Skip to content

Latest commit

 

History

History
79 lines (61 loc) · 3.14 KB

README-DEVS.md

File metadata and controls

79 lines (61 loc) · 3.14 KB

Developer documentation

Development

Install Docker if you haven't already, and then get the Python dependencies:

$ git clone https://github.com/refinery-platform/django_docker_engine.git
$ cd django_docker_engine
$ pip install -r requirements.txt
$ pip install -r requirements-dev.txt
$ ./manage.py test --verbosity=2

To run it end-to-end, use the included demo server:

$ ./manage.py runserver &
# In your browser, visit: http://127.0.0.1:8000/docker/my-container/
# You should get a "please wait" page: We're waiting for "my-container" to start.
$ docker run --name my-container --publish 80 --detach nginx:1.10.3-alpine \
  --label io.github.refinery-project.django_docker_engine.port=80
# In a second, you should get the nginx welcome page.
# If your container did something useful, then you'd be seeing that instead.

For tests to pass, you'll need to add one entry to your /etc/hosts:

127.0.0.1 container-name.docker.localhost

/etc/hosts does not support wildcard entries: If you are spending too much time in that file, you might try running DNS locally with dnsmasq.

Other demos

Two other demos are provided, primarily as targets for automated tests. To the extent possible, all three use the same configuration with symlinks, and each just adds a few differences to the base.

  • demo_host_routing: Checks that hostname-based routing works.
  • demo_path_routing_auth: Checks that user authentication in the parent application doesn't interfere with authentication in the container.

Release Process

To make a new release, branch, increment the version number in VERSION.txt, and make a PR on github. If it passes tests, merge to master, and Travis will push to PyPi. (Travis will try pushing to PyPi on every merge to master, but unless the version is new, the push will fail.)

Key Dependencies

  • docker-py: The official Python SDK for Docker. It uses much the same vocabulary as the CLI, but with some subtle differences in meaning. It's better than the alternatives: calling the CLI commands as subprocesses, or hitting the socket API directly.

  • django-revproxy: Reverse proxy for Django. There are other options, but this seems to be the most mature and active.

Related projects

  • sidomo: Wrap containers as python objects, but assumes input -> output, rather than a long-running process.

  • Dockstore: Docker containers described with CWL.

  • BioContainers: A set of best-practices, a community, and a registry of containers built for biology. Preference given to BioConda?