[Paraview] inquiries on parallel pvserver nodes

Thu Oct 15 05:35:06 EDT 2009

Hi guys,

an update of my problem..

if i wait long enough, the error message pops up and the server konsole.

[pc226][[25673,1],4][btl_tcp_endpoint.c:631:mca_btl_tcp_endpoint_complete_connect]
connect() failed: Connection timed out (110)
[pc226][[25673,1],5][btl_tcp_endpoint.c:631:mca_btl_tcp_endpoint_complete_connect]
connect() failed: Connection timed out (110)

since the server receives the initial call from the paraview, why wouldn't
it get a response from the paraview later as mentioned above??

Thanks again

On Wed, Oct 14, 2009 at 6:01 AM, yewyong <uyong81 at gmail.com> wrote:

> Hi Ken and All,
>
> i've solved the problem with my openmpi build.
> below are an example of it.
>
> yewyong at vrc1:~/installer/mpi_test> mpirun -np 14 -machinefile quad
> ./hello_parallel
> hello_parallel.cc: Number of tasks=14 My rank=5 My name=vrcwr220.
> hello_parallel.cc: Number of tasks=14 My rank=7 My name=pc249.
> hello_parallel.cc: Number of tasks=14 My rank=9 My name=pc205.
> hello_parallel.cc: Number of tasks=14 My rank=1 My name=vrc1.
> hello_parallel.cc: Number of tasks=14 My rank=3 My name=vrc1.
> hello_parallel.cc: Number of tasks=14 My rank=10 My name=pc226.
> hello_parallel.cc: Number of tasks=14 My rank=13 My name=vrcwr228.
> hello_parallel.cc: Number of tasks=14 My rank=0 My name=vrc1.
> hello_parallel.cc: Number of tasks=14 My rank=4 My name=vrcwr220.
> hello_parallel.cc: Number of tasks=14 My rank=2 My name=vrc1.
> hello_parallel.cc: Number of tasks=14 My rank=8 My name=pc205.
> hello_parallel.cc: Number of tasks=14 My rank=6 My name=pc249.
> hello_parallel.cc: Number of tasks=14 My rank=12 My name=vrcwr228.
> hello_parallel.cc: Number of tasks=14 My rank=11 My name=pc226.
>
> however, my problem still exist. my paraview doesn't seems to give any
> response when i'm trying to connect
> to the pvserver.
> *
> Starts the pvserver*
> yewyong at vrc1:~/installer/mpi_test> mpirun -np 14 -machinefile quad
> pvserver --use-offscreen-rendering
> Listen on port: 11111
> Waiting for client...
> Client connected.
>
> *And the client shows....*
> yewyong at pc226:~> paraview
> Waiting for server...
>
> and the paraview "hangs" with a "choose server" window. (no buttons are
> functioning)
> i had to go to the konsole to "ctrl+c" to actually exit the paraview.
>
> I tried having different sequence in machinefile to check whether it's the
> problem with different IP domain
>
> *my machine file (quad):*
> 202.185.77.110 slots=4 max-slots=4 (master node with 2 ethernet port;
> another is 192.168 seires)
> 192.168.0.220 slots=2 max-slots=2
> 192.168.0.205 slots=2 max-slots=2
> 192.168.0.249 slots=2 max-slots=2
> 202.185.77.219 slots=2 max-slots=2
> 202.185.77.220 slots=2 max-slots=2
>
> as per my findings, paraview will have problem connecting to pvserver
> whenever pvserver is build from nodes
> from different IP domain.
>
> ps, even with hamachi (vpn), when each nodes are having 5.xx.xx.xx series
> as IP, the problem persists.
>
> sorry for the lengthy "story telling" mail.
>
> appreciate all the help i can get.
>
> Thanks in advance.
>
> yewyong
>
>
>
>
>
> On Thu, Oct 8, 2009 at 1:22 AM, yewyong <uyong81 at gmail.com> wrote:
>
>> Hi Ken,
>>
>> Thanks for the swift reply, and for the reminder.
>> forgot to try out the mpirun before trying it out in pvserver.
>> tried a very simple "hello world" mpi job today. (to no avail)
>>
>> at first, i thought it was because of the difference of 192 and 202 domain
>> which causes the pause, then i applied hamachi (vpn tool) to all the nodes,
>> giving them a uniform LAN domain IPs, but still the pauses happens when
>> reaching np > 10.
>>
>> that all boils down to my openmpi build? but i am able to run >10 nps
>> pvserver with the same openmpi build when having the nodes when in a LAN
>> environment (not using hamachi)..
>>
>> no worries, i'm using the cvs version of the paraview build. (it states
>> "paraview 3.7 development" at the title window)
>>
>>
>> thanks again for all your advice in advance.
>>
>>
>> On Tue, Oct 6, 2009 at 10:06 PM, Moreland, Kenneth <kmorel at sandia.gov>wrote:
>>
>>>  I count 10 cores in the 192.168 domain.  My suspicion is that MPI is
>>> having a problem when running with more than 10 processes and therefore
>>> having to run on nodes in both 192.168 and 202.185.
>>>
>>> My advice is to first make sure that you can run any MPI job across all
>>> the nodes in your makeshift cluster.  Do this using a test MPI program that
>>> is simpler than pvserver.  Most MPI implementations come with a set of
>>> tests.  If you can’t find any on your system, there are many available for
>>> download; here is a good site:
>>> http://www.mcs.anl.gov/research/projects/mpi/mpi-test/tsuite.html
>>>
>>> By the way, are you really running a program called pvclient?  There has
>>> been no such executable since ParaView 2.6, which was released 2 years ago.
>>>  If that is the case, I recommend updated to the latest version of ParaView.
>>>
>>> -Ken
>>>
>>>
>>>
>>> On 10/6/09 6:22 AM, "yewyong" <uyong81 at gmail.com> wrote:
>>>
>>> Hi all,
>>>
>>> am trying to utilize some desktops in our lab (say "lab A") to visualize
>>> some virtual heart.
>>> on top of that, there are also some unused desktop from other lab ("lab
>>> B") which i plan to add in as the server.
>>> below are my "hostfile" used during mpirun.
>>>
>>> 192.168.0.10 slots=4 max-slots=4
>>> 192.168.0.220 slots=2 max-slots=2
>>> 192.168.0.205 slots=2 max-slots=2
>>> 192.168.0.249 slots=2 max-slots=2
>>> 202.185.77.219 slots=2 max-slots=2
>>> 202.185.77.220 slots=2 max-slots=2
>>> 202.185.77.221 slots=2 max-slots=2
>>>
>>> as we can see, the 192.168... series are from lab A and the 202.185..
>>> series are from lab B.
>>> the experiment started off fine as i'm also using a desktop from lab A
>>> (192.168.. series) to connect to the server.
>>> when we reach np 11, the paraview (pvclient) fails to response when we
>>> try to establish a connection with the server.
>>>
>>> are we missing out anything here?
>>>
>>> appreciate all your response and answer in advance
>>>
>>> Thanks
>>>
>>> yewyong
>>>
>>>
>>>
>>>
>>>
>>>    ****      Kenneth Moreland
>>>     ***      Sandia National Laboratories
>>> ***********
>>> *** *** ***  email: kmorel at sandia.gov
>>> **  ***  **  phone: (505) 844-8919
>>>     ***      web:   http://www.cs.unm.edu/~kmorel<http://www.cs.unm.edu/%7Ekmorel>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20091015/ab4a71cc/attachment.htm>