..
I lose my log file starting data if i run my simulation from latest time.. the new log file is written from latest time.

Normally, you should just copy your log file to a different filename, or write to a new filename upon restarting. The log file is just the screen output piped to a file, so it will not continue writing in the same file.