aws-ecs-cloudreactor-deploy is a Docker image that is able to deploy Tasks to AWS ECS and CloudReactor. aws-ecs-cloudreactor-deploy uses Ansible to execute a playbook that includes steps to resolve the configuration properties, before deploying to both AWS and CloudReactor.
Out of the box, it only supports unencrypted secrets, or secrets encrypted with Ansible Vault. It is geared toward code written in dynamic languages that do not require compilation. However, you can customize the image to support other ways of handling secrets and additional build steps such as compilation.
You can run custom build steps by adding steps to
deploy_config/hooks/post_build.yml as necessary.
If you need to use libraries (e.g. compilers) not available in the aws-ecs-cloudreactor-deploy image, your custom build steps can either:
1) Use the
docker command to build intermediate files (like JAR files or executables). Use
docker build to build images,
docker create to create containers, and finally,
docker cp to copy files from containers back to the host. When docker runs in the container, it will use the host machine’s docker service.
2) Use build tools installed in the deployer image. In this case, you’ll want to create a new image based on
``` FROM cloudreactor/aws-ecs-cloudreactor-deployer:1.1.0 # Example: get the JDK to build JAR files RUN apt-get update && \ apt-get -t stretch-backports install openjdk-11-jdk ... ```
Then set the
DOCKER_IMAGE environment variable to the name of your new image, or change the deployment command in
deploy.sh to use your new image instead of
cloudreactor/aws-ecs-cloudreactor-deployer. Your ansible tasks can now use
javac. If you create a Docker image for a specific language, we’d love to hear from you!
Also, check out multi-stage Dockerfiles as a way to build dependencies in the same Dockerfile that creates the final container. This may complicate the use of the same Dockerfile during development, however.
deploy.sh is what you’ll call on your host machine, which will run the Docker image for the deployer. The deployer Docker image has an entrypoint that executes the python script deploy.py, which in turn, executes ansible-playbook.
The Ansible tasks in
ansible/deploy.yml reference files that you can make available with Docker volume mounts. You can either modify
deploy.sh to add or modify existing mounts, or configure the files/directories with environment variables. The Ansible tasks also read environment variables which you can set in
You can configure some settings in
deploy.sh with environment variables if you want to avoid modifying it:
|Environment variable name||Default value||Description|
|DOCKER_CONTEXT_DIR||Current directory||The absolute path of the Docker context directory|
|DOCKERFILE_PATH|| ||Path to the Dockerfile, relative to the Docker context|
|CLOUDREACTOR_TASK_VERSION||Empty||A version number to report to CloudReactor. If empty, the latest git commit hash will be used.|
|PER_ENV_SETTINGS_FILE|| ||Path to a dotenv file containing environment-specific settings|
|EXTRA_DOCKER_RUN_OPTIONS||Empty||Additional options to pass to |
|DEPLOY_COMMAND|| ||The command to use when running the image. Can be set to |
|DOCKER_IMAGE|| ||The Docker image to run. Can be set to another name in case you extend the image to add build tools.|
|DOCKER_IMAGE_TAG|| ||The tag of the Docker image to run.|
The behavior of ansible-playbook can be modified with many command-line options. To pass options to ansible-playbook, add
--ansible-args to the end of the command-line for
deploy.sh. Follow that with all the options you want to pass to ansible-playbook. For example, to use secrets encrypted with ansible-vault and get the encryption password from the command-line during deployment:
./deploy.sh staging --ansible-args --ask-vault-pass
Alternatively, you can use a password file:
./deploy.sh staging --ansible-args --vault-password-file pw.txt
The password file could be a plaintext file, or a script like this:
#!/bin/bash echo `aws s3 cp s3://widgets-co/vault_pass.$DEPLOYMENT_ENVIRONMENT.txt -`
If you use a password file, make sure it is available in the Docker context of the container. You can either put it in your Docker context directory or add an additional mount option to the docker command-line.
You can customize the build even more by overriding any of the files in the
ansible directory of aws-ecs-cloudreactor-deployer with you own version, by passing a volume mount option to the Docker command line. For example, to override
deploy.yml, pass these options to the Docker command line in
-v $PWD/ansible_overrides/ansible.cfg:/work/ansible.cfg -v $PWD/ansible_overrides/deploy.yml:/work/deploy.yml
aws-ecs-cloudreactor-deploy uses templates contained in
/work/templates of its Docker image:
- The ECS task definition is created with the Jinja2 template
- The CloudReactor Task is created with the Jinja2 template
deploy/templates/cloudreactor_task.yml.j2. which produces a YAML file that is converted to JSON before sending it CloudReactor.
These templates use settings from the files described above. If you need to modify the templates, you can override the default templates similarly: