Runbooks¶
Operational procedures for managing the Freeze Design webshop.
Environments
Two environments are live, both on a Hetzner VPS with Docker Compose in /opt/webshop.
Staging (https://staging.freezedesign.nl, basic auth) runs
docker-compose.staging.yml and deploys automatically from the staging branch.
Production (https://www.freezedesign.nl, live since 2026-09-30, v1.60 since
2026-11-06) runs docker-compose.prod.yml and deploys tagged releases (vX.Y) from
main with scripts/deploy-production.sh. freezedesign.nl and prod.freezedesign.nl
redirect to www. DNS is at Openprovider and points straight at the VPSes; nothing
sits in front of the shop.
Available Runbooks¶
In the order of the site navigation:
- Deployment - Staging and production deploys, zero-downtime rollout, rollback, migrations
- Disaster Recovery - Database backups and restore, media backup (rclone), VPS rebuild, SSL renewal
- Monitoring - Sentry, Celery Beat schedule (Dutch time), healthchecks.io pings, celery-beat healthcheck and auto-restart, health endpoints, PostHog and the weekly snapshot, external uptime monitor (to do)
- Alerting - Step-by-step response procedures per alert type
- Discord Webhooks - Which Discord channel gets what, and the webhook env vars and secrets
- ntfy notifications - The application alerts on the self-hosted ntfy server (next to Discord during the transition)
- Eefje Feedback Intake - Intake-proces voor bugs en wensen via de feedback-repo
- agent_api - De alleen-lezen API voor de AI-agent op de Freeze AI-VM: sleutels, scopes en IP-toegang
- Synthetics - Het testproduct en de bestelcheck die elk uur vanaf de Freeze AI-VM draait
- Productfeed (Merchant Center) - De productfeed voor Google Merchant Center, Meta en Pinterest
- Inrichting na de groei-review - Afvinklijst van accounts, sleutels en schakelaars na de groei-review
- Nieuwsbrief (Resend) - Aanmelden met double opt-in, Resend-audience, versturen, storingen
- Reviews (Google) - Reviewverzoek na levering en het Google-reviewblok
- Winkelmail bij een bestelling - Mail "Nieuwe bestelling" aan de winkel bij elke betaalde order
- Betaalmethoden (Mollie) - Meer betaalmethoden dan iDEAL aanzetten via Mollie en de admin
- Vindbaarheid (Search Console) - Search Console, Bing Webmaster Tools, sitemap en structured data
- Dagplanning - De nachtelijke planner, de plan-PR,
/dagstarten de bewaker - Voor Ivar (de lijst) - Het ene vastgepinde issue met alles wat alleen Ivar kan doen
- E2E-quarantaine - Welke Playwright-tests geparkeerd staan en hoe ze weer meedraaien
- Staging-inlog wisselen - Naam en wachtwoord van de basic auth op staging vernieuwen, en het GitHub-secret bijwerken
- Catalogus van productie naar staging - Producten, prijzen en foto's van productie op staging zetten met
scripts/catalogus-prod-naar-staging.sh
Quick Reference¶
On a VPS, as the deploy user. Neither server has a docker-compose.yml, so select the
compose file once per SSH session:
Check Application Status¶
# Containers and their Docker health (celery-beat too, since #798)
docker compose ps
# Backend logs
docker compose logs backend --tail 100 -f
# Frontend logs
docker compose logs frontend --tail 100 -f
Restart Services¶
docker compose restart backend
docker compose restart frontend
docker compose restart celery celery-beat
A restart keeps the old image and the env file the container was created with. After a
change to backend/.env.production (or .env.staging), recreate instead:
docker compose up -d backend celery celery-beat.
Database Access¶
# Credentials come from backend/.env.production (POSTGRES_USER / POSTGRES_DB)
docker compose exec db psql -U "$POSTGRES_USER" -d "$POSTGRES_DB"
Incident Response¶
Priority Levels¶
| Level | Response Time | Examples |
|---|---|---|
| P1 | Immediate | Site down, payment failures |
| P2 | 1 hour | Major feature broken |
| P3 | 4 hours | Minor bugs, degraded performance |
| P4 | Next business day | UI issues, minor improvements |
Escalation Path¶
For a solo-operator setup, "escalation" means increasing your own response urgency. See the Alerting Runbook for detailed response procedures per alert type.