[alinuman15]

Hi All:-
I have a successful run of a two-phase flow case study by using interFoam solver(parallel running with 64 processors for 48 hours). However, The solver could not produce any time step solution despite the solution run smoothly as you can see from the file (the link below).
Any idea?
Thanks!
https://drive.google.com/file/d/1GwR...ew?usp=sharing

[piu58]

My recommendations:
- Simplify your model for testing it. Use less nodes.
- Make the writeInterval much shorter, to see if anything happens. Best you write at every step first
- Start with one processor. The course of the program can be followed much easier with only one processor.

[alinuman15]

Originally Posted by piu58

My recommendations:
- Simplify your model for testing it. Use fewer nodes.

1. How can I decrease the nodes and make sure that the model still able to capture the physical phenomenon? Now, I set the model with a number of cells more than eight million and not sure if it is enough for recognizing the capillary forces.

- Make the writeInterval much shorter, to see if anything happens. Best you write at every step first [/QUOTE]
2. My time step is too small and that will be reflected on the required memory after reconstruction. How can I avoid the expected aggravated required memory size?
- Start with one processor. The course of the program can be followed much easier with only one processor.[/QUOTE]
3- Based on 1 and 2, do you think I still can follow up with one processor?

Thank you so much for replying!

[alinuman15]

Dear Uwe:-
Update:-
I have time step ( 1e-6) and after your suggestion, I have changed the writeInterval from 0.0004 to 0.000009. After running the case, the solver just extracted the time step 0.0004 and stopped and reported following report:-

Create time

Create mesh for time = 0.000439557937677


PIMPLE: Operating solver in PISO mode

Reading field p_rgh

--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 4 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.
--------------------------------------------------------------------------
Now, I decreased the writeInterval even more (0.000005) and gave it another run. Waiting for results. But, still, need your feedback regarding the previous post.
Thanks again!

[piu58]

> decrease the nodes and make sure that the model still able to capture the physical phenomenon?

It is not necessary that the rougher geometry covers all phenomena in high accuracy. This approach should contain, however, all elements that the high-resolution has in it.

The main purpose of this simplified model is to check the calculation process and to find mistakes in modelling.

[jpeter3]

It looks like your case was terminated by an error in MPI mistery box that is supposed to run parallel cases. MPI does have nothing to do with cases it only provides a framework for parallel operations. That suggests that your problem is not in the openfoam case settings and solver but rather in the system you are running your case on. (just guessing though)

[alinuman15]

Originally Posted by jpeter3

It looks like your case was terminated by an error in MPI mistery box that is supposed to run parallel cases. MPI does have nothing to do with cases it only provides a framework for parallel operations. That suggests that your problem is not in the openfoam case settings and solver but rather in the system you are running your case on. (just guessing though)

As I tried it out:
It is an OpenFOAM setting issue since I tried it on my personal computer with parallel run => it produced the same thing. However, when I changed the write interval as Uwe suggested with the parallel run on supercomputer , it made a difference!