you want to use, check at the end of the file for ready-to-run images or create your own Docker tools.

How to create a docker image

Define

• Dockerfile: image definition
• requirements.txt: list python packages you want to install
• Makefile: to build the image
• other: every data you want to go into the working directory of the machine

Then run make root # builds and run the image as ROOT user

IMPORTANT: Always add the proxy environment variables in you Dockerfile. It is necessary to build the image.

NOTE: The project you are working on may have already defined a Dockerfile. Start from it and complete it.

Usefule commands

Build image

docker build -t <IMAGE NAME> .

Display images in the Docker Image registry

docker images

Run the app (foreground)

docker run -p <IMAGE NAME>

Run the image built with nvidia docker as root, mount a volume and get access to bash

nvidia-docker run --volume=<LOCAL HOST PATH>:<DOCKER IMAGE PATH> -it -u root <IMAGE NAME> bash

Run the app (background)

docker run -d -p <IMAGE NAME>

Stop the background app: get the id and stop it

docker container ls # get CONTAINER ID
docker stop <CONTAINER ID>

Stop all containers

docker stop $(docker ps -a -q)

Remove an image by its IMAGE ID

docker container ls # get IMAGE ID
docker rmi <IMAGE ID>

Remove dangling images (i.e. useless images)

docker rmi $(docker images -f dangling=true -q)

Remove all images

docker rmi $(docker images -a -q)

Remove container by CONTAINER ID

docker ps -a # Get CONTAINER ID
docker rm <CONTAINER ID>

Remove all containers

docker rm $(docker ps -a -q)

Save change made in a container to a new image

docker run -it <IMAGE NAME> bash
# echo test_commit > test_commit.txt # Do some modifs
# exit
docker ps -a # Get <CONTAINER ID>
docker commit <CONTAINER ID> <NEW IMAGE NAME>:tag_name(optional)
docker images # Check that the new image exists
docker run -it <NEW IMAGE NAME> bash
cat test_commit.txt

Save your image on a remote repository

# Tag your image 
docker tag <IMAGE NAME>:<IMAGE TAG> <DOCKER USERNAME>/<REPO NAME>:<IMAGE TAG>
docker push <DOCKER USERNAME>/<REPO NAME>:<IMAGE TAG>

Remove images with the same but with different tags

docker rmi <REPOSITORY NAME>:<TAG NAME>

Error handling

Can't delete image because child images dependencies

Error message: Error response from daemon: conflict: unable to delete 8b88f06b72d7 (cannot be forced) - image has dependent child images
# Solution: Find the child image and delete them
$ docker inspect --format='{{.Id}} {{.Parent}}' $(docker images --filter since=<IMAGE ID> -q)
sha256:1d1f94c972fc8d75a34135f9ff551617680c508f7ad17efda09048b4e805g1w3
# The first 12 characters are the child <IMAGE ID>: 1d1f94c972fc

Ready-to-run images

Tensorflow

# For tensorflow 1.2.0 gpu support
docker pull tensorflow/tensorflow:1.2.0-gpu