[Paraview] MPI Socket in use

Tue Apr 17 16:08:27 EDT 2012

Got it!!!!
Yes, there were, but I had not realized it before (not sure why they were floating around).  If I use the "correct" runmpi it now works.

Thank you much!

--Adam

On Apr 17, 2012, at 12:02 PM, Utkarsh Ayachit wrote:

> Are there multiple version of MPI installed by any chance?
> 
> Utkarsh
> 
> On Tue, Apr 17, 2012 at 2:51 PM, Adam Dershowitz
> <adershowitz at exponent.com> wrote:
>> Attached are the results of ldd libvtkParallel.so (looks to me like it is
>> linking MPI stuff).  And also CMakeCache.txt
>> 
>> 
>> 
>> 
>> 
>> -----Original Message-----
>> From: Utkarsh Ayachit [mailto:utkarsh.ayachit at kitware.com]
>> Sent: Tue 4/17/2012 6:14 AM
>> To: Adam Dershowitz
>> Cc: paraview at paraview.org
>> Subject: Re: [Paraview] MPI Socket in use
>> 
>> Do you mind posting your CMakeCache.txt file? Also do a "ldd
>> libvtkParallel.so". Let's verify that vtkParallel is linking against
>> MPI libs.
>> 
>> Utkarsh
>> 
>> On Tue, Apr 17, 2012 at 12:26 AM, Adam Dershowitz, Ph.D., P.E.
>> <adershowitz at exponent.com> wrote:
>>> I am sure that I use ccmake and changed PARAVIEW_USE_MPI to ON, (I also
>>> enabled python and pointed to the open mpi compiler, so that it then filled
>>> in most of the variables for MPI, as the note explains).  Then I did
>>> configure and generate.  Finally I did make, and sudo make install.  All
>>> seemed to work fine, and paraview runs fine, with a single processor.
>>> I even went so far and to make a new, empty directory, and rebuilt it
>>> there.  With the same results.
>>> I have also tried to explicitly put in the correct path, just to make sure
>>> that there isn't some other pvserver around:
>>> mpirun -np 3 /home/dersh/projects/ParaView-bin/bin/pvserver
>>> with exactly the same results.
>>> I have gone back, and now I see that MPIEXEC_MAX_NUMPROCS and
>>> VTK_MPI_MAX_NUMPROCS are set to 2.  But, I get the same error if I try to
>>> run it with -np 2, so clearly just using MPI is failing.  A
>>> 
>>> I realized that the problem is consistent with it not being built with
>>> MPI, but I definitely set it on.  Are there some other variables that have
>>> to be set?  Clearly something is not being built correctly, but I am not
>>> sure what it is.
>>> 
>>> Thanks,
>>> 
>>> --Adam
>>> 
>>> 
>>> 
>>> On Apr 16, 2012, at 7:09 PM, Utkarsh Ayachit wrote:
>>> 
>>>> You may want to verify PARAVIEW_USE_MPI flag again and ensure that the
>>>> pvserver you're running is indeed the one that has PARAVIEW_USE_MPI
>>>> set to ON. The problem you're seeing is typical when ParaView not
>>>> built with MPI.
>>>> 
>>>> Utkarsh
>>>> 
>>>> On Mon, Apr 16, 2012 at 8:10 PM, Adam Dershowitz
>>>> <adershowitz at exponent.com> wrote:
>>>>> I just built paraview on an opensuse linux box.  When I use the GUI, and
>>>>> a
>>>>> single connection it seems to work fine.  But, if I try to use multiple
>>>>> CPUs, or run with mpi, it fails.
>>>>> I do have OpenMPI installed.
>>>>> When I first started getting the error, I googled around some and found
>>>>> that
>>>>> maybe "make install" would help (I had been running it just in the build
>>>>> directory).  But I am getting the same error after installing.  I also
>>>>> added
>>>>> my openmpi libraries to LD_LIBRARY_PATH (when I first tried to run it I
>>>>> had
>>>>> other errors about a shared library).
>>>>> 
>>>>> I did build with PARAVIEW_USE_MPI set to on.  It looks as though one
>>>>> pvserver will run, but any additional ones finds a port in use.  Clearly
>>>>> something is not right about how MPI is being handled, but I am not sure
>>>>> how
>>>>> to fix it.
>>>>> 
>>>>> 
>>>>> If I try mpirun, here is error I get:
>>>>> 
>>>>> 
>>>>> mpiexec -np 3  pvserver
>>>>> Waiting for client...
>>>>> Connection URL: cs://cfd:11111
>>>>> Accepting connection(s): cfd:11111
>>>>> Waiting for client...
>>>>> Connection URL: cs://cfd:11111
>>>>> ERROR: In
>>>>> /home/dersh/projects/ParaView-3.14.1-Source/VTK/Common/vtkSocket.cxx,
>>>>> line
>>>>> 206
>>>>> vtkServerSocket (0xebd970): Socket error in call to bind. Address
>>>>> already in
>>>>> use.
>>>>> 
>>>>> ERROR: In
>>>>> 
>>>>> /home/dersh/projects/ParaView-3.14.1-Source/ParaViewCore/ClientServerCore/vtkTCPNetworkAccessManager.cxx,
>>>>> line 343
>>>>> vtkTCPNetworkAccessManager (0x661800): Failed to set up server socket.
>>>>> 
>>>>> Exiting...
>>>>> Waiting for client...
>>>>> Connection URL: cs://cfd:11111
>>>>> ERROR: In
>>>>> /home/dersh/projects/ParaView-3.14.1-Source/VTK/Common/vtkSocket.cxx,
>>>>> line
>>>>> 206
>>>>> vtkServerSocket (0xebd970): Socket error in call to bind. Address
>>>>> already in
>>>>> use.
>>>>> 
>>>>> ERROR: In
>>>>> 
>>>>> /home/dersh/projects/ParaView-3.14.1-Source/ParaViewCore/ClientServerCore/vtkTCPNetworkAccessManager.cxx,
>>>>> line 343
>>>>> vtkTCPNetworkAccessManager (0x661800): Failed to set up server socket.
>>>>> 
>>>>> Exiting...
>>>>> 
>>>>> -----------------------------------------------------------------------------
>>>>> It seems that [at least] one of the processes that was started with
>>>>> mpirun did not invoke MPI_INIT before quitting (it is possible that
>>>>> more than one process did not invoke MPI_INIT -- mpirun was only
>>>>> notified of the first one, which was on node n0).
>>>>> 
>>>>> mpirun can *only* be used with MPI programs (i.e., programs that
>>>>> invoke MPI_INIT and MPI_FINALIZE).  You can use the "lamexec" program
>>>>> to run non-MPI programs over the lambooted nodes.
>>>>> 
>>>>> -----------------------------------------------------------------------------
>>>>> mpirun failed with exit status 252
>>>>> 
>>>>> And suggestions would be greatly appreciated.
>>>>> 
>>>>> thanks,
>>>>> 
>>>>> --Adam
>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> Powered by www.kitware.com
>>>>> 
>>>>> Visit other Kitware open-source projects at
>>>>> http://www.kitware.com/opensource/opensource.html
>>>>> 
>>>>> Please keep messages on-topic and check the ParaView Wiki at:
>>>>> http://paraview.org/Wiki/ParaView
>>>>> 
>>>>> Follow this link to subscribe/unsubscribe:
>>>>> http://www.paraview.org/mailman/listinfo/paraview
>>>>> 
>>> 
>> 