Raising error from Fluent to PyFluent

[GautierR]

Hello,

I am having a super simple question that I wasn't able to answer from PyFluent documentation:

Is there a way to catch Fluent's solver error?

In other word if we doing something like:

try:
        run_calculation.iterate(iter_count=200)
except Exception as e:
        print(f"[fatal] case failed: {e}")
        solver.exit()

I would expect PyFluent to raise an error if there is a calculation error (like divergence, segfault or similar). It seems that's not the case.

Is there a possible solution to catch such error?

Thank you for the help
Gautier

[ansjschuetz]

Just as a workaround:
It appears you just want to make sure that the Fluent session exits on error.
You may be able to achieve that by executing the legacy TUI command "/file/set-batch-options , yes ,".
Let us know if you need help in finding out how to do that. (May be something like "<solver>.tui.file.set_batch_options('yes', 'yes', 'no')"...)

[GautierR]

Thanks for the feedback,

Unfortunately, the exit_on_error option is by default True (or "yes" in TUI). So this does not solve my issue, it seems that something is preventing Fluent to exit while it is launch with PyFluent.

I guess there might be some option in the PyFluent Fluent launcher to get error behavior similar to the headless "slurm" version. In the standard slurm command I used (fluent -g -slurm) , it killed itself on error.

[ansjschuetz]

Hello @GautierR ,
I've experimented a bit further using the script listed below, and I see the Fluent session stop and die immediately upon encountering the "dummy" error. (...both with Fluent 2024 R1 and current development state of 2026 R1.)

--> What kind of error is it that you cannot get it to exit upon?

import sys
import os

import ansys.fluent.core as pyfluent

## sys.path.append(os.path.join(os.path.dirname(__file__), "../.."))
## sys.path.append(os.path.join(os.path.dirname(__file__), "."))

# Set log level to info
# pyfluent.set_log_level("INFO")

## os.environ['AWP_ROOT261'] = '/home/jschuetz/Development/vNNN'
os.environ['AWP_ROOT241'] = '/ott/mirror/fluent/develop/v241'

if not os.getenv("FLUENT_PROD_DIR"):
    # Create a session
    solver = pyfluent.launch_fluent(
        dimension=3, precision="double", processor_count=1, ui_mode="no_gui"  ## Windows: "hidden_gui"
    )

print('\nsolver.settings.file.batch_options.exit_on_error: ',
         solver.settings.file.batch_options.exit_on_error(),
      '\n')

solver.tui.file.set_batch_options('yes', 'yes', 'no')

print('\nsolver.settings.file.batch_options.exit_on_error: ',
         solver.settings.file.batch_options.exit_on_error(),
      '\n')

print('\n\nPreparing for the error...\n\n')

solver.scheme.eval('(cx-error-dialog "dummy error" "nothing")')
## solver.scheme_eval.exec('(cx-error-dialog "dummy error" "nothing")')

print('\n\nGoing on AFTER the error -- shouldn\'t...\n\n')

--->

Information: The server has started and is running.

> 
solver.settings.file.batch_options.exit_on_error:  False 

pyfluent.tui WARNING: Currently calling the TUI commands in a generic manner. Please run `python codegen/allapigen.py` from the top-level pyfluent directory to generate the local TUI commands classes.
The following solver settings object method could also be used to execute the above command:
<solver_session>.file.batch_options.exit_on_error = True

solver.settings.file.batch_options.exit_on_error:  True 



Preparing for the error...



