Support calling orca through xvfb on Linux

[jonmmease]

Overview

This PR adds support for calling orca through Xvfb, simplifying the process of using plotly.py's image export functions in a headless Linux environment.

This is accomplished through the addition of a new use_xvfb parameter in plotly.io.orca.config. This parameter may be set to False, True, or 'auto', where:

• False: Never use xvfb
• True: Use Xvfb and raise an error if xvfb-run is not found on the system Path
• 'auto': Use Xvfb if running on Linux, no display serer is detected (determined by checking the DISPLAY environment variable), and xvfb-run is on the path.

use_xvfb will default to False for the remainder of version 3, but it will default to 'auto' in version 4. The new orca_defaults future flag may be used to enable this default behavior now.

Usage

To perform headless static image export on Linux, first install orca following the instructions at https://github.com/plotly/orca#installation. Then install Xvfb using the package manager for your Linux distribution. Then configure plotly.py to call orca through Xvfb using:

import plotly.io as pio
pio.orca.config.use_xvfb = True

Then export images as usual with:

pio.write_image(fig, 'filename.png')

Additional Changes

Add LD_PRELOAD to the list of environment variables to be excluded to be unset when calling orca to avoid breaking orca in Colab (See https://medium.com/@n0mad/how-i-trained-starcraft-2-ais-using-googles-free-gpus-44bc635b0418 for discussion of the Colab's use of tcmalloc).

[jonmmease]

Using plotly.py with orca in Colab

!pip install https://11956-14579099-gh.circle-artifacts.com/0/dist/plotly-3.8.0%2B2.g4a2ccc2e.tar.gz

!wget https://github.com/plotly/orca/releases/download/v1.2.1/orca-1.2.1-x86_64.AppImage -O /usr/local/bin/orca
!chmod +x /usr/local/bin/orca
!apt-get install xvfb libgtk2.0-0 libgconf-2-4

from _plotly_future_ import v4
import plotly.io as pio
import plotly.graph_objs as go

fig = go.Figure(data=[{'y': [3, 2, 4]}])
pio.to_image(fig)

[jonmmease]

orca setup for use in Kaggle kernels

!pip install https://11970-14579099-gh.circle-artifacts.com/0/dist/plotly-3.8.0%2B3.g27437f41.tar.gz
!conda install -y -c plotly plotly-orca
!apt-get -y install libxss1

If the Kaggle Python docker image (https://github.com/Kaggle/docker-python/blob/master/Dockerfile) was updated to include conda install -y -c plotly plotly-orca and apt-get -y install libxss1 then orca would work out-of-the-box with the changes in this branch since Xvfb is already included.

[jonmmease]

Unfortunately, it appears this approach won't work with Azure notebooks because sudo is required to use apt-get to install Xvfb and sudo isn't supported right now https://github.com/Microsoft/AzureNotebooks/issues/303.

[atan4583]

Hi @jonmmease

I installed orca using conda install -c plotly plotly-orca on AWS Linux server, but got this error when calling plotly:

/opt/conda/envs/pysig/lib/orca_app/orca: error while loading shared libraries: libgdk_pixbuf-2.0.so.0: cannot open shared object file: No such file or directory

According to this, I need to install libgtk2.0-0 and libgconf-2-4, but AWS AMI has its own Linux distribution which does not have the 2 packages. Though I also installed Xvfb on AWS AMI, it won't fix the orca error without the 2 missing packages.

So I tried the standalone orca binary route and downloaded orca-1.3.1.AppImage, made it executable and created a symbolic link named orca to point to the orca-1.3.1.AppImage. But when I tried orca --help , I got the following error

fuse: failed to exec fusermount: No such file or directory

Cannot mount AppImage, please check your FUSE setup.
You might still be able to extract the contents of this AppImage
if you run it with the --appimage-extract option.
See https://github.com/AppImage/AppImageKit/wiki/FUSE
for more information
open dir error: No such file or directory

I am not sure how to install FUSE on AWS linux, thus settled on trying AppImage with the --appimage-extract flag.

Followed the --appimage-extract flag example in your post , I installed Xvfb on AWS AMI san libgtk2.0-0 and libgconf-2-4 packages as they are not available on AWS linux AMI.

I then tried /opt/conda/envs/pysig/bin/orca-1.3.1.AppImage --appimage-extract. After the contents got extracted to /home/ec2-user/squashfs-root folder, I ran:

xvfb-run --auto-servernum --server-args "-screen 0 1280x800x24" /home/ec2-user/squashfs-root/orca "$@" > /opt/conda/envs/pysig/bin/orca
chmod 777 /opt/conda/envs/pysig/bin/orca

But orca --help returned this:

/opt/conda/envs/pysig/bin/orca: line 1: /home/ec2-user/squashfs-root/orca:: No such file or directory

I also tried orca-1.2.1-x86_64.AppImage in place of orca-1.3.1.AppImage to no avail. Does the glitch to do with not having libgtk2.0-0 and libgconf-2-4?

orca seems like a big hassle, cant find any help and use case examples beyond windows & ubuntu platform. No help on here nor stackoverflow nor plotly community.