Residual error for big mess (||r|| = 0.000000e+00)

mesh
1

Hi,

I have a problem with my program when it comes to big mesh.
I run the program below for permeability 1.3610**(-9) to 1.3610**(3), the error occurs only when the value is >= 1.36*10**(-1). Everything works fine for a small mesh. The mesh has been generated from Gmsh.

A small part of the program:

# Function spaces
V = VectorElement("Lagrange", mesh.ufl_cell(), 2)
Q = FiniteElement("Lagrange", mesh.ufl_cell(), 1)
TH = V * Q
W = FunctionSpace(mesh, TH)

K_p = 1.36*10**(-1)
K_interface = 1.36*10**(-1) 
K_wall = 1.36*10**(-9) 
mu = 0.00496
(u, p) = TrialFunctions(W)
(v, q) = TestFunctions(W)
a = (2*mu*inner(epsilon(u), epsilon(v)) + div(v) * p + q * div(u)) * dx(mark["v_domain"]) \
    + ((mu/ K_p) * inner(u, v) + inner(grad(p), v) + q * div(u)) * dx(mark["p_domain"]) \
    + ((mu/ K_wall) * inner(u, v) + inner(grad(p), v) + q * div(u)) * dx(mark["wall_domain"]) \
    + ((mu/ K_interface) * inner(u, v) + inner(grad(p), v) + q * div(u)) * dx(mark["interface_domain"])
    + (p*inner(v,n_s))*ds(mark["v_domain"])
L = inner(f_p, v) * ds(mark["p_domain"]) + inner(f_v, v) * ds(mark["v_domain"])
w = Function(W)
solve(a == L, w, bcs)
(u, p) = w.split(True)

And the error:

Traceback (most recent call last):
  File "case_2_2d.py", line 143, in <module>
    solve(a==L, w, bcs)
  File "/cluster/shared/fenics/nobackup/install/fenics-2019.1.0-2020-02-19/lib/python3.6/site-packages/dolfin/fem/solving.py", line 220, in solve
    _solve_varproblem(*args, **kwargs)
  File "/cluster/shared/fenics/nobackup/install/fenics-2019.1.0-2020-02-19/lib/python3.6/site-packages/dolfin/fem/solving.py", line 247, in _solve_varproblem
    solver.solve()
RuntimeError:

*** -------------------------------------------------------------------------
*** DOLFIN encountered an error. If you are not able to resolve this issue
*** using the information listed below, you can ask for help at
***
***     fenics-support@googlegroups.com
***
*** Remember to include the error message listed below and, if possible,
*** include a *minimal* running example to reproduce the error.
***
*** -------------------------------------------------------------------------
*** Error:   Unable to successfully call PETSc function 'KSPSolve'.
*** Reason:  PETSc error code is: 76 (Error in external library).
*** Where:   This error was encountered inside /cluster/shared/fenics/nobackup/build/fenics-2019.1.0-2020-02-19/dolfin/dolfin/la/PETScKrylovSolver.cpp.
*** Process: 0
***
*** DOLFIN version: 2019.1.0
*** Git changeset:  86617f912d4035520fa988d6af07a19d6069d5ee
*** -------------------------------------------------------------------------

Thank you in advance.

2

This error usually means that you are running out of memory. Try using iterative solvers or mumps. See for instance: Solving PDEs in Minutes - <br> The FEniCS Tutorial Volume I

3

Thanks for the answer.

I’m running the program in an HPC, so memory should not be an issue.
I have tried with mumps, preconditioners and parallel but still the same error.

any other tips?

4

Other things that helps on the assembly side is to wrap all scalar values as Constants, as it saves you from recompiling every time you change a parameter. I would carefully monitor the memory usage of your program, and at least use mumps. Especially since it works on a smaller mesh. It might be that you are hitting umfpacks 4 GB limit: https://fenicsproject.org/qa/4177/reason-petsc-error-code-is-76/

5

I had a similar issue with same error output. The reason is my huge mesh (DOF>100M) has several zero volume elements. You can use below code to check

import fenics as fe
# load mesh to fe_mesh
vols = [cell.volume() for cell in fe.cells(fe_mesh)] # get volumes
vol_st = sorted(tuple(zip(vols, range(len(vols))))) # sort from smallest to largest
for ver in fe.vertices(fe.Cell(fe_mesh, vol_st[0][1])): # print out the smallest element
    print(ver.point()[:], end=', ')
print(vol_st[0])

Ming

1 Like