Hello,

I have a linux server without a display attached. I try running blender -b <file>.blend -a which from the documentation, should run the render in the background. Instead I keep getting the following:

root@blender-test:~# blender -b file.blend -a
Blender 2.91.0 (hash 0f45cab862b8 built 2020-11-25 08:51:08)
found bundled python: /bin/2.91/python
Read blend: /root/file.blend
Unable to open a display
Aborted (core dumped)

Is there a way to run this headless using your container image?

Hello,

I want to use this image via DockerFile but I wonder how to get GPU support.

FROM nytimes/blender:3.0-gpu-ubuntu18.04

...

If I run blender cli inside my container I can't access the gpu devices: The command 'bpy.context.preferences.addons["cycles"].preferences.get_devices()' returns None.

Please guide me here.

Thank you!

I use your images with VS Code and the Remote-Containers add-on. This requires some setup which includes a devcontainer.json file to specify the run command. I thought that someone else might find this useful, so here's a short instruction that you could add to the wiki pages.

I've also added the functionality of having the Blender configuration persist between runs.

Cheers :)

Instructions

1. Create a project folder and open it in VS Code. Add a .devcontainer folder.
2. Create a devcontainer.json file inside it with the content as shown below.
3. Also create an empty .config/blender folder.
4. Execute the container by pressing Ctrl+⇧Shift+P and run the command Remote-Containers: Reopen in Container.
5. From the integrated terminal, run blender.

Folder structure

This is the folder structure before running the container.

.
├── .devcontainer
│   └── devcontainer.json
│   ├── .config
│   │   └── blender
├── Docker (Not required)
│   └── Dockerfile

 .
├── blender_file.blend
├── .devcontainer
│   ├── .config
│   │   └── blender
│   │       └── 2.92
│   │           └── config
│   │               ├── platform_support.txt
│   │               ├── recent-files.txt
│   │               └── userpref.blend
│   └── devcontainer.json
├── Docker
│   └── Dockerfile

devcontainer.json

{
	"name": "blender",
	//// Use this if you use a custom image.
	//"build": {
	//	"dockerfile": "../Docker/Dockerfile",
	// },
	// Use this if you only need to run the image directly. 
	"image": "nytimes/blender:latest",
	"settings": {
		"terminal.integrated.shell.linux": "/bin/bash"
	},

	"workspaceMount": "source=${localWorkspaceFolder},target=/root/${localWorkspaceFolderBasename},type=bind", 
	"workspaceFolder": "/root/${localWorkspaceFolderBasename}",
	// We also mount the config folder from the host system. This way, settings 
	// inside the container are kept between runs.
	"mounts": [
		"source=${localWorkspaceFolder}/.devcontainer/.config/blender,target=/root/.config/blender/,type=bind"
	],
	"runArgs": [
		"-it", 
		"--rm", // Removes the container on exit. Remove this if you want the container to keep running.
		"--gpus", "all",
		"-v", "/tmp/.X11-unix:/tmp/.X11-unix:rw", 
		"-v", "/tmp/.docker.xauth:/tmp/.docker.xauth:rw", 
		"-e", "DISPLAY=${localEnv:DISPLAY}", 
		"-e", "XAUTHORITY=/tmp/.docker.xauth", 
		"--device", "/dev/dri/card0:/dev/dri/card0"
	],
	"extensions": []
}

Issues

I've tried doing this is VS Codium (the open source version of VS Code), and I couldn't get it to work.

First of all, thank you for this useful repo! :)

1. Is it possible to run blender with GUI in the container? if so how can that been done?
2. If (1) is possible can you please upload an image version of blender 2.79b?

Thank you in advance!

Hi,

First of all, thanks for the great work you've made. The images you provide are very useful. However I'm currently trying to render frames using Nvidia OptiX but the container does not seem to be able to locate any other OptiX device than the CPU. It works perfectly with CUDA. Also, if I run it directly on the host machine, the OptiX device gets properly located (Nvidia T4).

Do you have any hints ?

Thanks in advance

As you may heard, blender 3.0 is ready. Just wondering do we have plans on it?

Hi,

So I've been working lately on my own Dockerfile trying to get blender setup in a container. I was successful to get it containerized, but I noticed that when I entered my container to render my Monkey.blend file using this command blender --background Monkey.blend --render-output /tmp --render-frame 1 it gave me the following error:

Blender 2.82 (sub 7) (hash 375c7dc4caf4 built 2020-03-12 05:30:40)
AL lib: (EE) ALCplaybackOSS_open: Could not open /dev/dsp: No such file or directory
found bundled python: /usr/bin/blender-2.82a-linux64/2.82/python
Read blend: /usr/bin/blender-2.82a-linux64/Monkey.blend
Unable to open a display
Aborted (core dumped)

So I thought maybe my Dockerfile was faulty, so I gave this repo (nytimes) a try, specifically this Dockerfile.

