[Arad]

In a parallel SU2 run (using parallel_computation.py) on many compute nodes:

Is there a way to stop the run in a peaceful manner (that is, not ctrl C but in a way that
all output and restart files are collected and written down before exit).
Common practices are sending a kill command with parameters to one of the compute nodes, or touching a file with a pre-coded name (like killSU2) in the running folder, that once located, initiate ordered run termination.
Could not trace any structure of that kind in the source. Did I miss something ?

Thanks
Eran

[talbring]

Hi Arad,

you can create a file "runtime.dat" in the working folder to set the new number of external iterations (so just one line "EXT_ITER=...").

[Arad]

Hi Talbring

Your suggestion worked nicely for direct solution.

However, in optimization run, using shape_optimization.py script,
the runtime.dat file had no effect. i have set there
FFD_ITERATIONS, FFD_TOLERANCE and EXT_ITER
but none was read or had an effect on the
running job. Is there other way to control an optimization run ?


Thanks

[jitendraseregar@gmail.com]

Originally Posted by talbring

Hi Arad,

you can create a file "runtime.dat" in the working folder to set the new number of external iterations (so just one line "EXT_ITER=...").

I have similar situation, where I have to stop CFD solver at particular iteration and import new surface gird file and continue running CFD solver with new file

[martinsimon]

Did anyone get the solution for the mp optim run?

[beatlejuice]

Hey Arad,

if you put the runtime.dat file into the folders :

DESIGNS/0.../DIRECT/
or
DESIGNS/0.../ADJOINT/

it should work ... at least EXT_ITER= ... works to finish the direct or the adjoint computation ... but it will proceed the optimization procedure ....

cheers