Project · Ship the Day 0 stack

Hints

• Pick ONE language and ONE stack. Resist the urge to compare three options — it's a Day 0 stack, not a benchmarking exercise.
• If you've never used systemd, the unit file in the demo is enough. Don't over-engineer it.
• For the restore drill, use Docker locally if you don't have a spare DB instance.
• If you're on a managed DB, the backup is already on. You still have to do the restore drill — managed backups are not verified by the provider.

Expected output

$ curl -s https://yourdomain.com/healthz
ok

$ curl -s https://yourdomain.com/readyz | jq
{ "db": "ok", "redis": "ok" }

$ sudo systemctl stop redis    # break a dep
$ curl -s -o /dev/null -w "%{http_code}\n" https://yourdomain.com/readyz
503

$ # restore drill
$ docker run -d --name drilldb -e POSTGRES_PASSWORD=t -p 5433:5432 postgres:16
$ zcat latest-backup.sql.gz | docker exec -i drilldb psql -U postgres
$ docker exec drilldb psql -U postgres -c "SELECT count(*) FROM users;"
 count
-------
 14237

Stretch goals

• Add automated unattended-upgrades + a reboot window. Verify a security patch gets applied next time you check.
• Move the systemd unit + Caddyfile + base config into a small Ansible (or Terraform + cloud-init) script so the box is reproducible.
• Add a second VM in the same region, point Caddy at both (round-robin), and see what breaks. (Spoiler: sessions, file uploads — preview of Module 6.)
• Run wrk against the Day 0 stack from a third VM. Record p50/p95/p99 for a /users listing endpoint. This is your baseline for the rest of the course.