Build command I used: docker build -t nytimes .
Run command I used: docker run -d -t --name nytimes.container nytimes

Afterwards I saw that it was running, so I then entered the container with docker exec -it nytimes.container /bin/bash

I then proceeded to use the following command to render my Monkey.blend (as a test) that I copied to the /tmp folder in the container.
blender --background /tmp/Monkey.blend --render-output /tmp --render-frame 1

But I receive the same error as my other container.
Blender 2.82 (sub 7) (hash 77d23b0bd76f built 2020-02-12 17:14:50)
AL lib: (EE) ALCplaybackOSS_open: Could not open /dev/dsp: No such file or directory
found bundled python: /bin/2.82/python
Read blend: /tmp/Monkey.blend
Unable to open a display
Aborted (core dumped)

Clearly I'm doing something wrong, or misunderstand how I need to correctly use the container. I double checked the /tmp folder but nothing was present in it, only the Monkey.blend file.

What is the correct way of doing this?

System & version:
host OS Ubuntu 18.04
Docker engine 19.03.8

Hello,

I've noticed that the current docker images are including header files from python 3.6.8 rather than 3.7.x, which is bundled with blender. Is there a reason for not using the 3.7.x headers?

# apt update
W: GPG error: https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64  InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY A4B469963BF863CC
E: The repository 'https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64  InRelease' is no longer signed.
The command '/bin/sh -c apt-get update' returned a non-zero code: 100

Refrence:
https://forums.developer.nvidia.com/t/notice-cuda-linux-repository-key-rotation/212772
NVIDIA/nvidia-container-toolkit#258

docker run --net=host -it --privileged --rm -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v "$HOME/.Xauthority:/root/.Xauthority:rw" --device=/dev/dri/card0:/dev/dri/card0 docker.io/nytimes/blender:latest /bin/blender

Thanks for providing blender docker images. I was trying to render a video with GPU support on a linux server using your pre-build image(nytimes/blender:2.82-gpu-ubuntu18.04)
I use the following command to render my project.

Firstly, I create the docker via:
nvidia-docker run -it -v /home/myProjec/myProject -p 9678:22 nytimes/blender:2.82-gpu-ubuntu18.04 /bin/bash the Docker version of my server is 19.02, for Docker 19.03 or later, one can use docker run --gpus all -it xxx to launch the container.

Then, I enter the container and change the dir to myProject. I have already prepared *.blend, *.py under myProject. Thus, I call blender to get the rendering result via:
blender -b *.blend -P *.py -o output -a

What confused me is that blender is not using CUDA for rendering.

1. For Cycles Render Engine, it took 2 seconds to render each frame locally on my windows PC with RTX 1080Ti. However, it took 9 seconds for each frame in the docker with RTX 2080Ti . When I call nvidia-smi to check the GPU memory usage on Linux server, there is no process running. To force CUDA rendering, I select GPU COMPUTE in *.blend file and add bpy.context.preferences.addons['cycles'].preferences.compute_device_type = 'CUDA' in *.py file, but neither operation works.
2. For EEVEE Render Engine, I receive this error Unable to open a display Aborted (core dumped). If I use vnc or xvfb to generate a virtual display, then OpenGL won't be call while rendering(https://yigityakupoglu.home.blog/)

Is there a way to fix Cycles/EEVEE CUDA rendering?
Thx!

What?

Right now we build and deploy all images (in manifest.json) on every push to main branch which consumes resources and is prawn to surfacing errors for legacy images that we no longer maintain. As suggested in #43 it'd be good to move to a model where we build and deploy

• Only images with LTS versions of Blender (right now 2.83 -> 2.93) as specified here https://www.blender.org/download/lts/
• Only images with LTS versions of Ubuntu (starting with 18)

Hi,

Thanks for this useful repos! Is it possible for you to add a guide about how to use add-ons in this docker image? Thanks!

Hello I have created the docker container from the Docker file that you provides but I can't able to access the software from the browser can you please let me know in what port I should map my docker container to access the application as you didn't mentioned the EXPOSE in the docker file

Thanks for the information but when I run the docker container from the image it is showing that "WARNING: IPv4 forwarding is disabled. Networking will not work."

How can I fix this issue?

