[Paraview] Problem with big files

[Christoph Moder]

Hi!

> In my opinion 2 Gig isn't enough for a 21 million unstructured cell
> dataset. If you have the option of a mini-cluster (4, 8 nodes) then
> that's the way to go. :)

Ok, good to know. I thought, if the binary file is only 800 MB big, 
about 1 GB RAM should be enough. But it didn't work even on a PC with 3 
GB RAM.

We have already thought about running it in parallel, but there are 
problems with MPI (using the MPICH library). The program crashes even 
with much smaller files, with a different error message (see below), and 
I have also no idea why.

> Is the 21 million cell dataset coming from a simulation that you are
> running at your institution? Is that simulation running on a cluster?
> Can you put paraview server on that cluster? 

At the moment, it is just a test file. But we want to use Paraview for 
simulations in geodynamics and seismology; the cluster for these 
simulations will be ready in a few weeks.

Of course it should be possible to run Paraview there, when the MPI 
problem is solved.

> At Sandia we don't usually put more than 2 million unstructured cells
> per paraview server process, that way operations and interactions are
> fairly responsive.

Also good to know.

Thank you very much, cheers,
Christoph

Typical MPI error messages:

#1:
==========
$ mpirun -machinefile ~/machinefile.txt -np 2 ./paraview
p1_27474:  p4_error: alloc_p4_msg: Message size exceeds P4s maximum 
message size: 321213977
rm_l_1_27475: (257.175131) net_send: could not write to fd=5, errno = 32
==========

#2:
==========
$ mpirun -machinefile ~/machinefile.txt -np 2 ./paraview
Process id: 1 >> ERROR: In 
/home/moder/paraview-3D/paraview-2.4.2-clone2/VTK/Rendering/vtkXOpenGLRenderWindow.cxx, 
line 1319
vtkXOpenGLRenderWindow (0xb576c78): bad X server connection.

p1_22329:  p4_error: interrupt SIGSEGV: 11
rm_l_1_22330: (1360.557878) net_send: could not write to fd=5, errno = 32
==========

What could be wrong?