[Paraview] Stream tracer with custom source fails in parallel mode

Paluszek, Lukasz lukasz.paluszek at airbus.com
Wed Mar 23 06:39:09 EDT 2016


Hello Andy,

I confirm that your solution works for all of the cases. As long as I am able to generate the seeding , save and then re-read a VTM, PVD or PVTU file then the Stream Tracer with Custom Source works in parallel.

What is worth noting is that in parallel the mask points filter with the “Proportionally Distribute Maximum Number Of Points” option requires that all processes contain a bit of the input data, i.e. it the direct input for this filter must be D3-partitioned data, otherwise mask points filter will hang or crash.

Regards,
Lukasz

[Beschreibung: Beschreibung: Description : cid:image001.png at 01CF0B8F.DB7831F0]
Łukasz Paluszek
Airbus Helicopters
Aerodynamics Department (ETGA)
86607 Donauwörth
Germany
Tel.: +49 (90) 671 8231
Fax.: +49 (90) 671 9122
Lukasz.Paluszek at airbus.com<mailto:Lukasz.Paluszek at airbus.com>

From: Andy Bauer [mailto:andy.bauer at kitware.com]
Sent: 18 March 2016 15:35
To: Paluszek, Lukasz
Cc: paraview at paraview.org
Subject: Re: [Paraview] Stream tracer with custom source fails in parallel mode

Hi Lukasz,
Here's a not too hacky workaround for doing your parallel use case 1. I think the main issue is that the seed source needs to be fully available on all processes. By using disk_out_ref.ex2 in parallel or the D3 filter and then the slice it is making this not true. What you can do is the following (all in parallel if you want):

  1.  Read in disk_out_ref.ex2
  2.  Create slice
  3.  Save slice output in vtm format
  4.  Read slice output back in
  5.  Create D3 from disk_out_ref
  6.  Do STCS with D3 and reader as the seeds

I think for readers that work can automatically partition the data in parallel you may be able to skip steps 2 and 3 though and use the slice output as the seeds. If you try this and it doesn't work please let us know. Also, I didn't verify my fix for your parallel 2, 3 and 4 use cases but if you try similar stuff for that and it doesn't work, please us know that as well.

Best,

Andy

On Fri, Mar 18, 2016 at 10:03 AM, Andy Bauer <andy.bauer at kitware.com<mailto:andy.bauer at kitware.com>> wrote:
Hi Lukasz,
I was able to replicate your crash in your first parallel crash with PV 5.0. I didn't try any of the other cases yet but am guessing that they're all related.
Thanks for the bug report.
Best,
Andy

On Fri, Mar 18, 2016 at 4:56 AM, Paluszek, Lukasz <lukasz.paluszek at airbus.com<mailto:lukasz.paluszek at airbus.com>> wrote:
Dear All,

I have been struggling to make the Stream Tracer with Custom Source (STCS) work in parallel. It works in serial mode without any issues but in parallel I am hitting a wall of obstacles. I made a little test with the disk_out_ref.ex2 in serial and parallel by and this is what I found. I am using Paraview v5.0.0 compiled from source with system libs linked against openmpi 1.7.1. I also briefly checked with other mpi flavours and under Windows with the MPI enabled binaries from paraview.org<http://paraview.org>, but the problems did not disappear.

Serial mode:

1)
Load file (just with the “V” variable)
Slice with Z=5
Stream Tracer with Custom Source (STCS)
Success

2)
Load file
Slice with Z=5
Mask points (10 points with random sampling)
STCS
Success

3)
Load file
Select some points with Select Points Through
Extract Selection
STCS
Success

Everything worked without issues. Now the same in parallel on 6 processes. Here is a little word of warning, the Exodus reader reads all the data onto just one process, so unless the D3 filter is used everything works in more or less serial mode.