I have the requirement to deploy the blend software inside the docker and want to access the software through browser by mapping the ports from the host to container

Hello,

I am trying access Blender's python interpreter within the container. I am running your docker image like this: docker run -ti --rm nytimes/blender:latest, but then when I try to run Blender like this: blender --python-console it fails like this:

root@46216929c17c:/# blender --python-console
Unable to open a display
Aborted (core dumped)
root@46216929c17c:/#

Starting Blender like this appears to work locally on my laptop (I am able to get to the interactive Python shell and import bpy, etc.). Do you happen to know how I can get around this error? I suppose I can just use -b and pass it a Python script directly, but I wanted to be able to access the console interactively for development.

Thanks!

Hello author, I am getting the following error，tanks for your replay
blender -b --enable-autoexec -noaudio --background --python taitui.py --debug

Hi, after making a couple of PRs (#34 #39 #43) I thought that the process for releasing new images could be simplified.

I ended up creating a new repo for CPU images: https://github.com/oliverpool/blender-cpu-image/pkgs/container/blender-cpu-image

Main differences:

• the release process is much easier: I just have to push a tag (corresponding to a blender version)
• the build happens via github actions so anyone can see the log
• the docker image is stored on github infrastructure ghcr.io/oliverpool/blender-cpu-image:3.3.0
• the tag contains the patch version (3.3.0 and not 3.3)
• the base image is (currently) ubuntu 22.04
• it is limited to CPU rendering

I also simplified the Dockerfile:

• python is shipped with blender, so no need for a separate install (I just add a symlink to /bin/python to ease its use)
• numpy is a recent version (so no reinstalling)

Feel free to have a look. I would really like to have some feedback!

I have come here with lots of hopes:)

I have used blender gpu image(3.3) version inside my docker.
My cluster is own aws and have good amount of gpu , cpu and ram.

My first question is :

• Is there any optimized way to render eeve in headless mode inside my docker container.
• My blender automatically killed inside the middle of rendering. Dont know the issue..why this is happening.
• Cycle rendering is to slow. I am exporting video of 30 second which takes around half an hour in 30 samples only. so cycle is not the way i would like to follow. Eeve is so much fast as compared to cycle but not supported in headless mode.

Any help would be much appriciated.

Great containers! When will 2.90 and future releases be added?

On ubuntu 21.04 :

root@Z390-AORUS-PRO:/home/ziomario/Scaricati/blender-cont# docker run --gpus all nytimes/blender:2.82-gpu-ubuntu18.04 nvidia-smi
Unable to find image 'nytimes/blender:2.82-gpu-ubuntu18.04' locally
2.82-gpu-ubuntu18.04: Pulling from nytimes/blender
d519e2592276: Pull complete
d22d2dfcfa9c: Pull complete
b3afe92c540b: Pull complete
13a10df09dc1: Pull complete
4f0bc36a7e1d: Pull complete
cd710321007d: Pull complete
514c7cd6e78a: Pull complete
44874f59d9f6: Pull complete
14971d62f48e: Pull complete
b608b01371f0: Pull complete
c8c606e5e9fd: Pull complete
5d4f90d0e554: Pull complete
c4d0341b3546: Pull complete
809b064dc084: Pull complete
08b51053f4e9: Pull complete
2a4b7a552e0c: Pull complete
Digest: sha256:7b1994a22eeff37116d9f76719e49f345522aafb2d41fbcb536a2eddfd9733e7
Status: Downloaded newer image for nytimes/blender:2.82-gpu-ubuntu18.04
docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].
ERRO[0052] error waiting for container: context canceled

Can we have a Blender 3.4 image?

It introduces Eevee headless render which is highly needed.

There's a PR but no activity: #51

@juniorxsound
First of all thank you for the support you all are providing through this repo. When i tried mounting the device on Azure VM i ran into below error. I'm using Ubuntu 18.04, blender 2.92 and Docker version 20.10.5, build 55c4c88.

Below shows the GPU setup of my remote server.

Initially when i ran the sh file (here) it showed below error:
xauth: file /root/.Xauthority does not exist
xauth: (argv):1: unable to read any entries from file "(stdin)"

Then i created .Xauthority file using below commands

touch ~/.Xauthority
xauth generate $DISPLAY . trusted

But it gives an error ===> xauth: (argv):1: unable to open display "."
Could you please help me on why these errors are generated and also on how to create a DISPLAY for the remote server?

Hi:
Many thanks for your help on blender 3.0 docker images.

Recently we have been working on upgrading the docker image, but somehow we failed to run it in CUDA mode(which do work with blender 2.9* images).

