Remove untagged images for ECR

remove-untagged-images-for-ecr

#!/bin/sh

set -eu

type aws > /dev/null 2>&1 && type jq > /dev/null 2>&1 || {
  echo 'Required aws-cli and jq'
  exit 1
}

repositories=$(aws ecr describe-repositories --query 'repositories[*].repositoryName' | tr -d "[\",\t\n\r\"]")

for repository in $repositories; do
  echo "$repository: Deleting untagged images"

  nextToken="initialized"

  while [ "$nextToken" != "null" ]; do
    if [ "$nextToken" == "initialized" ]; then
      response=$(aws ecr list-images --repository-name $repository --filter tagStatus=UNTAGGED --max-items 100)
    else
      response=$(aws ecr list-images --repository-name $repository --filter tagStatus=UNTAGGED --max-items 100 --starting-token $nextToken)
    fi

    nextToken=$(echo $response | jq -r '.NextToken')
    untaggedImages=$(echo $response | jq -c '.imageIds')

    if [ "$untaggedImages" == "[]" ]; then
      echo "$repository: Untagged image not found"
      break
    fi

    aws ecr batch-delete-image --repository-name $repository --image-ids $untaggedImages
  done

  echo "$repository: Deleted untagged images"
done

It works!

Didn't work for me initially because my default output format for the AWS CLI is configured as text. Adding this line to the top of the script fixed it for me: export AWS_DEFAULT_OUTPUT=json