Docker is awesome for developing twelve-factor apps.
This is a brief guide for satisfying key factors of twelve-factor with Docker tools. It explains what techniques you can use to develop a archetypical Rails / Postgres / Redis / web / worker application under Docker.
A future post will deep dive into a code base that demonstrates everything working together.
II. Dependencies — Explicitly declare and isolate dependencies
With these tools you can say that you want a Ubuntu 14.04 OS, Ruby 2.2.2, Node 0.11 system to run `bundle install` against.
III. Config — Store config in the environment
Docker containers rely heavily on linux environments for configuration.
docker-compose.yml has an environment hash where you explicitly define what environment variables will be set in a container. These can be default values or undefined values that will be inherited from the host at runtime.
With these tools you can say that your application expects a GITHUB_AUTH_TOKEN.
VII. Port binding — Export services via port binding
Docker containers rely heavily on port binding.
With this tool you can say that your application web server is listening on PORT 5000, and that you can access it from the host on PORT 5000.
IV. Backing Services — Treat backing services as attached resources
Docker containers share little to nothing with other containers and therefore should communicate with backing services over the network.
docker-compose.yml has a links hash where you specify the other containerized services that your application depends on. `docker-compose up` will start these backing services first and set environment variables in the app container with network connection information.
With these tools you can say that your application depends on a backing Postgres 9.4 and Redis 3.0 service and have your application connect to them via hostname and port information present in their environment.
VI. Processes — Execute the app as one or more stateless processes
By default, Docker containers are shared nothing processes with ephemeral storage.
docker-compose.yml defines a collection of services, each one with its own image or build recipe and command.
With this tool you can say that your application has a both a web and worker process.
XII. Admin processes — Run admin/management tasks as one-off processes
Docker images are easily run as one-off processes.
`docker run myapp CMD` will run an arbitrary command in the same environment as your web process.
With this tool you can run an interactive bash prompt or run one-off `rake db:migrate` process against your development Postgres database.
Without Docker, the OS X development tool chain is Homebrew for system dependency packages and development services like Postgres and Redis, Ruby’s Bundler for cross-platform development dependencies, a combination of shell scripts and foreman to coordinate running everything locally at the same time, and an independent Linux-based build service to package everything up for production.
There is nothing wrong with this workflow, however Docker offers a more elegant future.
With the right Dockerfile and docker-compose.yml files, we no longer need any OS X system dependencies, service packages, or cross-platform language dependencies. A simple `docker-compose up` can provide a complete Linux development environment that will easily port onto a production machine for any twelve-factor app.
Please direct feedback and/or questions via Twitter to @nzoschke or email to email@example.com.