From Laptop to Internet: How Deployment Works
On your laptop, your app runs with python run.py and only you can see it. Deploying means running it on a computer that is always on, reachable at a domain, over HTTPS, with real data, safely.
The pieces of a live web app
| Piece | Job | Examples |
|---|---|---|
| Hosting | runs your code 24/7 | Render, Railway, Fly.io, a VPS, AWS |
| App server | serves Python requests in production | Gunicorn (Flask/Django), Uvicorn (FastAPI) |
| Database | stores data permanently | managed PostgreSQL |
| Static/media files | images, CSS, uploads | the app, a CDN, S3 or R2 |
| Domain + DNS | yourapp.com → your host |
Cloudflare, Namecheap |
| TLS certificate | HTTPS | Let's Encrypt (usually automatic) |
| Environment variables | secrets and settings | the platform's dashboard |
| Monitoring | know when it breaks | health checks, logs, Sentry |
Hosting options
- Platform as a Service (PaaS): Render, Railway, Heroku, Fly.io. You push code; they build and run it. Least work, slightly higher cost. Best for most small apps.
- Virtual private server (VPS): Hetzner, DigitalOcean. A Linux machine you manage yourself. Cheapest for the power, but you handle security, updates and backups (see Linux System Administration).
- Serverless/edge: Vercel, Netlify, Cloudflare. Ideal for static sites and frontends; backends run as functions.
- Cloud (AWS, GCP, Azure): everything, at the cost of complexity.
The development vs production gap
| Development | Production |
|---|---|
debug=True |
debug off (debug pages leak code and allow code execution) |
| SQLite file | managed PostgreSQL |
flask run dev server |
Gunicorn with several workers |
secrets in .env |
secrets in the platform's environment settings |
| errors on screen | errors in logs and error tracking |
The twelve-factor ideas that matter most
- Config in the environment, not in code.
- Dependencies declared (
requirements.txtpinned). - Stateless processes: do not store uploads or sessions on the server's disk if you might run more than one instance or redeploy (disks on many platforms are wiped on deploy).
- Logs to stdout: the platform collects them.
- Dev/prod parity: keep environments as similar as possible.
Check yourself
- Why must debug mode be off in production?
- What is the trade-off between a PaaS and a VPS?
- Why should uploads not live on a PaaS server's local disk?
Ihr Fortschritt wird nur in diesem Browser gespeichert.
Zur Prüfung einreichen (10 Punkte)