FixBoard

A troubleshooting Q&A board for DevOps/Cloud/SRE/Platform Engineering students. Posts capture full problem context environment, logs, architecture so answers become a searchable, reusable knowledge base instead of one-off chat messages.

This is a teaching project. The stack is kept deliberately simple so the complexity budget stays available for the Kubernetes deployment lessons that come after this.

Stack

• Backend: Node.js + Express + PostgreSQL (backend/)
• Frontend: React (Vite build), served as a static bundle via Nginx (frontend/)
• Local orchestration: Docker Compose (this README)

Quick start

Requirements: Docker and Docker Compose.

docker compose up --build

That's it no manual setup required. This brings up three services:

Service	URL	What it is
frontend	http://localhost:8090	The web app
backend	http://localhost:3000	The API (migrations + seed run automatically on boot)
postgres	localhost:5432	Database (not needed directly)

Open http://localhost:8090 and log in with a seeded test account:

Role	Email	Password
Student	test@fixboard.dev	password123
Instructor	instructor@fixboard.dev	password123

To stop everything: docker compose down. To also wipe the database and uploaded files: docker compose down -v.

Configuring it

Defaults are baked into docker-compose.yml, so the command above works with zero configuration. To override anything (ports, credentials, JWT secret), copy the env file and edit it:

cp .env.example .env

docker compose picks up .env automatically. See .env.example for every variable.

Note: FRONTEND_PORT and POSTGRES_PORT are read at container start time, but BACKEND_PORT is also baked into the frontend's JavaScript bundle at build time (Vite inlines it as the API base URL). If you change BACKEND_PORT, rebuild: docker compose up --build.

Architecture notes

• File uploads land on a named Docker volume (uploads_data) mounted into the backend container at /app/uploads, and are served back out under /uploads/*. The backend only ever treats this as "a writable directory it was handed" — swapping it for an EFS-backed PVC at the Kubernetes stage is a volume-mount change, not a code change.
• Soft delete: instructor deletes (of posts or comments) set a deleted_at timestamp rather than removing rows. This sidesteps cascade edge cases (a deleted comment with replies, a deleted post's attachments/notifications) while keeping the FK graph intact. Deleted rows are simply filtered out of every read query.
• Notifications are a minimal unread-count model, not an inbox: a row is written when someone comments on your post (not when you comment on your own), the frontend polls GET /notifications/unread-count for the nav badge, and opening My Posts marks everything read.

Repo layout

backend/    Express API — see backend/README.md for endpoints, schema, local (non-Docker) run instructions
frontend/   React app — see frontend/README.md for local (non-Docker) dev instructions
docker-compose.yml
.env.example

For day-to-day frontend/backend development (hot reload, debugging one service in isolation) see the README in each subdirectory — Compose is the "run the whole thing like production" path, not the dev loop.

End-to-end flow this was verified against

Signup (student) → browse dashboard → filter by tag → create a post with an architecture image and a log file → a second student comments on it → original poster sees an unread badge on My Posts → opens the post, marks a comment as the solution → an instructor deletes a test post and a test comment.