Skip to content
This repository has been archived by the owner on Mar 21, 2019. It is now read-only.

uisautomation/media-webapp

Repository files navigation

This repository has moved to https://gitlab.developers.cam.ac.uk/uis/devops/media/webapp/

Streaming Media Service Webapp

CircleCI codecov

Documentation for developers, including a "getting started" guide, can be found at https://uisautomation.github.io/media-webapp.

Useful developer links

When a development environment is created via ./compose.sh development up, the following endpoints are available:

Additionally, endpoints for a mock hydra OAuth2 authorisation provider are created. See compose/base.yml for the exposed ports for those services.

Short, short version

Firstly, install docker-compose.

Secondly, copy the example secrets.env file to secrets.env and populate it with the various secrets required.

Then, most tasks can be performed via the compose.sh script:

# Start development server
$ ./compose.sh development

# Start development server in background
$ ./compose.sh development up -d

# View logs
$ ./compose.sh development logs

# Stop the development server
$ ./compose.sh development down

# Run tests
$ ./compose.sh tox run --rm tox

# Start a server using the production Docker image
$ ./compose.sh production build
$ ./compose.sh production up -d
$ ./compose.sh production exec production_app ./manage.py migrate

Additionally the tox.sh and manage_development.sh wrapper scripts provide convenient ways to run tox and management commands:

# Rebuild all testenvs
$ ./tox.sh -r

# Run only the flake8 tests
$ ./tox.sh -e flake8

# Run the migrate management command using the development images
$ ./manage_development.sh migrate

# Run tests and write coverage/documentation to build directory
$ ./compose.sh tox run -v $PWD:/tmp/workspace -e TOXINI_ARTEFACT_DIR=/tmp/workspace/build --rm tox

Developer guides

The following developer guides have been written:

  • [guides/uploading.md](Uploading new media items).

Dockerfile configuration

The following environment variables need to be set when running the application via Docker:

  • DJANGO_SECRET_KEY
  • DJANGO_DB_ENGINE, DJANGO_DB_HOST, DJANGO_DB_PORT, DJANGO_DB_PASSWORD, etc.
  • JWPLATFORM_EMBED_PLAYER_KEY
  • JWPLATFORM_API_KEY and JWPLATFORM_API_SECRET
  • OAUTH2_CLIENT_ID and OAUTH2_CLIENT_SECRET
  • CELERY_BROKER_URL: broker URL for celery. Defaults to amqp://guest:guest@localhost:5672//.

The following environment variables are optional:

  • DJANGO_SETTINGS_MODULE. Module on the Python import path which is used to load settings. Default is "mediawebapp.settings.docker".
  • SMS_ENABLE_DEBUG_THIS_IS_DANGEROUS. Enables DJANGO DEBUG if set.
  • DANGEROUS_DISABLE_HTTPS_REDIRECT. Disables HTTP to HTTPS redirect.

See the Dockerfile.

Connecting the React world and the Django world

This section talks about the configuration required to get the React frontend app to be served by the Django application retaining the live code recompiling offered by create-react-app. It also outlines how we build the frontend in production without littering the production container with various node-related detritus.

In production, we want the frontend to be served by the Django app and we'd like to be able to test this when developing. To allow Django to serve the app, we have some whitenoise configuration which arranges to serve files from a frontend build directory (configured via a setting) from /. If a request to /foo can be satisfied by a file from this directory, it is so otherwise the Django app takes a turn to satisfy the request.

We have an additional docker-compose service which makes use of the watch script we added to the frontend to watch for changes and build a new bundle of HTML/JavaScript in a docker volume. This same volume is mounted in the Django app container and is the directory whitenoise looks at.

The whitenoise configuration above is intended for use in production as well. We arrange for the frontend production build files to be in a directory in the production container and tell whitenoise to serve files from it.

The subtlety is in how those files get there.

We want the production image to remain small and so installing the entirety of node, npm and the node_modules directory seems troublesome since we'll never actually use them when running the container. Instead we make use of a useful but little-known feature in Docker called a multi-stage build. With this feature a Dockerfile can actually use multiple stages and only keep the last stage. We use this feature to compile the frontend JavaScript inside a node container and then copy the compiled JavaScript out of that container and in to the actual production container. This has the strong advantage that all of the node-related stuff then gets thrown away after the image is built.

Loading the legacy statistics schema

A current dump of the legacy statistics schema is available from a private URL. This dump can be downloaded and imported to the development app using the copy-legacy-stats.sh script. Once the application is up, the script can be run as follows:

./compose.sh development exec development_app /tmp/copy-legacy-stats.sh <stats_url>

(the stats url can be discovered in the main deployment template of the media project's deployment repository)

About

Web application for the University Media Platform

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published