This post walks through the config as it currently runs. Not a tutorial. Not a clean-slate build guide. Just what’s actually on the box, what each piece is doing, and which of the “this looks clever” choices turned out to be worth it.

What OPNsense replaced

The previous edge was a consumer router running vendor firmware. It worked. It also had opinions about DNS, UPnP, and what “internal” meant that didn’t line up with mine. The moment I wanted per-VLAN firewall rules and per-host DNS overrides, I’d hit a wall.

OPNsense runs on a small x86 box with two NICs. One is WAN, one is a trunk into the switch carrying tagged VLANs. The whole thing idles under 15% CPU with everything I’ve thrown at it.

Interfaces

Three interfaces that matter, plus the ones that don’t.

• WAN — the public side. spoofmac is set, blockpriv and blockbogons are both on. Anything showing up from RFC1918 or reserved space on the WAN gets dropped before it hits a single rule. This is a default, but it’s a default I’ve seen disabled “temporarily” on other people’s firewalls more times than I’d like.
• LAN (192.168.1.0/24) — the trusted network. Hosts, servers, my workstation, the things I wouldn’t want on a segment with a cheap IP camera.
• VLAN20_IOT (192.168.20.0/24) — everything that speaks a protocol I didn’t write and can’t audit. Cameras, smart switches, the TP-Link stuff, a Ring doorbell, an Apple TV. Anything that phones home on a schedule.
• VLAN30_GUEST (192.168.30.0/24) — guest Wi-Fi. Scoped tight and deliberately boring.

The WireGuard interface exists as a stub. The actual WireGuard endpoint runs on Boreas (the Raspberry Pi I wrote about previously) and terminates on LAN. Port 51820/udp is port-forwarded through OPNsense to it. The virtual interface here is reserved for a future on-box tunnel I haven’t needed yet.

Inbound: what’s actually exposed

A small, deliberate list. If something isn’t on this list, the outside internet can’t see it.

• 32400/tcp → Alecto (Plex). Plex’s own account server handles connection brokering; this is the fallback direct path.
• 25/tcp, 993/tcp → Alecto (SMTP inbound, IMAP-over-TLS). Yes, a residential SMTP listener is a lightning rod. I accept the noise in exchange for owning my own mail flow.
• 9001/tcp+udp → the Tor middle relay host. Advertised to the Tor directory as the relay’s ORPort.
• 443/tcp, 51820/udp → Boreas. 443 is Nginx Proxy Manager fronting Overseerr for friends and family behind Cloudflare. 51820/udp is the WireGuard handshake port for remote access.

Port 80 exists as a disabled rule. I leave disabled-but-documented rules in the config on purpose; they’re a paper trail for “this used to be exposed, here’s why it isn’t anymore.” Deleting them loses that context.

Nothing else is forwarded. No UPnP. No NAT-PMP. If a device wants to be reachable from outside, I add the rule by hand.

VLAN policy

The LAN rule is broad: trusted hosts can reach anything. The two VLANs get narrower treatment.

IoT VLAN (VLAN20)

• Can reach the router for DNS (53) and NTP (123). Nothing else on LAN.
• Can reach WAN on 80/443. That’s it.
• Blocked from every other internal subnet.

Guest VLAN (VLAN30)

• Can reach WAN on 80/443. Nothing else.
• Blocked from LAN, IoT, and the router’s management ports.
• Uses its own DHCP scope (192.168.30.50–200) with the router itself as DNS.

Guest users regularly ask why their work VPN client doesn’t connect. That’s not a bug. Guest is guest; if somebody needs a real tunnel, they’re not on guest.

DNS is the interesting part

I run Dnsmasq (not Unbound) across all three interfaces. Domain is home.lan. Upstream is Cloudflare 1.1.1.1 and 1.0.0.1. DNSSEC is off, which I’ll come back to. Authoritative DHCP is on, so DHCP leases produce DNS records for every host automatically.

The interesting part isn’t the resolver. It’s the firewall rules around it.

Forcing IoT through the router

Most cheap smart devices ship hardcoded with Google DNS (8.8.8.8) or similar. They don’t care what DHCP tells them; they use whatever the vendor baked in.

So there’s a NAT redirect on the IoT VLAN that rewrites any outbound connection to *:53 and sends it back to 192.168.1.1:53. The device thinks it’s talking to Google. It’s actually talking to OPNsense. From the device’s point of view, nothing changes. From my point of view, every DNS lookup for every IoT device is now logged locally and subject to the same filters as everything else.

The alias that drives this (TP_LINK) started as one device and grew organically. It now covers the usual IoT suspects including the NVR.

Killing the DNS escape hatches

DNS hijacking works until a device decides it would rather use DNS-over-HTTPS or DNS-over-TLS to bypass your resolver entirely. Modern OSes will happily do this without asking.

Three rules cover the escape routes:

• block-dns-tls-853 — drops TCP/853 outbound (DoT).
• block-doq-quic-853 — drops UDP/853 outbound (DoQ).
• block-doh — drops outbound to an alias called doh_providers.

doh_providers is a dynamic URL table alias. It pulls from the awesome-dnscrypt DoH server list on a 30-minute refresh. The alias updates itself; I don’t maintain it.

This isn’t perfect. DoH hiding inside a general HTTPS flow on port 443 is indistinguishable from any other HTTPS traffic without deep inspection, and I’m not terminating TLS at the edge. What the blocklist catches is lazy implementations that connect directly to a known public DoH provider’s IP, which is most of them.

The goal isn’t to stop a determined adversary. It’s to stop my smart TV from quietly ignoring my DNS server because its firmware author thought that was a reasonable default.

Aliases that earn their keep

Most aliases are boring host groups. A few are worth calling out.

• doh_providers — the dynamic DoH blocklist described above.
• crowdsec_blocklists / crowdsec6_blocklists — external aliases driven by CrowdSec. These hold the IPs that CrowdSec has decided are hostile. Any firewall rule referencing these aliases gets the current set of bad actors without me maintaining a blocklist manually.
• CLOUDFLARE_IPS — Cloudflare’s edge ranges. Used to scope inbound rules so that services meant to sit behind Cloudflare only accept connections from Cloudflare’s IP space. Anything hitting those services directly at my IP gets dropped.
• TP_LINK — the IoT forcing-function alias. Named after what started it. Now overloaded with everything in the “must be hijacked for DNS” bucket.
• TOR_RELAY — single-host alias pointing at the Tor machine. Useful for bundling rules even when the alias has one entry, because when the hostname inevitably moves, I change it in one place.

CrowdSec is doing the heavy lifting

The earlier home lab post mentioned Suricata. That’s no longer what’s on the box.

OPNsense now runs CrowdSec as the main threat-intel layer:

• Agent enabled — parses local logs and identifies scanning, brute-force, and common exploit patterns.
• LAPI enabled — the local API that the bouncer talks to.
• Firewall bouncer enabled — actively drops traffic from IPs flagged by the agent or by the CrowdSec community feeds.
• Rules enabled — the community scenarios load and run automatically.

CrowdSec is cheaper on CPU than Suricata was, and its blocklist model fits the home lab better than a full IDS pipeline. Suricata gave me floods of alerts I wasn’t going to investigate. CrowdSec gives me “this IP has been a problem elsewhere, dropped at the firewall.” Less signal, but the signal it produces is directly actionable because the action already happened.

I lost some visibility moving away from Suricata — notably per-flow inspection rules for specific CVEs. That trade-off has been fine for this environment. If this were an enterprise edge I’d run both.

Observability

A firewall with no telemetry is a box with opinions you can’t verify.

• NetFlow v9 is enabled on LAN and WAN, with WAN set to egress-only. Export target is a local collector at 127.0.0.1:2056 that ships flow records into the metrics stack on Alecto. Active timeout 30 min, inactive 15 sec.
• vnstat runs against the WAN interface for long-term bandwidth counters. It’s the “how much did I actually use this month” graph, not a security tool.
• Zabbix agent is enabled, reporting to Alecto. CPU, memory, interface counters, states table, per-rule hits. The Zabbix server decides what to alert on; the firewall just emits data.

The alerting philosophy is the same as the one in the honeypot post. The firewall is not the thing that should decide what’s worth waking me up over. Its job is to produce honest signal. Somebody else’s job is to filter it.

Hardening that isn’t glamorous

None of this is interesting to read. All of it matters more than the DoH blocklists.

• Web GUI is HTTPS-only. HSTS is on. The cert is self-signed because the GUI is only reachable from LAN and I’d rather pin the cert than introduce an ACME dependency for a single internal endpoint.
• Console menu is disabled. Physical access to the box doesn’t immediately give you the OPNsense menu. It gives you a login prompt.
• NAT reflection is off — fewer surprises with internal traffic taking unexpected paths.
• SSH is bound to LAN only, group-restricted to admins. It’s one of the few places I’m still running password auth with root allowed. Not because I think it’s fine, but because the exposure surface (LAN-only) makes it low-risk and a key-only migration is on the list below.
• Bogon updates run monthly. Reserved/unallocated prefixes change less often than people think, but they do change.
• Backups go to Google Drive via os-gdrive-backup on a schedule. The XML config is small enough that a month of daily snapshots costs nothing and saves hours if I ever have to rebuild.

Gateway monitoring

Single WAN, no failover. The gateway monitor pings 1.1.1.1 continuously with alert thresholds at 30% packet loss (low) and 35% (high), and 300ms latency (low). If the line is bad enough to notice, OPNsense will flag the gateway before I do.

No multi-WAN yet. The second ISP option in my area isn’t worth the monthly cost for the uptime I actually need. If that changes, gateway groups are already half-configured.

What I’d change

A few things are genuinely outstanding, not just aspirational.

• Turn on DNSSEC. Dnsmasq supports it; I disabled it years ago when a misbehaving internal service was breaking on stale signatures. That reason is long gone.
• Move SSH to key-only, disable root login. The exposure is LAN-only today but the migration is trivial and overdue.
• Ship syslog off-box. Right now logs live on OPNsense and get rolled locally. A remote syslog destination (somewhere on LAN, not off-site) would survive a firewall rebuild and give me history during incident response.
• Terminate TLS on a reverse proxy for the services that sit behind Cloudflare. Right now the tunnel ends on Boreas with NPM, which is fine. Moving cert management to an ACME plugin on OPNsense for the few public services would remove a moving part.
• Split LAN. The LAN rule is still “trusted hosts can reach everything.” That worked when LAN was five devices. It’s less defensible now. A management VLAN separate from a workstation VLAN is the next logical step.

Closing

The interesting parts of this config aren’t the exotic features. They’re the boring ones: a DNS policy that assumes devices will lie, per-VLAN egress rules that default to deny, inbound exposure that fits on a single screen, and telemetry that lets somebody else decide what’s worth acting on.

OPNsense makes all of that reachable from a UI without punishing you for running CLI alongside it. That’s a rarer combination than it sounds.

If you’re standing one up from scratch and wondering where to start: segment the network, hijack DNS on the untrusted segments, block the DNS escape hatches, and wire up telemetry before you wire up rules. Everything else is refinement on top of that.

I’ve always enjoyed tinkering with operating systems and finding ways they improve day-to-day life. I’m not a cloud hater. Cloud services are useful and I still use them. I self-host because it’s fun.

With most SaaS tools, you’re limited by design choices you had no part in. My biggest self-hosted system is a Plex machine. I watch what I want, how I want, for roughly the cost of electricity. There’s also been a serious learning component: networking, security, general IT practice. That alone has made it worth running.

Topology

Starting at the internet edge and working inward:

Router / Firewall I settled on OPNsense. It met and exceeded what I needed. The box runs intrusion detection, Unbound DNS, and a handful of other security-focused services.

Switching Traffic hits a 24-port unmanaged gigabit switch with SFP ports. Nothing exotic, but most ports are in use.

Flat Network The network is currently flat, so traffic flows directly to access points, servers, Raspberry Pis, NVR systems, gaming consoles, and everything else.

The topology is simple. The interesting part is what the devices are doing, not how complex the diagram looks.

Core Infrastructure

Alecto

Hardware

• Ryzen 7 1700X
• 1 TB NVMe
• 4 TB HDD
• GTX 1050 Ti

Nothing exotic, but it handles everything I need with around 85% idle time.

Software Ubuntu LTS as the host OS, Docker for everything else: media acquisition, media consumption, networking, local services, metrics, and automation.

Docker makes backing up and restoring critical services significantly easier, which is the main reason I keep everything containerised.

Services

Media Acquisition

The pipeline follows a simple request, acquire, process, library chain.

• Overseerr
• Prowlarr
• Sonarr / Radarr
• qBittorrent
• Deluge
• Unpackerr
• cross-seed

Prowlarr manages indexers. Private trackers have far less fake or malicious content than public ones, which matters later in the chain.

