[Paraview] Re: Documentation for Paraview Distributed Options?

[Joseph Insley]

Jeremy,

This problem looks very familiar, I was running into similar issues.  
Although, I am using mpirun to start both the server and the 
render-server, so I'm not certain whether this solution will work for you 
or not, but it's probably worth a try.  

When I start the --render-server, I also use the --machines=node_file 
option.  node_file contains the list of nodes that the render-server will 
run on, format of the file is one hostname per line.  In your case, 
node_file would just contain COMPUTER2.  So my commandline looks something 
like:
mpirun -np 8 -machinefile $PBS_NODEFILE /soft/bin/paraview --render-server --machines=$PBS_NODEFILE
 
If I understand the process correctly, when the client connects to the 
render-server, it gets this list of machines then passes it on to the 
server, who uses it to connect to the render-server node(s).  

I hope that helps.  
joe.

On Thu, 9 Sep 2004, Jeremy Stout wrote:

>After doing a lot of experimentation yesterday, I believe I found the
>commands I need to use, but have come across a new problem (see
>below). Let's say I want to run the client and server on COMPUTER1 and
>the render server on COMPUTER2. I think I need to use the following:
>
>COMPUTER2:
>paraview --render-server
>
>COMPUTER1:
>paraview --server
>paraview --client-render-server --render-server-host=COMPUTER2 --host=localhost
>
>The problem with this arrangement is that the render-server on
>COMPUTER2 thinks the server is running on COMPUTER2 and waits for a
>connection from it. Also, it seems that the server program is looking
>for the render server program locally. Needless to say, that doesn't
>work and I get the following error message from the server program:
>
>Process id: 0 >> ERROR: In
>/usr/local/packages/paraview-1.6.1/VTK/Parallel/vtkSocketCommunicator.cxx,
>line 453
>vtkSocketCommunicator (0x9737b78): Can not connect to localhost on port 32850
>
>In case it helps, here is the output from each of the programs:
>
>[stout at COMPUTER2 stout]$ paraview --render-server
>Listen on port: 22221
>RenderServer: Waiting for client...
>connected to port 22221
>Client connected.
>open with port 0
>WaitForConnection: id :0  Port:32850
>
>[stout at COMPUTER1 stout]$ paraview --server
>Listen on port: 11111
>Waiting for client...
>connected to port 11111
>Client connected.
>Connect: id :0  host: localhost  Port:32850
>Process id: 0 >> ERROR: In
>/usr/local/packages/paraview-1.6.1/VTK/Parallel/vtkSocketCommunicator.cxx,
>line 453
>vtkSocketCommunicator (0x9737b78): Can not connect to localhost on port 32850
>
>Process id: 0 >> ERROR: In
>/usr/local/packages/paraview-1.6.1/VTK/Parallel/vtkSocketCommunicator.cxx,
>line 536
>vtkSocketCommunicator (0x9737b78): Could not send tag.
>
>(repeat the error message ad nauseum)
>
>[stout at COMPUTER1 stout]$ paraview --client-render-server 
>--render-server-host=COMPUTER2 --render-port=22221 --host=localhost
>Connect to localhost:11111
>Connect to node27:22221
>
>What do I need to do in order to get the server and render server to
>connect to one another? I've tried lots of different combinations of
>commands without any success.
>
>Jeremy Stout
>
>
>On Wed, 8 Sep 2004 10:25:00 -0400, Jeremy Stout <stout.jeremy at gmail.com> wrote:
>> After resolving yesterday's issue, I'm moving on and trying to figure
>> out how to use Paraview on my organization's computer cluster. I've
>> been looking through the 1.4 User's Guide at the different options
>> that can be given on the commandline. Unfortunately, some of the
>> explanations for the options I am interested in are sparser than what
>> I would like.
>> 
>> Before I go any farther, let me briefly explain my computer cluster
>> setup. It consists of one main node and 20+ computation nodes. The
>> main node has Internet access and is primarily used to submit jobs to
>> the cluster. The computation nodes can communicate with the main node
>> and each other. MPI is used to provide parallel support.
>> 
>> I'm primarily interested in using my computer cluster for rendering
>> work. I'm thinking that I will have to setup one of the computation
>> nodes as a render server and have the other nodes connect to it. I
>> would then have the server and client run on the main node.
>> 
>> Looking through the options for the program, I am not sure how to make
>> this happen. So, let me ask the following questions:
>> 
>> 1.  If I want to run the rendering process on 4 nodes, what options
>> would I use to do that? Do I have to tell one node to be a "render
>> server" and the others to be a "render node" that connects to the
>> render server? In the user's guide, I saw mention of a "machine file,"
>> but I didn't see any mention made for the format of the file or how it
>> was supposed to be used.
>> 
>> 2.  Going through the mail archives, I saw mention of support being
>> integrated for a render server/server/client mode. What would be the
>> options for the client to connect if the servers are running on
>> different machines? It seems like "host" only lets you specify the
>> server you are connecting to.
>> 
>> Any help that can be offered would be appreciated.
>> 
>> Jeremy Stout
>>
>_______________________________________________
>ParaView mailing list
>ParaView at paraview.org
>http://www.paraview.org/mailman/listinfo/paraview
>
>

====================================================================
joseph a. insley				insley at mcs.anl.gov
mathematics & computer science division		(630) 252-5649
argonne national laboratory                     (630) 252-5986 (fax)
http://www.mcs.anl.gov/~insley
====================================================================