Skip to content
This repository has been archived by the owner on Jun 1, 2022. It is now read-only.

CAVaccineInventory/vial

Repository files navigation

VIAL

VIAL = Vaccine Information Archive and Library.

This is the Django application that powers https://help.vaccinateca.com/ and provides the backend data for https://www.vaccinateca.com/

Project background: Spinning up a new Django app to act as a backend for VaccinateCA

Where this is hosted

API exports

VIAL exports data to static JSON (and GeoJSON) in Google Cloud buckets for use by other applications. These exports are triggered by Google Cloud Scheduler - see https://console.cloud.google.com/cloudscheduler?project=django-vaccinateca

Velma and Scooby

Scooby is our app for filing reports against locations.

Velma is our app for verifying details about new imported locations and handling potential duplicate merges.

Auth0 user permissions

This app is built around the Django admin, but uses Auth0 for authentication.

User permissions are controlled using Auth0 roles. Users can be assigned these roles in the Auth0 interface at https://manage.auth0.com/dashboard/us/vaccinateca/roles

The following Auth0 roles grant access to VIAL, and place the user in a VIAL permissions group with the same name as the Auth0 role. The permissions assigned to those groups can be seen and edited in the VIAL groups interface, by users with the VIAL super-user role.

  • Vaccinate CA Staff
  • Reports QA
  • VIAL data corrections
  • VIAL super-user

Membership of these groups should be controlled entirely through Auth0 role assignments. If a VIAL super-user adds or removes someone from one of these groups using the VIAL interface that user will have their membership reset next time they sign into VIAL.

If you want to grant permissions to specific users within VIAL independent of their AUth0 roles you can do so by editing that user's list of permissions directly on the edit user page.

Architectural principles for this app

  • Write code (and issue comments and commit messages) with the expectation that the entire repository will be open to the public some day. So keep secrets out of the code, and don't be uncouth!
  • As few moving parts as possible. Right now this means:
    • The app is written in Django
    • All data is stored in PostgreSQL - even data that might be a better fit for a dedicated logging system or message queue. We'll adopt additional storage mechanisms only when PostgreSQL starts creaking at the seams.
  • Django migrations are great. We use these enthusiastically, with a goal of making schema changes boring and common, not exciting and rare.

API documentation

Comprehensive documentation for the VIAL web API is available in docs/api.md. This documentation should be updated in sync with changes made to the APIs - any time the documentation becomes out of sync with the implementation should be considered a bug and fixed ASAP!

This documentation is also included in VIAL deployments:

Setting up a development environment

The recommended development environment for VIAL uses Docker Compose. It has been tested using Docker for Mac, which can be installed using brew install --cask docker.

Check out the vial repository:

git clone [email protected]:CAVaccineInventory/vial

Then cd vial and start Docker Compose:

docker-compose up

This should run three containers: database_1, running PostgreSQL (and PostGIS), web_1 running a Django development server and migrations_1 which will run any database migrations and then stop.

The web_1 container will start a server on port 3000 - you can then access the web application at http://0.0.0.0:3000/

If you run into issues during environment setup, start with the FAQ for troubleshooting.

If you encounter any issues in setup, please document them and add them to the FAQ.

Visit it at http://0.0.0.0:3000/ - it's important to use 0.0.0.0:3000 as that is the URL that is allow-listed for logins by the Auth0 configuration. Click "sign in" and sign in with an Auth0 account.

Once you have signed in and created an account you should grant yourself super-user access so you can use every feature of the admin site. You can do that by running the following:

docker exec -it vial ./manage.py shell
>>> User.objects.all().update(is_superuser=True, is_staff=True)
>>> <Ctrl+D> to exit

Finally, run a command to import the details of 3,000+ counties into your database. Make sure you are signed in using your super-user account, then visit this page:

http://0.0.0.0:3000/admin/tools/

And click the "Import 3,000+ counties" button.

Importing sample data from live or staging

Two scripts are provided for importing data from our live or staging instances into your development environment, using the export and import APIs.

To use these scripts, you will need two API keys: one for your development environment (which you can create at http://0.0.0.0:3000/admin/api/apikey/) and one for the production or staging environment that you wish to import data from.

To import locations:

docker exec -it vial python /app/scripts/dev_copy_locations.py \
   --source-token '17:ce619e0...' \
   --destination-token '4:09b190...' \
   --source-url 'https://vial.calltheshots.us/api/searchLocations?size=100'

Where source-token is the API key from production, and destination-token is your API key created for your local development environment.

