Mayavi off screen rendering not working for VTK 7.1.0

[jan-janssen]

I use vtk with the MesaOS driver as I want to run the rendering as part of a jupyter notebook in headless mode, some more details to the package configuration can be found here:
conda-forge/vtk-feedstock#20

When I use the VTK off screen example it is working fine:

#! ./local/bin/python

from vtk import (vtkSphereSource, vtkPolyDataMapper, vtkActor, vtkRenderer,
        vtkRenderWindow, vtkWindowToImageFilter, vtkPNGWriter)

sphereSource = vtkSphereSource()
mapper = vtkPolyDataMapper()
mapper.SetInputConnection(sphereSource.GetOutputPort())

actor = vtkActor()
actor.SetMapper(mapper)

renderer = vtkRenderer()
renderWindow = vtkRenderWindow()
renderWindow.SetOffScreenRendering(1)
renderWindow.AddRenderer(renderer)

renderer.AddActor(actor)
renderer.SetBackground(1, 1, 1)

renderWindow.Render()

windowToImageFilter = vtkWindowToImageFilter()
windowToImageFilter.SetInput(renderWindow)
windowToImageFilter.Update()

writer = vtkPNGWriter()
writer.SetFileName("sphere.png")
writer.SetInputConnection(windowToImageFilter.GetOutputPort())
writer.Write()

But when I test the mayavi off screen example, it is not working:

# Author: Prabhu Ramachandran <prabhu@aero.iitb.ac.in>
# Copyright (c) 2007, Enthought, Inc.
# License: BSD Style.

from os.path import join, abspath, dirname

# The offscreen Engine.
from mayavi.api import OffScreenEngine

# Usual MayaVi imports
from mayavi.scripts.util import get_data_dir
from mayavi.sources.api import VTKXMLFileReader
from mayavi.modules.api import Outline, ScalarCutPlane, Streamline


def main():
    # Create the MayaVi offscreen engine and start it.
    e = OffScreenEngine()
    # Starting the engine registers the engine with the registry and
    # notifies others that the engine is ready.
    e.start()

    # Create a new scene.
    win = e.new_scene()

    # Now setup a normal MayaVi pipeline.
    src = VTKXMLFileReader()
    src.initialize(join(get_data_dir(dirname(abspath(__file__))),
                        'fire_ug.vtu'))
    e.add_source(src)
    e.add_module(Outline())
    e.add_module(ScalarCutPlane())
    e.add_module(Streamline())
    win.scene.isometric_view()
    # Change the size argument to anything you want.
    win.scene.save('offscreen.png', size=(800, 800))


if __name__ == '__main__':
    main()

The error message I get is:

ERROR: In ../Rendering/OpenGL2/vtkXOpenGLRenderWindow.cxx, line 301
vtkXOpenGLRenderWindow (0x30ced20): bad X server connection. DISPLAY=Aborted

So to me it seems like the off screen initialisation is not correctly forwarded to VTK. Unfortunately I was unable to locate the initialisation of VTK inside mayavi myself. To me it seems to be happening in tvtk_access.py, still I do not see an option to force offscreen rendering directly.

[Cadair]

I found that by setting ETS_TOOLKIT='null' I could reproduce this error using:

>>> from mayavi import mlab
>>> mlab.options.offscreen = True
>>> mlab.test_plot3d()
ERROR: In ../Rendering/OpenGL2/vtkXOpenGLRenderWindow.cxx, line 301
vtkXOpenGLRenderWindow (0x2e027d0): bad X server connection. DISPLAY=Aborted (core dumped)

[jan-janssen]

If ETS_TOOLKIT='null' is not set, not even the import works. So for the moment the script above fails when I try to create a scene new_scene()

[prabhuramachandran]

@jan-janssen -- Can you please test with latest master, it should be fixed now. You will still need the export ETS_TOOLKIT=null but the rest should work now.

[Cadair]

I have tested this using the osmesa conda-forge vtk packages and I can't make it work without an X server. I think I have the env setup properly.