Error: dummy error
Error Object: "nothing"
Traceback (most recent call last):
  File "/home/jschuetz/Development/PyFluent/exit_on_error/eoe.py", line 33, in <module>
    solver.scheme.eval('(cx-error-dialog "dummy error" "nothing")')
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/ansys/fluent/core/services/scheme_eval.py", line 374, in eval
    return self._eval(val, suppress_prompts)
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/ansys/fluent/core/services/scheme_eval.py", line 296, in _eval
    response = self.service.scheme_eval(request, metadata)
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/ansys/fluent/core/services/scheme_eval.py", line 97, in scheme_eval
    return self.__stub.SchemeEval(request, metadata=new_metadata)
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/grpc/_interceptor.py", line 277, in __call__
    response, ignored_call = self._with_call(
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/grpc/_interceptor.py", line 329, in _with_call
    call = self._interceptor.intercept_unary_unary(
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/ansys/fluent/core/services/interceptors.py", line 172, in intercept_unary_unary
    return self._intercept_call(continuation, client_call_details, request)
  File "/home/jschuetz/.venv/lib/python3.10/site-packages/ansys/fluent/core/services/interceptors.py", line 162, in _intercept_call
    raise new_ex from None
RuntimeError: Error Occurred!

[GautierR]

You have below an example of the Cortex error on can get from divergence in AMG solver:

turbulent viscosity limited to viscosity ratio of 1.000000e+05 in 246907 cells        
Stabilizing k to enhance linear solver robustness.
Divergence detected in AMG solver: k

I am also attaching fluent's transcript where you can see that even if the solver give error at the first iteration, it is still continuing.

fluent_transcript.txt

Fluent version is 2025 R1 and PyFluent 0.35.

By the way, the error I have (turbulence viscosity ratio exploding) is something I could not understand (I have investigate deeper and all the BCs and material properties for the setup are correct, there is no reasons for this to happens as mu_t = alpha^×(rhok)/omega). The only workaround I found is to save my fluent case and reload it just after (see below). For me this show some weakness for the current PyFluent workflow solution.

# Write Input Summary and Base Case
print("=== Write Input Summary and Base Case ===")
r.results.report.summary(write_to_file=True, file_name="report.sum")
r.file.write_case(file_name=f'{case_name}_base.cas.h5')

# Reload the saved Case (workaround to turbulent viscosity ratio exploding)
r.file.read_case(file_name=f'{case_name}_base.cas.h5')

But even if this issue is solved by this workaround I would like the simulation to stop and raise an error when a Cortex error occurs.

[ansjschuetz]

If writing and reading case & data changes behaviour in this way, that's probably some kind of a bug.
Can you provide (preferably small) case & data files for us to reproduce it?

But please first try to eliminate your UDF as the potential source of the unexpected behaviour...

In the transcript file you've shared, I don't see any trace of setting the "batch options" to exit immediately on any error.
Can you please introduce the lines from the python lines I shared into your script and try again?

[ansjschuetz]

Hello @GautierR , can you give an update on the status of this?

[GautierR]

The change on "exit_on_error" is not reported in the transcript, but it's there on my PyFluent script (most of the PyFluent commands are not display in the transcript, only the stdout we get in standard fluent are).

I am sending the current version of the script (run_case_reduced.py), you can see that I have commented the "read_case" command thus giving me the cortex error not catched by PyFluent. I tried to make a minimal folder for one specific case, you can run the script from the case folder: python ../run_case_reduced.py --src_dir=.. with a python env with PyFluent installed.

As you can see I start from scratch to avoid any issue from different .case file version. If I share with you a .case.h5 file you will not have the error because it only occurs when I start from scratch and do not reload the case.

Unfortunately, I don't have much time to investigate whether the issue comes from my UDF. In any case, I still think the error should be catch by PyFluent.

---

You can also see in my script that I faced other issues with PyFluent (I am not sure whether I should create new discussion or file a github issue for that):

1. Loading UDF using the API doesn't seem to be possible with the version 2025R1

2. For the wall BC, the conduction_layers qdot doesn't seems to work with a udf option:

wall_interface.thermal.conduction_layers[0].qdot.option = 'udf'
            To modify shell conduction model settings use define/models/shell-conduction/settings menu.
            Error: GENERAL-CAR-CDR: invalid argument [1]: improper list
            Error Object: ()

I tried with the other attribute thin_wall but this time it simply say that the object is not active even if wall_interface.thermal.enable_shell_conduction = True.