We want to enable GPU with the following function(which works before):

def enable_gpus():
    cycles_addon = bpy.context.preferences.addons['cycles']
    cycles_addon.preferences.compute_device_type = 'CUDA'
    bpy.context.scene.render.engine = "CYCLES"
    bpy.context.scene.cycles.device = "GPU"

But we don't why, bpy.context.preferences.addons['cycles'].preferences.devices returns an empty list. as a result , we are not managed to change compute_devide_type.

As a workaround, we run blender 3.0 in graphics mode, and edit => preference => system, choosing CUDA manually. then we are able to see more than 1 devices in bpy.context.preferences.addons['cycles'].preferences.devices

With a headless blender it is normally possible to use a python script to assign specific devices before rendering, but for some weird reasons, all the version (2.83, 2.90, 2.91) containerized with this repo that I tried seems to ignore the instructions given by the file. For example, if I use the following command

docker run --gpus all --ipc=host -v /mnt/vol_b/blender_input:/volume/blender_input -v /mnt/vol_b/blender_output:/volume/output nytimes/blender:2.90-gpu-ubuntu18.04 blender -b /volume/blender_input/water_sim.blend -E CYCLES -P /volume/blender_input/device_select.py -a

The expected behavior is that a render will be started with the script device_select.py (shown below) activating only the CUDA graphics card.

import re
import bpy
import time
from datetime import datetime
scene = bpy.context.scene
scene.cycles.device = 'GPU'
print("~~~~~> scene device: {}".format(scene.cycles.device))
prefs = bpy.context.preferences
cprefs = prefs.addons['cycles'].preferences
cuda, opencl = cprefs.get_devices()
print("~~~~~> cprefs: {}".format(cprefs))
print("~~~~~> cuda devices: {}".format(cuda))
for compute_device_type in ('CUDA'):
  try:
    cprefs.compute_device_type = compute_device_type
    print('~~~~~> compute device found type set', compute_device_type)
    break
  except TypeError:
    pass
for device in cuda:
    print('~~~~~> Activating', device)
    device.use = (device.type != 'CPU')
    print("~~~~~> device {} usage status: {}".format(device, device.use))

for device in opencl:
    print('~~~~~> Deactivating', device)
    device.use = (device.type != 'CPU')
    print("~~~~~> device {} usage status: {}".format(device, device.use))

Unfortunalety, this doesn't work with the dockerized blender found in this repo. But it does locally on my machine.

Here’s all the things I verified and confirmed are:

• The cuda drivers are properly loaded in the docker container
• I tested with a different project (simple default cube project) and I get the same results
• I loaded and tried different versions of blender using the following containers nytimes/blender:2.83-gpu-ubuntu18.04, and nytimes/blender:2.91-gpu-ubuntu18.04 with the same problem
• Blender runs fine otherwise and renders properly on CPU
• The script print the modifications made before launching the animation render and the parameters are as expected

Any help would be appreciated.

Cheers!

My goal is to generate some images using the Eevee renderer on google cloud run. There is no GPU device, so I believe I need to fake one (as mentioned in #3 )

Here is what my Dockerfile looks like:

FROM nytimes/blender:3.0-cpu-ubuntu18.04
WORKDIR /flask_app
RUN apt-get update && apt-get install xvfb mesa-utils x11-utils python-opengl -y
COPY requirements.txt requirements.txt
RUN pip install -r requirements.txt
RUN pip install gunicorn pyglet pyvirtualdisplay

COPY * ./
RUN chmod a+x boot.sh

ENTRYPOINT ["./boot.sh"]

I have a python script that builds my scene and renders it (works as expected when run locally). To the top of this file I added:

from pyvirtualdisplay import Display
Display().start()

However I am still not able to get the image to render. Is there anything else I need to do to get Eevee working?

Hi, thanks for this excellent resource, please could you add official support for Ubuntu 20.04? Regards, Arie

numpy 1.24.0 got release on 18 Dec 2022 (https://pypi.org/project/numpy/#history).

Since we are not specifying the fixed version of numpy in the build (https://github.com/nytimes/rd-blender-docker/blob/master/generate.py#L107) it installs the latest one automatically.

GLTF import is failing with the latest numpy version and Blender 3.3:

