Docker-All-We-Need

1. System Requirements

Ubuntu: 18.04

2. Install the docker

Take a look the reference Docker Installation. I have built a script that automatically installs the docker engine.

bash install_docker.sh

3. Docker with Travis CI

Download a sample repo which uses docker with travis CI

git clone https://github.com/leviethung2103/WEB1066-probot-hello
cd WEB1066-probot-hello
# Go to master_docker branch 
checkout master_docker

We need to have the Dockerfile inside the folder.

Test the project locally to see it works or not, on your computer:

Install the docker (already done) - Step 1
Build the project with Dockerfile - Step 3.1
Test the project inside Docker container - Step 3.2

3.1 Build locally

docker build -t probot-hello .

3.2 Test the project

docker run -it --rm probot-hello npm run test

--rm: remove the container on exit.

4. Develop locally

Changes to source code get reflected in the container and node restarts our app with the lastest changes

docker run -it --rm |
			-e APP_ID=abc \
			-e PRIVATE_KEY=none \
			-w /home/node/probot-hello-dev \
			-v "$(pwd)":/home/node/probot-hello-dev \
			-p 3000:3000 probot-hello \
			bash -c 'npm install && npm run dev'

npm run dev: watcher to changes, see the real time updates in the application.

Dockerfile Sample

# https://hub.docker.com/_/node/
FROM node:8-onbuild

# use debug to troubleshoot
ENV LOG_LEVEL=error
# Required env settings determined by GitHub App
ENV APP_ID=1234
ENV WEBHOOK_SECRET=development
ENV WEBHOOK_PROXY_URL=https://localhost:3000/
ENV PRIVATE_KEY="someprivatestring"

# see https://github.com/nodejs/docker-node/blob/e3ec2111af089e31321e76641697e154b3b6a6c3/docs/BestPractices.md#global-npm-dependencies
ENV NPM_CONFIG_PREFIX=/home/node/.npm-global
ENV PATH=$PATH:/home/node/.npm-global/bin

# Lets install our app into /home/node
COPY . /home/node/probot-hello
RUN chown -R node:node /home/node/probot-hello

# setup our app
# non-root user  https://github.com/nodejs/docker-node/blob/e3ec2111af089e31321e76641697e154b3b6a6c3/docs/BestPractices.md#non-root-user
USER node

WORKDIR /home/node/probot-hello
RUN npm install

5. NVIDIA-Docker

5.1 Installation

We need to install the NVIDIA driver, no need to install the CUDA toolkit on the host. The CUDA toolkit will be installed inside the Container.

Take note of your Docker version with docker -v. Versions earlier than 19.03 require nvidia-docker2 and the --runtime=nvidia flag. On versions including and after 19.03, you will use the nvidia-container-toolkit package and the --gpus all flag. Both options are documented on the page linked above.

bash install_nvidia_docker.sh

5.2 Usage

With the release of Docker 19.03, usuage of nvidia-docker2 packages are deprecated.

The NVIDIA GPUs are now natively supported as devices in the Docker runtime.Please note that this native GPU support has not landed in docker-compose yet.

Now one can simply pass --gpus option for GPU-accelerated Docker based application.

Check the docker version

$ docker -v 
Docker version 19.03.11, build 42e35e61f3

Example

docker run -it --rm --gpus all ubuntu nvidia-smi

Test nvidia-smi with the lasted official CUDA image

docker run --gpus all nvidia/cuda:10.0-base nvidia-smi

new image: nvidia/cuda:10.0-base

Starting a GPU enabled container on specific GPUs 1 and 2

docker run --gpus '"device=1,2"' nvidia/cuda:10.0-base nvidia-smi

How about if we don't specify the tag --gpus ?

$ docker run nvidia/cuda:10.0-base nvidia-smi
docker: Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused "exec: \"nvidia-smi\": executable file not found in $PATH": unknown.
ERRO[0001] error waiting for container: context canceled

6. Docker Compose

Compose is a tool for defining and running multi-container Docker applications.

With Compose, you can use a YAML file to configure your application's services. Then, with a single command, you create and start all the services from your configuration.

Using Compose is basically a three-step process:

Define your app's environment with a Dockerfile so it can be reproduced anywhere.
Define the services that make up your app in docker-compose.yml so they can be run together in an isolated environment.
Run docker-compose up and Compose starts and runs your entire app

A docker-compose.yml looks like this:

version: '2.0'
services:
  web:
    build: .
    ports:
    - "5000:5000"
    volumes:
    - .:/code
    - logvolume01:/var/log
    links:
    - redis
  redis:
    image: redis
volumes:
  logvolume01: {}

Compose has commands for managing the whole lifecycle of your application:

Start, stop, and rebuild services
View the status of running services
Stream the log output of running services
Run a one-off command on a service

6.1 Docker Compose Features

The features of Compose that make it effective are:

For more information, take a look DockerCompose.

