Learning Forum

[kaeden.freston]

I’m seeing repeat failures of Fluent–Mechanical coupling that don’t appear model-specific.

Environments

• Local (Windows): Ansys 2025 R1 (v251) via Workbench.

• Cluster (Linux/SLURM): Ansys 2023 R1 (v231), single-node jobs, System Coupling shows Intel MPI.

Symptoms

• Local v251: The stock Reed Valve FSI Simulation tutorial (no edits) fails during initial participant connection; System Coupling reports that one participant exited during handshake.

• Cluster v231: My minimal two-way FSI prototype launches and connects, Mechanical begins solving, then System Coupling aborts on the first data exchange with “Low-level communication error −4: Write error” while making a remote call to Fluent.

What I’ve tried

• Reproduced with both my prototype and the unmodified Ansys tutorial.

• Multiple runs on both environments with the same outcome.

• No custom MPI flags; default SC-assigned ports; single node.

Ask

• Known causes/fixes for error −4 (Write error) when SC calls Fluent?

• Recommended diagnostics (Fluent/SC logging flags, environment variables) to pinpoint socket/RPC shutdowns.

• Any version-specific issues with v251/v231 affecting SC ↔ Mechanical/Fluent handshake?

• MPI/fabric or port/hostname settings you recommend for single-node SC runs (e.g., loopback vs host IP, firewall notes)?

I can upload the SC/Fluent/MAPDL logs (local and cluster) if needed.

Thank you

[Rahul]

Share SC/Fluent/MAPDL logs images

[kaeden.freston]

Here are the output logs. Please let me know if there's anything else I can provide. Thank you for your help. 

[Rahul]

Is this log for your prototype or ansys tutorial? 

From logs, it looks like it is failing in Fluent with the error: chip-exec function 'none' not found. 
Are you using UDMI in your UDF? If you are using UDMI, please set User-defined memory.

[kaeden.freston]

Thanks for your reply. This log is for my prototype. Just to clarify, I was not originally using a UDF — today, I tried experimenting with a “noop” Scheme script because of the repeated "Error: chip-exec: function "none" not found" messages during iterations. That was a workaround I tried based on advice that this might be an Ansys bug, but I've only been able to define the function on the host side.

Since I never had a UDF to begin with, I’m not using UDMI or any user-defined memory. From what I can tell, the error is coming from some execute-commands hook in the case/data that’s trying to call chip-exec none every iteration.

Could you advise:

• What’s the correct way to check whether UDMI is expected/required for this case?

• If no UDFs are involved, is there a way to disable or clear the execute-commands list so nothing tries to call "none"? I've already tried deleting all exisiting commands but there are none to delete. 

• Or is there a recommended way to cleanly define a placeholder when no UDF is present?

I want to make sure I’m not mis-configuring my setup or missing a step with user-defined memory allocation. Thank you

[Rahul]

You mentioned that the Reed Valve FSI Simulation tutorial also fails with this error. Share logs from that run to debug the issue.

[kaeden.freston]

Here are the logs from running the Ansys Tutorial Reed Valve FSI Co-simulation. Please let me know if there's anything else I can provide. This error appears to be almost identical to that of my prototype. 

[Rahul]

Are you able to solve the Fluent only simulation (without coupling)? Let's run this to isolate the issue to see if it's coming from Fluent or system coupling.

[kaeden.freston]

I ran the Fluent-only simulation on the cluster (no System Coupling, just fluent.cas.gz). After Fluent finishes “Building…” I still get:

 

Error: CAR: invalid argument [1]: wrong type [not a pair] Error Object: ()

This happens before I even try to iterate. When trying to iterate, I get the same chip-exec function error.

• I confirmed there are no UDFs or UDMI in this model.

• Execute-Commands shows no entries.

• Disabling residual monitors and saving/reloading did not resolve it.

• I also reproduced the same CAR error using the official Reed Valve tutorial.

• Importantly, I’ve submitted Fluent-only cases to the cluster in the past without any issues. The error only shows up with cases exported as part of the two-way FSI workflow (with Mechanical and System Coupling).

[Rahul]

As this issue is just happening while running the files exported for system coupling analysis, could you please share the settings used?

[kaeden.freston]

Fluent setup used in my two-way FSI workflow:

• General

  • Time: Transient

• Materials

  • Imported blood material

• Cell Zone Conditions

  • Fluid region changed from air → blood

• Boundary Conditions

  • Wall: wall

  • Inlet: velocity inlet

    • Velocity profile with average velocity chosen so that wall shear stress ≈ 1.5 Pa

  • Outlet: pressure outlet

    • Gauge Pressure = 0.0 Pa

• Dynamic Mesh

  • Dynamic mesh enabled

  • Mesh methods: Smoothing + Remeshing

    • Smoothing:

      • Method = Diffusion

      • Diffusion function = boundary-distance

      • Diffusion parameter = 1

    • Remeshing:

      • Methods = Local Cell, Local Face, Region Face

      • Parameters: Use Defaults, Size Remeshing Interval = 1

  • Dynamic Mesh Zones:

    • Inlet → Stationary

    • Outlet → Stationary

    • Wall → System Coupling

      • Solver Options: Activate Solution Stabilization

        • Scale Factor = 0.5

        • Method = coefficient-based

    • Region → Deforming

      • Meshing Options: Smoothing + Remeshing

      • Zone parameters:

        • Minimum length scale = 1.1 × (zone scale info min)

        • Maximum length scale = 0.9 × (zone scale info max)

• Solution → Monitors → Residuals

  • Convergence criteria:

    • Continuity = 1e-5

    • X-, Y-, Z-velocity = 1e-5

• Solution → Run Calculations

  • Time step size = 1

  • Number of time steps = 1

  • Max iterations per step = 400