bad initial partitioning

[martijn]

Hi,
I created a mesh using gmsh, which imports fine in Elmer, but when I start to run the solver (NS equation) Elmer says:

Code: Select all

Bad initial partitioning: 6 nodes do not belong anywhere!
MeshSplitter failed - aborting

The geometry consists of a tube with pillars in it. I would like to simulate an airflow through this pipe. The number of 6 nodes caused me to suspect the centres of the top and bottom planes of the pillars, since these are not part of the body.
After some experimenting I found that the problem only occurs when I use the MPI solver (enabled, with 2 processes, since I have a dual core CPU). When I try a simpler geometry with one pillar/hole, I get the same results: runs fine with a single core, but complains about 2 nodes not belonging anywhere when I enable the MPI solver.
What am I doing wrong? I attached the simple problem to this message.

[raback]

Hi,

Maybe you have some nodes that are not related to any of the bulk elements? The partitioning algorithm can't seem to be able to handle those as they are not included in the graph. You could try to include the string '-removeunused' in the Parallel Settings / Divide box. This make ElmerGrid plug-in to eliminate the nodes that are not needed by any higher dimensional element. For example,

Code: Select all

ElmerGrid 2 2 %msh -metis %n -removeunused

I Hope this helps.

-Peter

[martijn]

That works, thanks!