Sonarr and Radarr handle TV and movies. Quality profiles are simple: HD and 4K. That has covered everything so far. Both monitor RSS feeds from configured indexers and push matched torrents to the downloader automatically.

I run two download clients. qBittorrent handles the entire arr stack. Deluge handles manual downloads and non-media content. Dynamic save paths split movies and TV cleanly for Plex.

Unpackerr handles automatic extraction for downloads that arrive as archives. cross-seed finds identical or near-identical torrents across trackers and advertises that I already have the data, which improves speeds and availability for others.

The only recurring issue is Sonarr or Radarr occasionally grabbing a fake title. Aggressive regex-based filters have mostly resolved it.

Media Consumption

• Plex
• Tautulli
• Overseerr
• Homepage

Plex is the primary player. Mature, stable, available on every device, and accessible for non-technical users. I’ve tested Jellyfin and like it, but haven’t switched.

Tautulli gives visibility into Plex usage: playback activity, per-user bandwidth, transcoding load. That data makes decisions around limits and capacity easier.

Overseerr lets users request titles themselves rather than messaging me. Requests still require approval, but that takes seconds instead of a back-and-forth conversation.

Homepage is a single customisable dashboard with a high-level view of everything running. It doesn’t replace Zabbix or Grafana for monitoring, but it’s useful for day-to-day glancing.

Networking and VPN Containment

Torrent clients are routed through Gluetun, a dedicated VPN container running WireGuard. The downloaders have never touched my LAN directly and never see my public IP.

Gluetun runs in strict kill-switch mode. If the VPN drops, traffic stops. There’s no fallback to my home connection. Given the provider’s SLA, this hasn’t been an issue in practice.

No inbound ports need to be open, which reduces exposure further. Speed degradation from the VPN hasn’t been noticeable.

Observability

Prometheus, Node Exporter, cAdvisor, and Grafana cover system-level metrics: CPU load, memory usage, container behaviour. Critical alerts go to Telegram. I’m refining thresholds so only actionable issues send a notification.

Automation

Watchtower handles container updates on a schedule at 03:00. If an update breaks something, rolling back means redeploying from the same configuration paths. Docker’s stateless container model makes that straightforward.

Portainer handles anything that needs a UI.

Network Edge

OPNsense

OPNsense sits between the internet and all internal systems. UPnP is disabled. No device exposes itself automatically. Only explicitly required services are permitted outbound or inbound.

All traffic is statefully inspected. DNS is forced through Unbound. Suricata monitors inbound and outbound traffic for known malicious patterns. Devices can’t quietly phone home, bypass DNS filtering, or accept unsolicited inbound connections without generating an alert.

Services get published deliberately, not accidentally exposed.

Boreas

A Raspberry Pi running Nginx Proxy Manager and WireGuard. This started as a workaround for a previous router that lacked VPN support. After moving to OPNsense, the separation made enough sense to keep.

Boreas is my remote access point back into the LAN. Nginx Proxy Manager exposes Overseerr to friends and family outside the local network. Both services sit behind Cloudflare, primarily to obscure my real IP.

The throughput on the Pi is better than you’d expect for the hardware.

Chronos

A Tor middle relay running as a small contribution to online privacy. No exit node: the ISP complaints and CAPTCHA overhead aren’t worth it. A middle relay provides value without the operational noise. It’s low-maintenance and largely invisible once configured.

Failures and Lessons

In the past year, two things actually broke:

• Disks filled up from log spam. Entirely my fault. Log rotation is properly configured now.
• Incorrect or fake titles downloaded. Better filters and denied extension lists resolved most of it.

Beyond that, issues have been minor misconfigurations and occasional reboots.

What I’d Do Differently

I’d spread services across more hosts if I could go back. Some hardening measures were probably over-engineered, but I don’t regret that trade-off. Deploying the arr stack earlier would have saved time.

What’s Next

Hardware: managed switch, better access points, a more capable GPU for transcoding.

Monitoring: a consolidated Zabbix dashboard with proper alerting.

Networking: VLANs.

Self-hosted AI models aren’t on the list. Cloud tools cover what I need without the overhead.

Closing

Running this lab has mostly taught me patience. Getting multiple devices, containers, and services working together takes iteration. It’s improved my understanding of networking and containerised systems more than any course I’ve taken.

I keep running it because it’s fun and I learn from it. That’s enough reason.