[Paraview] Using paraview with pvserver

Herwig Zilken h.zilken at fz-juelich.de
Tue Mar 5 11:14:39 EST 2013


Dear Patrick,

unfortunately I have no solution for your problem, but I want to say
that we run into the same problems at the moment with parallel ParaView
to visualize big data sets.
In sequential mode and with smaller datasets ParaView is more or less
stable.

But when we switch to parallel mode, ParaView crashes a lot in our tests.
We also see the strange behavior of ParaView to multiplicate the memory
footprint of the data in parallel mode. As far as I know that depends on
the reader (on the data format).
The VTK legacy reader seems not to  be able to distribute data parts to
the ParaView processes, instead all processes read all data, which is
really cumbersome for big data visualization.
At the moment I really wonder what the best data format (reader) is to
distribute big data equally on the paraview processes. (For structured
and for unstructured data.)
Hdf5/Xmdf? VTK XML?

I hope that there is a solution, so that big data visualization on a
visualization cluster is possible with ParaView for real world data :) .


Regards,

Herwig





> but runing pvserver in parallel on the server does'nt.
> If I launch :
> mpirun -np 4 /share/apps/paraView-server-3.08.1/bin/pvserver
> (with or without --use-offscreen-rendering) it crashes after loading
> the datas, while rendering (I think).
>
> On the client the message is:
> ERROR: In
> /home/utkarsh/Dashboards/MyTests/NightlyMaster/ParaViewSuperbuild-Release/paraview/src/paraview/VTK/Parallel/Core/vtkSocketCommunicator.cxx,
> line 812
> vtkSocketCommunicator (0x177a380): Could not receive tag. 188969
>
> On the server, the message is:
> mpirun noticed that process rank 2 with PID 25736 on node kareline
> exited on signal 9 (Killed).
>
> I also noticed some strange behavior:
> 1) after selecting the file to read and before  clicking on the apply
> button, 3 of the 4 pvserver processes use 100% cpu.
> 2) after clicking the apply button, the 4 pvserver processes uses the
> same memory amount than the sequential run (8Gb each)
> So I think something is wrong in building my parallel version.




------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------


More information about the ParaView mailing list