Four years in: a two-node Proxmox cluster, six VLANs and 25+ self-hosted services, some of them used daily by people outside this household. Web services are reachable without a single open port; the game and voice servers that do need one are isolated in their own VLAN. Everything below is designed, built and operated by hand.
A two-node Proxmox VE cluster with shared TrueNAS storage, a dedicated backup server, and a Raspberry Pi quorum device to prevent split-brain during single-node failures.
Each workload class lives in its own VLAN with tailored firewall rules. This limits blast radius if any service is compromised and enforces least-privilege networking by default.
local.screedy.com. Forwarder zone for screedy.com — local overrides resolve to Traefik VIP, unknown records forwarded to Cloudflare. Secondary Technitium syncs via zone transfer.25+ self-hosted services across six VLANs, all accessible via Traefik with wildcard TLS certs issued through a Cloudflare DNS challenge — no open WAN ports required for web access.
local.screedy.com, forwarder for screedy.com. Secondary node provides HA for DNS resolution.Defence in depth — from VLAN segmentation at the network layer up through zero-trust tunnels, automated TLS, SIEM alerting, and per-service access controls.
*.local.screedy.com is never published to public DNS — unreachable externally by design*.local.screedy.com and *.screedy.com via Cloudflare DNS challengeserversTransport with appropriate cert verification settingsMetrics, logs, and security events aggregated into a coherent observability pipeline — so problems surface before they become outages.
/metrics on :8080), Unpoller (UniFi network metrics), and itself. Basic auth enforced by Traefik middleware.The decisions that shaped this setup — and the trade-offs considered.
service.screedy.com to the Traefik VIP via Technitium's forwarder zone. External clients get Cloudflare's public record. Same domain, different resolution path — no internal traffic hairpins through Cloudflare, and no UX difference for users.Every technology actively used, managed, or operated in this environment.
Active migrations and planned improvements to move the lab closer to a fully HA, zero-trust production posture.