Development and Deployment with Docker¶

Connect ships with support for Docker.

The Dockerfile will install all the required dependencies for the build, as well as a script called proclaunch. As the Dockerfile does not have a CMD or ENTRYPOINT which launches all of Connect (nor does it have all the required settings pre-defined), to launch each process you’ll need to run proclaunch web, proclaunch scheduler, and/or proclaunch worker to run the correct process.

By default there are not enough environment settings to fully launch Connect. You’ll want to create a .env file to pass into docker which contains the settings in Connect Settings.

In order to have dev/production parity, you’ll want to also launch docker containers for Memcached and RabbitMQ and set the BROKER_URL and CACHE_URL to link to those. docker-compose.yml contains example code for running web, scheduler and worker processes.

Note

The first time you build the dockerfile (either via the standard docker build process or via docker-compose build) your machine will have to pull in all the dependencies for both frontend and backend code. Subsequent builds will skip these steps. If you want to re-pull dependencies, the easiest step is to change requirements.txt for python packages, package.json for node packages, or bower.json for static dependencies. Docker will scan these files for changes during each build process, and if there is a change it will re-run that part of the build process.

Warning

While docker-compose.yml is a good starting-point for knowing what settings should be defined, it’s written for development.