You can pass different arguments in the query string for --source-url to retrieve different subsets of our data. If you wanted to import every location in Puerto Rico for example you could use:

   --source-url 'https://vial.calltheshots.us/api/searchLocations?state=PR&all=1'

Be careful with the all=1 argument - used carelessly you could accidentally pull 60,000+ records into your local environment!

Source locations are raw, unprocessed location data gathered by our vaccine-feed-ingest mechanism.

You can import these in a similar way to locations, using this script:

docker exec -it vial python /app/scripts/dev_copy_source_locations.py
  --source-token '17:ce619e0...' \
  --destination-token '4:09b19...' \
  --destination-url 'https://vial.calltheshots.us/api/searchSourceLocations?state=RI&source_name=vaccinefinder_org'

This will import all of the source locations in Rhode Island that were originally imported from vaccinefinder_org.

Running the tests

Run the tests like this:

docker exec -it vial pytest

You can pass extra arguments to run specific tests - for example:

  • docker exec -it vial pytest -k test_admin_location_actions_for_queue - run specific test
  • docker exec -it vial pytest api/test_export_mapbox.py - run the tests in the vaccinate/api/test_export_mapbox.py module
  • docker exec -it vial pytest --lf - run tests that failed during the last test run

Code formatting and linting

This repository uses Black and isort to enforce coding style as part of the CI tests.

Run docker exec -it vial black . and docker exec -it vial isort . to ensure your code is formatted correctly, then enjoy never having to think about how to best indent your Python code ever again.

Run docker exec -it vial /app/scripts/run-flake8 to check for missing or unused imports.

Run docker exec -it vial /app/scripts/run-mypy run the mypy type checker.

Run docker exec -it vial bash -c 'cd /app && /app/scripts/lint-migrations' to verify that migrations do not have any backwards-incompatible changes that could cause problems during a deploy while the site is serving traffic.

Rebuilding the containers

Any time you need to apply changes to requirements.txt or Dockerfile.dev you should stop the conatiners and run this:

docker-compose build

Then start the environment again with:

docker-compose up

Logging SQL

You can set the DJANGO_DEBUG_LOG_ALL_SQL=1 environment variable to log all SQL executed by Django to the console. This can be useful for things like understanding how complex migrations work:

DJANGO_DEBUG_LOG_ALL_SQL=1 ./manage.py migrate

How to deploy

Deploys work by pushing to the production branch in this repository. The scripts/deploy.sh script can run this for you - it needs to be run outside of your Docker container as it uses your regular GitHub SSH credentials:

% scripts/deploy.sh
You are about to deploy the following commits:

25badf0 Ran isort
f7c45bb  Report field internal notes labeled private notes
613fb89 Tell #vial-app-server about deploys

Type 'yes' to confirm they look right - paste the above into #vial-app-server so people know about the deploy

Copy and paste the commits into #vial-app-server on Discord so people know about the deploy and type "yes".

Errors in both production and staging are recorded in Sentry. You can view those here - ask in #access-requests if you do not have access to Sentry.

Django SQL Dashboard

https://vial.calltheshots.us/dashboard/ and https://vial-staging.calltheshots.us/dashboard/ offer an interface for running read-only SQL queries against our database and bookmarking the results, using Django SQL Dashboard.

You can create saved dashboards at https://vial.calltheshots.us/admin/django_sql_dashboard/dashboard/ - these will then be available at URLs like https://vial.calltheshots.us/dashboard/closest-locations/

Only a specific list of tables are available through that interface. To make a new table available follow these instructions:

  1. Make sure you're auth'd at the command line as someone who can connect to the database; you may need to bootstrap to a service account; the process there is:
    # Fetch a key
    gcloud iam service-accounts keys create ./staging-key.json --iam-account [email protected]
    # Enable it:
    gcloud auth activate-service-account --key-file=./staging-key.json
    
  2. Download and start the cloud SQL proxy:
    ./cloud_sql_proxy -instances=django-vaccinateca:us-west2:staging=tcp:5432
    
  3. Look up the django-staging-env secret in Secret Manager, click the triple-dots, and show the current value; pull out the secret for the vaccinate user from the DATABASE_URL line.
  4. Connect with that password:
    psql -U vaccinate -h localhost -p 5432 vaccinate
    
  5. Grant the rights:
    grant select on
      public.api_apikey,
      public.api_log
    to "read-only-core-tables";
    
    grant select on
      public.api_apikey,
      public.api_log
    to "datascience";
    

...then repeat that for production. The read-only-core-tables role is used by Django SQL Dashboard, and the datascience role is used by Mode.

About

The Django application powering calltheshots.us

Resources

License

Stars

Watchers

Forks

Languages