DevOps for Startups - What You Actually Need (And What to Skip)

Most DevOps content is written for companies with 50-person platform teams. Here's what actually matters when you have 2–5 engineers.

---

The Minimal Stack That Scales to $1M ARR

Hosting:     1–2 VPS (Hetzner, DigitalOcean, or dedicated server)
Containers:  Docker + Docker Compose
CI/CD:       GitHub Actions
SSL:         Let's Encrypt via Certbot or Caddy
Monitoring:  Uptime Robot (free) + Sentry (errors) + basic logging
Backups:     Automated daily DB backups to S3 or Backblaze B2

That's it. No Kubernetes, no Terraform, no service mesh. Introduce complexity when you have a specific problem, not in anticipation of problems.

---

CI/CD: The One Thing You Must Have

Every commit merged to main should:

1. Run tests
2. Build Docker image
3. Deploy to staging
4. (On approval) Deploy to production

GitHub Actions makes this 50 lines of YAML. If you're deploying by SSH-ing into a server and running git pull, you will ship broken code to production eventually.

# .github/workflows/deploy.yml (simplified)
on:
  push:
    branches: [main]
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Run tests
        run: npm test
      - name: Build and push Docker image
        run: docker build -t myapp:${{ github.sha }} .
      - name: Deploy to server
        run: ssh deploy@myserver "docker pull myapp:${{ github.sha }} && docker compose up -d"

---

Docker: Use It From Day One

Docker ensures "works on my machine" == "works in production." Your development environment, CI, staging, and production all run identical containers.

Docker Compose handles multi-container setups (app + database + Redis + Nginx) with one file. For most startups this is all you need.

---

Monitoring: The Minimum

Uptime Robot - pings your endpoints every 5 minutes, alerts via Telegram/email when down. Free tier is sufficient.

Sentry - catches unhandled errors in frontend and backend with stack traces. Free tier: 5k errors/month. Invaluable.

Server metrics - CPU, memory, disk via Netdata or Grafana + Prometheus when you need it. Don't set this up on day one.

---

When to Actually Consider Kubernetes

• You have 10+ microservices
• You need auto-scaling (traffic patterns that 10× in minutes)
• You have a dedicated DevOps engineer

Before that: Docker Compose + 2 servers (primary + hot standby) handles more load than most startups ever see.

---

Backup Strategy (Non-Negotiable)

1. Daily PostgreSQL dump → compressed → uploaded to Backblaze B2 ($0.006/GB)
2. Keep 30 days of backups
3. Restore test monthly - backups you've never tested aren't backups

One pg_dump command in a cron job. This has saved multiple clients from total data loss.

Talk to us about infrastructure setup →