← all talks

Homelab Renaissance: 21 Stacks, 200 Services, Full Autonomy

Homelab Renaissance: 21 Stacks, 200 Services, Full Autonomy. By Erwan Rougeux. October 25, 2025, Sunway University, Kuala Lumpur, Malaysia. Fork It Community.

Who am I?

VP, International Growth @ Clever Cloud (sovereign PaaS).

Passionate about software development & cloud architecture; open-source and privacy advocate.

Love privacy-by-design.

Productively paranoid: if it's not self-hosted, encrypted, and monitored, I don't sleep.

Cartoon avatar of Erwan

The shift: from owning to renting

  • We used to own CDs, DVDs, books → local, offline, durable.
  • Now: streams, subscriptions, pay-per-use, geo-locks.
  • Access can be revoked; data gets extracted; costs creep up.
  • Privacy & resilience suffer.
Illustration: a DVD and books turning into streamed media

Why self-hosting?

  • Control: data locality, portable formats, reversible choices.
  • Cost: avoid subscription sprawl; reuse commodity hardware.
  • Privacy: reduce data exhaust; least-privilege access.
  • Resilience: independent recovery path; fewer external dependencies.

Architecture overview

  • Ingress: Traefik (TLS), Keycloak (SSO/MFA), CrowdSec (reputation).
  • Orchestration: Docker + Portainer (per stack).
  • Network privacy: Pi-hole + Unbound, Gluetun (egress VPN).
  • Access: Tailscale & Headscale (LAN-over-WAN).
  • Storage: Unraid + ZFS (SSD cache → HDD pool).
  • Observability & DR: Prometheus/Telegraf + Grafana; ZFS replication.

Stack map

  • Security & network: Traefik • Keycloak • CrowdSec • Pi-hole • Unbound • Gluetun • Tailscale/Headscale
  • Orchestration & storage: Docker • Portainer • Unraid • ZFS
  • Data & collaboration: Nextcloud • OnlyOffice • Imaginary • Full-Text Search • ClamAV • Redis • DB
  • Media & streaming: Radarr • Sonarr • Lidarr • Readarr • Prowlarr • FlareSolverr • Calibre-Web/CalibreVNC • AutoM4B • Plex • Kodi
  • Home & sensing: Home Assistant • Grocy • ebusd • ESPHome • Node-RED • MQTT • Speedtest Tracker • Signal
  • AI & dev: Ollama • Open WebUI • code-server
  • CCTV: Frigate • go2rtc
  • Passwords: Vaultwarden (OTP stored elsewhere)
  • Sensors & lab: InfluxDB (Mycodo)

What it looks like

The full homelab network map, blurred on purpose: dozens of stacks and hosts wired together

The real map, deliberately blurred: addresses and hostnames stay at home.

Some stacks

In depth

Personal files

  • Nextcloud as the personal hub (files / photos / mobile sync).
  • OnlyOffice for docs; Full-Text Search; Imaginary for images.
  • ClamAV / Redis / DB behind the scenes; quotas & lifecycle rules.
  • Backups per dataset (ZFS snapshots → replicated).

Media & streaming

  • *arr pipeline: Prowlarr/FlareSolverr → Radarr/Sonarr/Lidarr/Readarr.
  • Libraries & metadata (Calibre, LazyLibrarian, Calibre-Web, AutoM4B).
  • Plex for remote playback; Kodi DB on the TV.
  • Compliance: personal-use copies of media I own; no redistribution; I comply with French and local laws.

Home automation & sensing

  • Home Assistant + Node-RED scenarios; ESPHome devices.
  • Frigate + go2rtc + Coral TPU for CCTV.
  • ebusd for heating; MQTT as the glue.
  • Speedtest Tracker; push alerts via Signal.
  • Mycodo + InfluxDB for the mushroom lab (sensors, graphs, alarms).

AI, dev & specialty stacks

  • Ollama + Open WebUI: local LLMs, private by default.
  • code-server: edit configs safely from a browser.
  • Manyfold: asset manager for 3D models; Foundry VTT for tabletop RPG.

How to operate it?

~1h / week

Operate it in ~1h / week

  • Upgrades (per stack): docker compose pull && docker compose up -d (pinned tags).
  • Health: Grafana alerts; a quick glance when I'm at my desk.
  • Front-door sanity: test 1–2 Traefik routes; Keycloak MFA; CrowdSec anomalies.
  • Storage & quotas: Nextcloud usage / pool capacity panels.
  • ZFS snapshots & sends are automated by cron.

Underlay: hardware, storage, orchestration

  • Main host (~10 years old): 64 cores, 256 GB RAM.
  • 12 × 12 TB HDD + 8 × 1 TB SSD; Unraid + ZFS (SSD cache → HDD).
  • Docker everywhere; Portainer templates & visibility.
  • Secondary host (older 4 TB disks) as ZFS replication target.

Full autonomy* — the explicit exception

  • I self-host ~200 services for my digital life… except email.
  • Solo email is hard: deliverability & IP reputation, DKIM/SPF/DMARC, blocklists, inbound spam filtering, rDNS, 24/7 uptime.
  • Pragmatic setup: own the domain; outsource mailboxes; optionally mirror IMAP locally; use a trusted SMTP relay for notifications.

The goal isn't No-Cloud; it's Know-Cloud.

Happy to share templates, compose files, and the weekly runbook.

Questions?