What I did was:

$ export ETS_TOOLKIT='null'
$ conda create -n mayavi_osmesa -c conda-forge mesalib pyqt numpy vtk
$ pip install git+https://github.com/enthought/mayavi
$ python -c "from mayavi import mlab; mlab.options.offscreen=True; mlab.test_plot3d(); mlab.savefig('test.png')"
ERROR: In ../Rendering/OpenGL2/vtkXOpenGLRenderWindow.cxx, line 445
vtkXOpenGLRenderWindow (0x1c58c80): bad X server connection. DISPLAY=Aborted (core dumped)

[Cadair]

ok actually the pure VTK example given in the OP dosen't work either, so I assume there is something wrong with my env. I get the same thing with a straight pip install git+https://github.com/enthought/mayavi so I am assuming I am missing something in vtk somewhere.

[prabhuramachandran]

Thanks for testing, can you check if the installed VTK has vtkOSOpenGLRenderWindow, i.e.
import vtk; print(hasattr(vtk, 'vtkOSOpenGLRenderWindow')). If it doesn't, VTK was probably not built with osmesa support. If it does, I will need to investigate carefully.

[Cadair]

It does not have that. I don't suppose you know what controls if that exists or not?!

[prabhuramachandran]

When VTK is built you need to set VTK_OPENGL_HAS_OSMESA to ON.

[Cadair]

hmmm, as far as I can tell the conda-forge osmesa package should do that. I will go open an issue there.

[prabhuramachandran]

I think newer mesa releases do not allow both mesa and osmesa so recent VTK releases may automatically turn off osmesa when vtk_use_x is turned on. But I am not entirely sure about this as many things could have changed ...

[prabhuramachandran]

If possible could you try with an older version like VTK-7.x or whenever this was known to work?

[dappelha]

I'm having the same error message on Ubuntu 18, vtk 8.1.2, mayavi 4.6.2

[tjiagoM]

I'm also having this issue with mayavi. Did anyone find any solution? I'm having this issue when using mayavi in a server, although I have the display being forwarded to my local computer.

[max9111]

I had quite a similar problem when using Mayavi in a Jupyter-Notebook provided by a remote Jupyter-Hub server. The problem was that that there is no X Server is running.

I solved it with installing xvfbwrapper and the Xvfb back-end. Before doing anything with mayavi start a Display.

Example

#The following 3 lines must be before doing anything with mayavi (even importing)
from xvfbwrapper import Xvfb
vdisplay = Xvfb(width=1920, height=1080)
vdisplay.start()

from mayavi import mlab
mlab.init_notebook()
s = mlab.test_plot3d()
s

If there is no more elegant way, I think it would be good to mention this in the manual.

[tjiagoM]

@max9111 do you remember doing anything else? I tried your code after installing xvfbwrapper. But after ssh'ing to my server using the -X flag, I still get the error:

ERROR: In ../Rendering/OpenGL2/vtkXOpenGLRenderWindow.cxx, line 606
vtkXOpenGLRenderWindow (0x55b43e61d590): Cannot create GLX context. Aborting.

[max9111]

@tjiagoM Maybe you have forgotten to install xfb itself (sudo apt-get install xvfb)?

[tjiagoM]

@max9111 oh, no, initially I made that mistake but they would throw an error saying that no xvfb was found/active (or something similar)

[mirestrepo]

@max9111 I'm struggling with the same set-up, mayavi on a remote jupyterhub. Do you have a sample docker image used by the hub? I tried using xvfbwrapper as you suggested but it didn't do the trick for me. Thanks in advance!

[max9111]

@mirestrepo
Unfortunately don't have a docker image.

My setup is:

• A virtual Machine running on Kubuntu 18.04 (using Virtual Box) on a remote Workstation which runs on CentOS7.

• Installed The Littlest JupyterHub

• Make Jupyterhub available from outside (Port-forwarding, Firewallsetup)

• Installed XfVB in the virtual machine like that
  The crucial Point here is to get a working X-Server (so general tutorials how to do something GUI related in a Docker image might also help) I would suggest trying the simple script from the link to check this. In this case you would also get a meaningful error message.

• If this works it is important that you start the X-Server before doing any mayavi related stuff, like shown in the example above. If you import mayavi from mayavi import mlab before initializing the X-Server mayavi will crash.

[mirestrepo]

@max9111 I managed to skip the xvfb wrapper by launching xvfb at entry point for my docker ENTRYPOINT ["tini", "-g", "--", "xvfb-run"] but I can only get mlab to work with the png backend... I haven't tried building vtk with OSMesa though...

[yuehaowang]

The xvfb-run may come to your assistance:

 xvfb-run python your-py-script.py

Note you need to install xvfb before running the command above.