[Paraview] EXTERNAL: Re: EXTERNAL: Re: EXTERNAL: Re: CxxOverlappingAMRExample with Catalyst Live

Tue Apr 5 13:21:31 EDT 2016

Hi Andy,

Your explanation made it a lot clearer for me. I tested it out with multiple pvservers running on a single node and I was able to connect with my simulation. Success!

However, when I tried running pvservers across more than one node, the Catalyst connection failed. I saw a bunch of error messages that look like this:

ERROR: In <src_dir>/VTK/Common/System/vtkSocket.cxx, line 481
vtkClientSocket (…): Socket error in call to connect. Connection refused.

ERROR: In <src_dir>/VTK/Common/System/vtkClientSocket.cxx, line 53
vtkClientSocket (…): Failed to connect to server localhost:22228

ERROR: In <src_dir>/VTK/Parallel/Core/vtkSocketCommunicator.cxx, line 714
vtkSocketCommunicator (…): Can not connect to localhost on port 22228

There were a bunch more of these with various socket numbers.

Do you have any idea what might be causing this problem?

In my coproc.py file I have a line that looks like this:

    coprocess.DoLiveVisualization(datadescription, “localhost”, 22222)

Do I have to call DoLiveVisualization differently when running across multiple nodes? Do I need to set the host differently maybe? Or is the problem more likely to be at my end?

Many thanks

Kind regards,
Jim

From: Andy Bauer [mailto:andy.bauer at kitware.com]
Sent: 02 April 2016 15:51
To: Eliot Jim AWE <Jim.Eliot at awe1.awe.co.uk>
Cc: paraview at paraview.org
Subject: EXTERNAL: Re: EXTERNAL: Re: EXTERNAL: Re: [Paraview] CxxOverlappingAMRExample with Catalyst Live

Hi Jim,
I inlined some response below...
Best,
Andy

On Fri, Apr 1, 2016 at 12:41 PM, <Jim.Eliot at awe.co.uk<mailto:Jim.Eliot at awe.co.uk>> wrote:
Hi Andy,

Thank you for all the help that you have provided.

I feel like I don’t quite understand how Catalyst handles the simulation data. If I use ParaView Catalyst for large parallel simulations, and I try to fetch the entire dataset, does all the data in the problem get sent to the client on my workstation?

The way it works now is that when you click on the icon next to a filter, it fetches that data set to whichever pvserver it's connected to, including the built-in server. So yes, in this case it would try to fetch all of the data to your client. We've done some work on generating images through Catalyst and then sending that image to the client. That was done with PyFR.

If I am running pvserver in parallel, would that be able to distribute the data between the processes? How do I tell Catalyst to connect to the pvservers and the pvservers to connect to the client? I guess I can use –client-host and –server-port on the pvservers and point them to the paraview client but how do I make the Catalyst -> pvserver connection?

Yes, that should attempt some form of partitioning when it's sent from the sim/Catalyst to the parallel pvserver. The way to do the connection in this case is to tell the Catalyst Python script to link to where pvserver is running ( coprocessor.DoLiveVisualization(datadescription, "localhost", 22222)). Then, run the simulation, pvserver and the GUI client, have the client connect to pvserver and then in the GUI go through the Catalyst->Connect... point to where the simulation is running. This should connect pvserver to the sim. If there's firewalls then it may be a bit more complex connection. I don't know to what level/size data this has been pushed to.

Thank you for the information regarding AMR grids – it sounds like this approach will work for my application.

Kind regards,
Jim

From: Andy Bauer [mailto:andy.bauer at kitware.com<mailto:andy.bauer at kitware.com>]
Sent: 31 March 2016 20:18
To: Eliot Jim AWE <Jim.Eliot at awe1.awe.co.uk<mailto:Jim.Eliot at awe1.awe.co.uk>>
Cc: paraview at paraview.org<mailto:paraview at paraview.org>
Subject: EXTERNAL: Re: EXTERNAL: Re: [Paraview] CxxOverlappingAMRExample with Catalyst Live

Hi Jim,
Yes, you can have a Catalyst instrumented code connect to something running pvserver and have the pvserver connect to the client. At the end of the Python generated scripts there should be a line that looks like:
    coprocessor.DoLiveVisualization(datadescription, "localhost", 22222)
Here you just have to edit it to point to something other than localhost.

As for overlapping structured grids, there's no convenient composite/multiblock data set for that like there is for Cartesian grids. What you would need to do is use vtkStructuredGrids with blanking done manually and then just add them in as blocks in a vtkMultiBlockDataSet. It won't have the structure that the AMR specific composite data sets have so it may be a bit slower for some filters but it should work pretty much like you want for most stuff. Please let us know if you try this out and get unexpected/undesired behaviour.
Best,
Andy

On Thu, Mar 31, 2016 at 12:05 PM, <Jim.Eliot at awe.co.uk<mailto:Jim.Eliot at awe.co.uk>> wrote:
Hi Andy,

Thanks for your reply. Currently I am looking at the small examples but I would like to be able to fetch larger geometries in future. Is it possible for the simulation to connect to a parallel pvserver which will send images back to the client?

