Our project uses Docker to containerize a local dev environment and for deploying the app to production. This document describes the general setup for local dev, staging, and production.
There are three services configured in /docker-compose.yml, and managed using docker-compose. The project's main README.md includes the commands to manage the containers locally.
- node react-scripts/webpack dev server (port 3000)
- expungeservice uwsgi server hosting Flask app (port 5000)
Named Docker volumes provide persistent storage of node_modules files. To update node dependencies, edit package.json, and restart the container with make frontend_restart. Use make frontend_logs to watch the NPM install output, and use CTRL-C to stop watching once you see something like:
node_1 | You can now view record-expunge-pdx in the browser.
node_1 |
node_1 | Local: http://localhost:3000/
node_1 | On Your Network: http://192.168.80.4:3000/
node_1 |
node_1 | Note that the development build is not optimized.
node_1 | To create a production build, use npm run build.
Running tests either frontend and backend requires the respective container to be "up". Use the following make targets to run tests:
| target | description |
|---|---|
| test | run both frontend (no watch) and backend tests |
| frontend_test | run frontend tests in the node container in "watch" mode |
| frontend_test_no_watch | run frontend tests in the node container and exit |
We run a continuous integregation testing via GitHub Actions. This both frontend and backend tests, via the same image tags as used in development.
See main.yml
See DevOps README
Launch the docker-compose services with:
make up
The services are specified in the file docker-compose.yml. This target first pulls the dev-tagged docker image, and official image for node. It then creates and launches a new docker container for each service. It runs the docker stack in detached mode, so you can follow the set of container logs with docker-compose logs, or each specific one with make backend_logs or make frontend_logs. make backend_build will build the image recordsponge/expungeservice:dev if it can't pull or python dependencies change.
Stop the running containers with:
make down
To force a rebuild of the expungeservice docker image (if you change any package dependencies), use:
make backend_build
While the full dev stack is running, you can use the app in two different ways:
The frontend stack uses react-scripts which starts a hot-module-reloading dev server for quick iterative work. This is run inside the node service, which runs from the stock node:13.13.0-alpine image and mounts a persistent named volume to house the node_modules. This service is listening at:
The dev server is configured to proxy the backend API endpoints, so having the backend service running is required. Check out the make targets beginning with frontend_ for commented details.
In the case that one needs to blow away the node_modules tree and reinstall, use the following steps:
make frontend_downdocker volume rm recordexpungpdx_node_modulesmake upmake frontend_logsto watch install output, ctrl-c when done.
NOTE: a staging/prod deploy uses the output from npm run build and does not run any node service.
The backend stack uses a Docker Hub published image recordsponge/expungeservice with the :dev tag. This runs a reloading uwsgi server, handling the Flask app at app.py. This service is listening at:
The uwsgi server is run with --py-autoreload, so it will hot reload on detected source file chnages. The local src/backend and src/frontend trees are bind mounted into the container. Frontend source is in there so the Flask app can serve static files out of src/frontend/build, where an npm run build command would produce deployable artifacts. Check out the make targets beginning with backend_ for commented details.
To see which containers are running (and with the optional flag to see stopped containers also):
docker-compose ps