6.2 Install Docker Compose

Check the version of Docker Compose

$ docker-compose --version

bash scripts/install_docker_compose.sh

6.3 Download a Tensorflow Docker Image

The official TensorFlow Docker images are located in the tensorflow/tensorflow Docker Hub repository. Image releases are tagged using the following format:

Tag	Description
`latest`	The latest release of TensorFlow CPU binary image. Default.
`nightly`	Nightly builds of the TensorFlow image. (unstable)
`version`	Specify the version of the TensorFlow binary image, for example: 2.1.0
`devel`	Nightly builds of a TensorFlow `master` development environment. Includes TensorFlow source code.

Each base tag has variants that add or change functionality:

Tag Variants	Description
`tag``-gpu`	The specified tag release with GPU support. (See below)
`tag``-py3`	The specified tag release with Python 3 support.
`tag``-jupyter`	The specified tag release with Jupyter (includes TensorFlow tutorial notebooks)

There are two ways to get the docker images

Use the docker-compose.yml

sudo docker-compose -f tensorflow-docker.yml up --build

Use directly the terminal

docker pull tensorflow/tensorflow                     # latest stable release
docker pull tensorflow/tensorflow:devel-gpu           # nightly dev release w/ GPU support
docker pull tensorflow/tensorflow:latest-gpu-jupyter  # latest release w/ GPU support and Jupyter

Inside of tensorflow-docker.yml

version: '3'
services:
  tf: 
      image: tensorflow/tensorflow:latest-gpu
      ports:
        - 8888:8888
      volumes:
        - .:/notebooks

List all the images

docker image ls

Start a Tensorflow Docker container

docker run [-it] [--rm] [-p hostPort:containerPort] tensorflow/tensorflow[:tag] [command]

Example

docker run -it -p 8888:8888 tensorflow/tensorflow:latest-gpu-jupyter

However, this one only uses the Tensorflow-CPU

from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 189652975046659030
, name: "/device:XLA_CPU:0"
device_type: "XLA_CPU"
memory_limit: 17179869184
locality {
}
incarnation: 6518940841554504313
physical_device_desc: "device: XLA_CPU device"
]

How to use the tensorflow wit enabled-GPU?

just simply pass the tag --gpus all

docker run -it -p 8888:8888 --gpus all  tensorflow/tensorflow:latest-gpu-jupyter

It will shows the GPU like this

[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 14274143214298123097
, name: "/device:XLA_CPU:0"
device_type: "XLA_CPU"
memory_limit: 17179869184
locality {
}
incarnation: 8827210151315510378
physical_device_desc: "device: XLA_CPU device"
, name: "/device:XLA_GPU:0"
device_type: "XLA_GPU"
memory_limit: 17179869184
locality {
}
incarnation: 4945486739601633121
physical_device_desc: "device: XLA_GPU device"
, name: "/device:GPU:0"
device_type: "GPU"
memory_limit: 4890624000
locality {
  bus_id: 1
  links {
  }
}
incarnation: 5081808267847125318
physical_device_desc: "device: 0, name: GeForce GTX 1660, pci bus id: 0000:01:00.0, compute capability: 7.5"
]

[Advanced] Setup host machine for build and running our docker container

Install the container based on Dockerfile or docker-compose.yml

Setup to show GUI tracking 360 cam (X11 forwarding mode)
- Modify .env file
  - SHOW_GUI=True
- Setup X11

$ XSOCK=/tmp/.X11-unix
$ XAUTH=/tmp/.docker.xauth
$ touch ${XAUTH}
$ xauth nlist ${DISPLAY} | sed 's/^..../ffff/' | xauth -f ${XAUTH} nmerge -

Build and start the docker container

# Adds docker to X server access control list (ACL)
$ xhost + local:docker

# Build and run the services
$ docker-compose up --build

# Removes docker out of X server ACL if you are not working with
$ xhost - local:docker

Docker Compose with GPUs

Basically, after installing the nvidia-docker, we can run the container via docker run. However, it's a little tricky when using the docker compose up.

In order to use the gpus as well as cuda library, we need to specify the runtime argument inside the docker-compose.yml.

By doing that, we can easily run the gpus via docker-compose command.

Install the nvidia-docker2 is a must.

sudo apt-get install -y nvidia-docker2

Configure the /etc/docker/daemon.json file.

"runtimes": {"nvidia": { "path": "/usr/bin/nvidia-container-runtime","runtimeArgs": [] } }

in your docker-compose YML you should add: runtime: nvidia

Verified verison: Cuda 10.1 - Devel - Ubuntu 18 without cudnn

FROM nvidia/cuda:10.1-devel-ubuntu18.04 as nvidia_base

Reference

Docker Installation

Continous Integration

TensorflowDocker

leviethung2103/Docker-All-We-Need