Putting interactive usage to one side for a moment, is non-uniform AMR supported in Catalyst batch? I.e. overlapping structured grids . I couldn’t find any references to non-uniform AMR in ParaView while researching.

Kind regards,
Jim

From: Andy Bauer [mailto:andy.bauer at kitware.com<mailto:andy.bauer at kitware.com>]
Sent: 31 March 2016 16:08
To: Eliot Jim AWE <Jim.Eliot at awe1.awe.co.uk<mailto:Jim.Eliot at awe1.awe.co.uk>>
Cc: paraview at paraview.org<mailto:paraview at paraview.org>
Subject: EXTERNAL: Re: [Paraview] CxxOverlappingAMRExample with Catalyst Live

Hi Jim,
Yes, currently Overlapping AMR data sets are not supported by Catalyst Live. This is because VTK doesn't support sending that data object. That would have to be changed in vtkCommunicator::Send(vtkDataObject* data, int remoteHandle, int tag).
A workaround would be to use a Threshold filter to convert to a multiblock of unstructured grids, use the merge blocks filter and then extract this output to the client. You'll get some warnings about the data set not being a polydata (think about running a large simulation and trying to transfer all of that to a separate server or client). I'm assuming since you're trying to fetch the entire data set that the use case isn't for an overly large data set.
Cheers,
Andy

On Thu, Mar 31, 2016 at 9:44 AM, <Jim.Eliot at awe.co.uk<mailto:Jim.Eliot at awe.co.uk>> wrote:

Good afternoon,

I have been trying to use Catalyst Live with the CxxOverlappingAMRExample from the ParaView Catalyst Example Code repository [1] but I see an error when I try to visualise the data from a ParaView client. The client connects to the simulation successfully and PVTrivialProducer1 appears in the Pipeline Browser. The Information tab shows it as an Overlapping AMR Dataset with 3 levels. However, when I click on the PVTrivialProducer1 icon, ParaView prints an error then segfaults. The error is something like:

    Warning: In <src-dir>/VTK/Parallel/Core/vtkCommunicator.cxx, line 394
    vtkSocketCommunicator (…): Cannot receive vtkOverlappingAMR

I have tried this with both ParaView 4.4.0 and 5.0.0 with the same results. The example works fine when I am using non-Live Catalyst (i.e. it writes out a bunch of vtm files that I can read into ParaView).

So I am wondering, are AMR meshes not supported by Catalyst Live? Do I have to handle AMR meshes differently when using Catalyst interactively?

Many thanks,
Jim

[1] https://github.com/Kitware/ParaViewCatalystExampleCode

Jim Eliot
High Performance Computing Group
AWE, Aldermaston, Reading, RG7 4PR

The information in this email and in any attachment(s) is commercial in confidence. If you are not the named addressee(s) or if you receive this email in error then any distribution, copying or use of this communication or the information in it is strictly prohibited. Please notify us immediately by email at admin.internet(at)awe.co.uk<http://awe.co.uk>, and then delete this message from your computer. While attachments are virus checked, AWE plc does not accept any liability in respect of any virus which is not detected. AWE Plc Registered in England and Wales Registration No 02763902 AWE, Aldermaston, Reading, RG7 4PR

_______________________________________________
Powered by www.kitware.com<http://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

Search the list archives at: http://markmail.org/search/?q=ParaView

Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview

The information in this email and in any attachment(s) is commercial in confidence. If you are not the named addressee(s) or if you receive this email in error then any distribution, copying or use of this communication or the information in it is strictly prohibited. Please notify us immediately by email at admin.internet(at)awe.co.uk<http://awe.co.uk>, and then delete this message from your computer. While attachments are virus checked, AWE plc does not accept any liability in respect of any virus which is not detected. AWE Plc Registered in England and Wales Registration No 02763902 AWE, Aldermaston, Reading, RG7 4PR

The information in this email and in any attachment(s) is 
commercial in confidence. If you are not the named addressee(s) 
or 
if you receive this email in error then any distribution, copying or 
use of this communication or the information in it is strictly 
prohibited. Please notify us immediately by email at 
admin.internet(at)awe.co.uk, and then delete this message from 
your computer. While attachments are virus checked, AWE plc 
does not accept any liability in respect of any virus which is not 
detected.

AWE Plc
Registered in England and Wales
Registration No 02763902
AWE, Aldermaston, Reading, RG7 4PR

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20160405/7469d2d5/attachment.html>