OK, you might want to look at the command line in the qsub script though
since it doesn't make use of the environment variables. Not sure if
that's really your issue though.
I've used PV with openmpi, mvapich2, mpich2, cray mpt, and intel mpi.
Very rarely seen any interactions between pv and various
implementations. Once or twice hit some weird memory bugs in mvapich.
On 02/20/2012 10:14 AM, Rakesh Hammond wrote:
Alan, Burlen
Thanks for the prompt response - I will try with the master of 3.14.0
Burlen - We don't need to do tunnelling for this cluster (we did for
other clusters), because the compute nodes of viz cluser are visible
across the network (ie I am able to ssh from a compute node on the viz
cluster back to my workstation).
I am now pinning my hope on testing with a different MPI than Intel -
Can I ask what MPI do you use with ParaView?
Regards,
Rakesh
On 20 February 2012 18:03, Scott, W Alan <[email protected]
<mailto:[email protected]>> wrote:
I know that there were some connection issues with “older”
versions of PV 3.14.0. This problem would be in the source for
3.14.0-RC2. Give master a try, and see if that helps. We did
find the PV_CLIENT_HOST variable problem – it is fixed in master.
Alan
*From:*[email protected]
<mailto:[email protected]>
[mailto:[email protected]
<mailto:[email protected]>] *On Behalf Of *Rakesh Hammond
*Sent:* Monday, February 20, 2012 9:54 AM
*To:* [email protected] <mailto:[email protected]>
*Subject:* [EXTERNAL] [Paraview] Paraview-3.12.0 Client-Server
Hi,
I am trying to get ParaView-3.12.0 working on our visualisation
cluster. However I am getting errors such as the following
"Socket error in call to conect : Connection refused"
As a bit of a background, I have ParaView-3.10.1 working in
client-server mode on the same cluster. On the server, we have
IntelMPI-4.0.1 and Slurm as the scheduler.
I tried debugging the problem by printing stuff out in
vtkSocket.cxx file - I always get the message which says
"Connected" from slurm out file. The ParaView-3.12.0 client then
just hangs.
I did get it to work by using a machinefile (ie outside slurm) and
it worked in client-server to my work station.
I am not really sure what is the cause of this - my suspicion is
that this either to do with IntelMPI-4.0.1 and ParaView-3.12.0 or
with Slurm-2.1.4 (which is what is available on the cluster).
Can some one please give me some hints how I can debug this further?
Any pointers are greately appreciated.
Given my level of desperation, I tried with Paraview-3.14.0-RC2
and things have really changed here - for example, what happend to
the PV_CLIENT_HOST variable? Is this no longer set?
I also note that when you set up a profile to connect to the
server in reverse-connection mode, it no longer allows you to
specifiy host name - is this intentional?
Regards,
Rakesh
_______________________________________________
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
_______________________________________________
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