1)
Load file
D3
Slice with Z=5
STCS
Crash
“
Generic Warning: In /home/caldron-b/paluszek/linux/Software/Paraview/v5.0.0/ParaView-v5.0.0-source/VTK/Parallel/MPI/vtkMPICommunicator.cxx, line 72
MPI had an error
------------------------------------------------
MPI_ERR_IN_STATUS: error code in status
------------------------------------------------

--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 5 in communicator MPI_COMM_WORLD
with errorcode 18.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
“
2)
Load file
D3
Slice with Z=5
Mask points (10 points, random sampling & proportional distribution)
Paraview hangs:
“
ERROR: In /home/caldron-b/paluszek/linux/Software/Paraview/v5.0.0/ParaView-v5.0.0-source/VTK/Filters/Core/vtkMaskPoints.cxx, line 325
vtkPMaskPoints (0x31c8480): No points to mask

ERROR: In /home/caldron-b/paluszek/linux/Software/Paraview/v5.0.0/ParaView-v5.0.0-source/VTK/Filters/Core/vtkMaskPoints.cxx, line 325
vtkPMaskPoints (0x4b0dea0): No points to mask
“
If you do not switch proportional distribution then Paraview spits an error “No points to mask” but you doesn’t crash or freeze. Generating STCS from the masked points will crash Paraview though.

3)
Load file
D3
Slice with Z=5
D3 the slice
Mask points (10 points)
D3 the points
STCS
Crash
“
Generic Warning: In /home/caldron-b/paluszek/linux/Software/Paraview/v5.0.0/ParaView-v5.0.0-source/VTK/Parallel/MPI/vtkMPICommunicator.cxx, line 72
MPI had an error
------------------------------------------------
MPI_ERR_TRUNCATE: message truncated
------------------------------------------------

--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 1 in communicator MPI_COMM_WORLD
with errorcode 15.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
”

4)
Load file
D3
Select some points with Select Points Through
Extract Selection
STCS
Crash


I also tried with more combinations with and without the D3 filter on the subsequent filters and on my industrial geometries but I could never make the Stream Tracer with Custom Source work in parallel. I found a related bug report http://www.paraview.org/Bug/view.php?id=15342 and an old post https://cmake.org/pipermail/paraview/2009-December/014909.html but no solutions to this problem. Interestingly the common Stream Tracer filter works in parallel with both point and high resolution line sources.

I would be grateful if you could try to first reproduce this behavior and confirm that this is not specific to my installation. If anyone had an idea how to fix that I would be more than happy.

Regards,
Lukasz



[Beschreibung: Beschreibung: Description : cid:image001.png at 01CF0B8F.DB7831F0]
Łukasz Paluszek
Airbus Helicopters
Aerodynamics Department (ETGA)
86607 Donauwörth
Germany
Tel.: +49 (90) 671 8231<tel:%2B49%20%2890%29%20671%208231>
Fax.: +49 (90) 671 9122
Lukasz.Paluszek at airbus.com<mailto:Lukasz.Paluszek at airbus.com>


AIRBUS HELICOPTERS DEUTSCHLAND GmbH

Sitz der Gesellschaft / Registered Office: Donauwörth

Registergericht / Registration Court: Amtsgericht Augsburg HRB 16508

Vorsitzender des Aufsichtsrates / Chairman of the Supervisory Board: Guillaume Faury

Geschäftsführung / Board of Management: Dr. Wolfgang Schoder, Vorsitzender / CEO; Johannes Kleidorfer; Dr. Klaus Przemeck; Martin Schübel; Stefan Thomé



The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised.

If you are not the intended recipient, please notify Airbus immediately and delete this e-mail.

Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately.

All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free.







_______________________________________________
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



AIRBUS HELICOPTERS DEUTSCHLAND GmbH

Sitz der Gesellschaft / Registered Office: Donauwörth

Registergericht / Registration Court: Amtsgericht Augsburg HRB 16508

Vorsitzender des Aufsichtsrates / Chairman of the Supervisory Board: Guillaume Faury

Geschäftsführung / Board of Management: Dr. Wolfgang Schoder, Vorsitzender / CEO; Johannes Kleidorfer; Dr. Klaus Przemeck; Martin Schübel; Stefan Thomé

<html><head></head><body><font color="black" face="arial" size="2">
The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised.
If you are not the intended recipient, please notify Airbus immediately and delete this e-mail.
Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately.
All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free.
</font>
</body>
</html>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20160323/93741353/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 16151 bytes
Desc: image001.png
URL: <http://public.kitware.com/pipermail/paraview/attachments/20160323/93741353/attachment-0001.png>


More information about the ParaView mailing list