Bash script to remove Docker Images using a whitelist and keeping a specified number of copies ... beyond what "docker system prune" does

docker_image_purge.sh

#!/bin/bash
#
# Usage
# ~~~~~
# docker_image_purge list
# docker_image_purge [ remove ]
#
# By default, this command only shows what Docker images it'll remove.
# Use the optional "remove" command line parameter to actually remove them !
#
# Internal variables that may be configured ...
# - DO_NOT_DELETE: Whitelist of Docker image repositories not to purge
# - MAXIMUM_SIMILAR_IMAGES: Number of images to preserve
#
# To Do
# ~~~~~
# - Command line parameter: Specific Docker image repository to clean-up
# - Pretty print Docker images, in order of grouped repository image size

LIST=${1:-DONT_LIST}
REMOVE=${1:-DONT_REMOVE}

MAXIMUM_SIMILAR_IMAGES=2

if [ "$LIST" == "list" ]; then
  docker images | tr -s " " | sort | awk '{print $3,$7,$1,$2}' | column -c 80 -t
  exit 0
fi

DO_NOT_DELETE=\
nvidia/cuda\|\
python\|\
tensorflow/tensorflow

free_disk_space() {
  disk_partition=$1
  df -h $disk_partition | tail -n +2 | {
    read -r -a tokens
    disk_space_available=${tokens[3]}
    echo "Free space on " $disk_partition "is" $disk_space_available
  }
}

echo Before: $(free_disk_space /home/users)

docker_images() {
  docker images | tail -n +2
}

DOCKER_IMAGES=$(docker_images)

docker_images_count() {
  cut -d" " -f1 | sort | uniq --count
}

sort_numeric() {
  sort --numeric-sort --reverse
}

docker_image_count_list() {
  echo "$DOCKER_IMAGES"         |   \
  echo "$(docker_images_count)" |  \
  echo "$(sort_numeric)"
}

docker_image_ids_for_repository() {
  desired_image_repository=$1

  echo "$DOCKER_IMAGES" | while read -r line
  do
    echo "$line" | {
      read -r -a tokens
      docker_image_repository=${tokens[0]}
      docker_image_id=${tokens[2]}
      if [ $desired_image_repository == $docker_image_repository ]; then
        echo $docker_image_id
      fi
  }
  done
}

filtered_docker_image_ids() {
  docker_image_count=$1
  docker_image_repository=$2

  echo $docker_image_repository | egrep -v $DO_NOT_DELETE >/dev/null
  if [ $? -eq 0 ]; then
    if [ $docker_image_count -gt $MAXIMUM_SIMILAR_IMAGES ]; then
      echo "$DOCKER_IMAGES"         |   \
      echo "$(docker_image_ids_for_repository $docker_image_repository)"
    fi
  fi
}

lines=$(docker_image_count_list)

echo "$lines" | while read -r line
do
  echo "$line" | {
    read -r -a tokens
    docker_image_count=${tokens[0]}
    docker_image_repository=${tokens[1]}

    FILTERED_DOCKER_IMAGE_IDS=$(filtered_docker_image_ids $docker_image_count $docker_image_repository | tail -n +$(($MAXIMUM_SIMILAR_IMAGES + 1)) | uniq)

    image_id_count=`echo "$FILTERED_DOCKER_IMAGE_IDS" | wc -w`

    if [ $image_id_count -gt 0 ]; then
      if [ "$REMOVE" == "remove" ]; then
        action="Removing"
      else
        action="Could remove"
      fi
      echo $action $image_id_count $docker_image_repository "image(s)"
      if [ "$REMOVE" == "remove" ]; then
        docker rmi --force $FILTERED_DOCKER_IMAGE_IDS
      fi
    fi
  }
done

echo After: $(free_disk_space /home/users)

if [ "$REMOVE" != "remove" ]; then
  echo 'Add "remove" parameter to actually remove the Docker images'
fi