DockerTutorial

Simplest Docker Tutorial

Hello World

We will use https://github.com/shekhargulati/python-flask-docker-hello-world

git clone https://github.com/shekhargulati/python-flask-docker-hello-world.git
cd python-flask-docker-hello-world
docker build -t simple-app .
docker run -it -p 5000:5000 simple-app

Installing Docker on Ubuntu

curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
sudo add-apt-repository \
   "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
   $(lsb_release -cs) \
   stable"
sudo apt-get update
sudo apt-get install -y docker-ce
sudo usermod -aG docker $USER

Installing nvidia-docker

distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | \
sudo tee /etc/apt/sources.list.d/nvidia-docker.list

sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

Essential Commands

Preparing Image

Check https://hub.docker.com/search?q=&type=image

docker pull <image_name>  # Download image from docker hub
docker images  # Show all images

Running

Simple Run

Running shell directly. Check <image_name> from docker images

docker run -it <image_name> /bin/bash

Example

docker run -it ubuntu

Running shell background daemon

docker run -it -d <image_name>

Running with GPUs

docker run --gpus all -it <image_name> /bin/bash
docker run --gpus 2 -it <image_name> /bin/bash  # Run two gpus available.
docker run --gpus '"device=0,1,2,3"' -it <image_name> /bin/bash  # Run 4 GPU id=0,1,2,3

Exit shell without stopping container.

<CTRL+P> + <CTRL+Q>

Share storage with system

docker run -v <system_directory>:<container_directory> ...

Example

docker run -it -v /home/yoongi/Projects:/data pytorch/pytorch:latest /bin/bash

Expose port

docker -p <system_port>:<container_port> ...

Example

docker run -it -p 5555:5555 -p 5556:5556 --gpus 1 yoongicomcomai/bert-as-service-base:latest /bin/bash

Using GUI programs such as cv2.imshow()

xhost +local:docker
docker run -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -e QT_X11_NO_MITSHM=1

Check running containers

docker ps  # Check running containers
docker ps -a  # Show all containers including stopped containers

Save running container as image

docker commit <container_id> <username/imagename:tag>  # Save running docker as image
docker stop <container_id>  # Check 'docker ps'
docker rm <container_id>  # Delete a stopped container. Check 'docker ps -a'

Upload to DockerHub

Sign up https://hub.docker.com/

# 1. Rename local image same as dockerhub name
docker tag <image_name:tag> <dockerhub_username/imagename:tag>

# 2. Login to docker hub
docker login

# 3. Push layers
docker push <dockerhub_username/imagename:tag>

Delete docker image

# 1. Check what image to delete
docker images

# 2. Delete an image. ### Becarefull Not to delete other's images ###
docker rmi <image_id>

Build Dockerfile

docker build -t <image_name> <path to directory containing Dockerfile>

How to make Dockerfile

Example of Dockerfile

https://github.com/Laeyoung/Zooming-Slow-Mo-CVPR-2020/blob/master/Dockerfile

# 1. base image
FROM pytorch/pytorch:1.1.0-cuda10.0-cudnn7.5-devel

# 2. apt install
# Each time RUN docker creates intermediate container. Use '\' to coninuously run commands.
RUN apt update && \
  apt install -y git wget ffmpeg libsm6 libxext6 libxrender-dev libglib2.0-0

# 3. pip install
COPY ./pip.conf ~/.pip/pip.conf
RUN pip install numpy opencv-python lmdb pyyaml pickle5 matplotlib seaborn

# 4. install flask and expose 80 port
RUN pip install flask Flask-Limiter
EXPOSE 80

# 5. download pre-trained model
RUN mkdir /app
WORKDIR /app
RUN wget --no-check-certificate 'https://docs.google.com/uc?export=download&id=1xeOoZclGeSI1urY6mVCcApfCqOPgxMBK' -O model.pth

# 6. copy codes
COPY . .

# 7. set ENTRYPOINT and CMD
ENTRYPOINT bash /app/entrypoint.sh
CMD []

docker build -t <yoongi/zooming-slow-mo:latest> ./Dockerfile

Dockerfile commands (FROM, RUN, WORKDIR, CMD, ...)

https://rampart81.github.io/post/dockerfile_instructions/

Example Situation (Flask app)

1. Pull python3 docker image

docker pull python:3.8.3-slim

2. Run docker image

docker run -it python/python:3.8.3-slim /bin/bash

3. Install git & clone repository

apt update
apt install git
git clone https://github.com/YoongiKim/DockerTutorial.git
cd DockerTutorial/
ls -al

4. Install Flask by pip

pip install flask

5. Run Flask app

python app.py

6. Test web page

Connect web-browser to 127.0.0.1:5000 (It won't connect)

7. Expose port 5000

We didn't opened port 5000. So we should save, stop, re-run with -p 5000:5000 option

7-1. Exit without stopping container.

If we just exit we will lose changes. So press key CTRL+P + CTRL+Q

<CTRL+P> + <CTRL+Q>

7-2. Show running containers

docker ps

Then it will show you running containers

CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS              PORTS               NAMES
e0bbebd22254        python:3.8.3-slim   "/bin/bash"              7 minutes ago       Up 7 minutes                            blissful_borg

7-3. Save changes

# docker commit <CONTAINER ID> <NEW IMAGE NAME>
docker commit e0bbebd22254 simple-app:latest

7-4. Stop running container

docker stop e0bbebd22254

7-5. Re-Run with 5000 port opened

# -p outside_port:inside_port
docker run -it -p 5000:5000 simple-app:latest /bin/bash

7-6. Test web site again

Connect web-browser to 127.0.0.1:5000

8. Push

Now we can upload our own image to docker hub.

docker images
docker tag simple-app:latest yoongicomcomai/simple-app:latest
docker login
docker push yoongicomcomai/simple-app:latest

9. Pull

Later we can download our image

docker pull yoongicomcomai/simple-app:latest

10. Dockerfile making

If you want to make Dockerfile what we've done

Create file named Dockerfile and type

FROM python:3.8.3-slim
RUN apt update && \
    apt -y install git && \
    git clone https://github.com/YoongiKim/DockerTutorial.git && \
    pip install flask
EXPOSE 5000
WORKDIR /DockerTutorial
ENTRYPOINT ["python"]
CMD ["app.py"]

docker build -t simple-flask-app .
docker run -it -p 5000:5000 simple-flask-app