VtkLocalView and conda Paraview

[edmondb]

Hello,

I was using the conda-forge route to try and see if I could run the paraview.simple through trame alone since it said

When using the conda approach you won’t need to download ParaView as it will get installed by conda for you.

I'm complicating the situation by running on an AWS EC2 machine trying to do local rendering. Each time I try to call simple.Render, I get a Seg fault. Any help would be greatly appreciated!

python trame-tutorial/05_paraview/SimpleCone.py --host 0.0.0.0 --port 9090

Loguru caught a signal: SIGSEGV
Stack trace:
36            0x5878ee python() [0x5878ee]
35      0x7f8074829e40 __libc_start_main + 128
34      0x7f8074829d90 /lib/x86_64-linux-gnu/libc.so.6(+0x29d90) [0x7f8074829d90]
33            0x587a39 Py_BytesMain + 57
32            0x5b447e Py_RunMain + 894
31            0x5b6f02 PyRun_SimpleFileExFlags + 418
30            0x456537 python() [0x456537]
29            0x5bd220 python() [0x5bd220]
28            0x5c1217 python() [0x5c1217]
27            0x59398b PyEval_EvalCode + 27
26            0x4e66c9 PyEval_EvalCodeEx + 57
25            0x4e6717 _PyEval_EvalCodeWithName + 71
24            0x4e6a8a python() [0x4e6a8a]
23            0x4ec6d4 _PyEval_EvalFrameDefault + 19780
22            0x4f8053 python() [0x4f8053]
21            0x4ec6d4 _PyEval_EvalFrameDefault + 19780
20            0x504f6c python() [0x504f6c]
19            0x4e6a8a python() [0x4e6a8a]
18            0x4ed53f _PyEval_EvalFrameDefault + 23471
17            0x5057c8 PyObject_Call + 344
16            0x507300 python() [0x507300]
15      0x7f80475d301d /home/ubuntu/miniconda3/envs/pv/lib/python3.9/site-packages/paraview/modules/vtkRemotingViews.so(+0x1d301d) [0x7f80475d301d]
14      0x7f8050ea30ab vtkSMViewProxy::StillRender() + 379
13      0x7f80514ce055 vtkPVSessionBase::ExecuteStream(unsigned int, vtkClientServerStream const&, bool) + 53
12      0x7f80514d0687 vtkPVSessionCore::ExecuteStreamInternal(vtkClientServerStream const&, bool) + 263
11      0x7f807410c3be vtkClientServerInterpreter::ProcessStream(vtkClientServerStream const&) + 30
10      0x7f807410bed4 vtkClientServerInterpreter::ProcessOneMessage(vtkClientServerStream const&, int) + 196
9       0x7f807410bdb7 vtkClientServerInterpreter::ProcessCommandInvoke(vtkClientServerStream const&, int) + 1319
8       0x7f804d00a9c1 vtkPVRenderViewCommand(vtkClientServerInterpreter*, vtkObjectBase*, char const*, vtkClientServerStream const&, vtkClientServerStream&, void*) + 6913
7       0x7f8050de5ad6 vtkPVRenderView::StillRender() + 118
6       0x7f8050df2bb9 vtkPVRenderView::Render(bool, bool) + 2345
5       0x7f804fe07454 vtkXOpenGLRenderWindow::Render() + 52
4       0x7f804fd5b732 vtkOpenGLRenderWindow::Render() + 194
3       0x7f8051f67740 vtkRenderWindow::Render() + 624
2       0x7f80512de7ed vtkXRenderWindowInteractor::Initialize() + 237
1       0x7f804ff10fb9 XSync + 25
0       0x7f8074842520 /lib/x86_64-linux-gnu/libc.so.6(+0x42520) [0x7f8074842520]
(   1.279s) [paraview        ]                       :0     FATL| Signal: SIGSEGV
Segmentation fault (core dumped)```

[jourdain]

The problem is similar to that one.

ParaView from conda is built to leverage X. On AWS, use the OSMesa one if you don't have a GPU.

[edmondb]

Is there an OSMesa version of ParaView purely through conda or a way to specify the OSMesa version instead of the default conda version leveraging X?

Thank you for the quick reply and the link to the other thread! I might try doing the OpenGL version using an EC2 that has GPUs as that is preferred from conda.

[jourdain]

Even so, it won't work unless you use EGL. I don't think conda has any option for those headless path.

[edmondb]

Okay, thank you!