AttributeError: module 'numpy' has no attribute 'bool'. Did you mean: 'bool_'?
    raise AttributeError("module {!r} has no attribute "
  File "/usr/bin/3.3/python/lib/python3.10/site-packages/numpy/__init__.py", line 284, in __getattr__
    poly_smooths = np.empty(num_polys, dtype=np.bool)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_mesh.py", line 607, in set_poly_smoothing
    set_poly_smoothing(gltf, pymesh, mesh, vert_normals, loop_vidxs)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_mesh.py", line 424, in do_primitives

    do_primitives(gltf, mesh_idx, skin_idx, mesh, tmp_ob)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_mesh.py", line 46, in create_mesh
    return create_mesh(gltf, mesh_idx, skin_idx)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_mesh.py", line 25, in create
    mesh = BlenderMesh.create(gltf, pynode.mesh, pynode.skin)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_node.py", line 223, in create_mesh_object
    obj = BlenderNode.create_mesh_object(gltf, vnode)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_node.py", line 51, in create_object
    obj = BlenderNode.create_object(gltf, vnode_id)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_node.py", line 30, in create_vnode
  [Previous line repeated 2 more times]
    BlenderNode.create_vnode(gltf, child)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_node.py", line 44, in create_vnode
    BlenderNode.create_vnode(gltf, child)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_node.py", line 44, in create_vnode
    BlenderNode.create_vnode(gltf, child)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_node.py", line 44, in create_vnode
    BlenderNode.create_vnode(gltf, 'root')
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_scene.py", line 36, in create
    BlenderScene.create(gltf)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_gltf.py", line 40, in _create
    BlenderGlTF._create(gltf)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_gltf.py", line 33, in create
    BlenderGlTF.create(gltf_importer)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/__init__.py", line 1221, in unit_import
    return self.unit_import(self.filepath, import_settings)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/__init__.py", line 1206, in import_gltf2
    return self.import_gltf2(context)
  File "/usr/bin/3.3/scripts/addons/io_scene_gltf2/__init__.py", line 1173, in execute
Traceback (most recent call last):
  poly_smooths = np.empty(num_polys, dtype=np.bool)
/usr/bin/3.3/scripts/addons/io_scene_gltf2/blender/imp/gltf2_blender_mesh.py:607: FutureWarning: In the future `np.bool` will be defined as the corresponding NumPy scalar.  (This may have returned Python scalars in past versions.
  bpy.ops.object.delete({"selected_objects": objs})

Hi everyone, thanks for putting together this image! Just had a quick question, and hope it's alright to post here.

If I want to use this in a staged build, what are the minimal files I need to copy over?

For example...

# --- stage 1 ---
FROM nytimes/blender:3.2-cpu-ubuntu18.04 as blender_build

# --- stage 2 ---
FROM some-other-image/example as final_build
COPY --from=blender_build . /root/blender/  # <-- should I copy everything?
# ... rest of build script

I'm just a bit confused at the outset, does this launch Blender as you would normally run it (with a GUI), but using Docker? Or do you need some sort of X server to get the GUI out (when using https://github.com/nytimes/rd-blender-docker/wiki/Using-the-Blender-GUI-in-containers)

Please could Blender 4.1 support be added? It has some fixes that are required for a project I am working on.

I was trying to build the the image for RPi4b but unfortunately end up with error.

Please add build instruction/ images for ARM/v6

Hello.

In my script that I run inside docker I have troubles with opencv-python module.
The thing is, when I install opencv in system python and use cv2.imread('/image/path') it reads image normally. Also, when I enter in /bin/blender_version/python/bin/python_version and run cv2.imread('image/path'), it also works normally.
But, when I run blender with script, such as /bin/blender --python my_script.py -b -- -a arg1 ... and when my script reaches the part with cv2.imread('/image/path') it reads empty image. To be precise cv2.imread returns None.
And to add a bit of information, it prints this warning just when attempting to cv2.imread :
libpng-1.5.13 but running with 1.6.35 opencv.

Only thing I changed with Dockerfile provided in this repo is additionally installing the opencv-python in blender provided python (and also in python provided in this repo).
Also installed this package because blender fails to even try running my script in background without it:
libgtk2.0-dev

I used version blender2.81 (cpu and gpu), blender2.82 (cpu and gpu), blender2.83(cpu and gpu).

Also, for more additional info I am currently running this on windows host. But I'll try running this on linux in few hours, but I'm writing here before because I really can't even start to think where this might me failing me.

Thanks in advance!

Edit: Forgot to add libgtk2.0-dev dependency in issue description.

Edit 2: Tested it with my arch host machine and still the same result. Also just now I remembered to add docker version. On windows host docker -v output is: Docker version 19.03.8, build afacb8b and on arch output is: Docker version 19.03.11-ce, build 42e35e61f3

Blender 3.3 is out 🎉

Are you guys going to add it as an image? I can try to provide a PR if wanted.

Now that Blender released version 2.83 LTS - add images for 2.83 and update the latest tag

Hi, thanks for clean Blender Docker repo.

have you managed to run 2.92 version inside Docker?
I can't manage to make it work.
My command blender -noaudio -b --engine CYCLES standard1-16-9.360p.blend -f 20 -- --cycles-device CPU
But still getting Unable to open display