Appearance
Nginx Proxy Manager
Nginx Proxy Manager (NPM) provides HTTPS entry points for homelab web services. This deployment runs NPM and OpenResty as native services in an unprivileged Debian LXC, rather than Docker.
First prepare the proxy guest. Private proxy inventory, access policies, captive-portal integration, and AI procedures are in NPM · Internal Operations.
Access and deployment
| Setting | This homelab |
|---|---|
| Guest | LXC 101, npm |
| Address | 10.120.40.101 |
| Domain | npm.home.rwk.me |
| Dashboard | https://npm.home.rwk.me |
| Direct administration | http://10.120.40.101:81 |
| Proxy traffic | TCP 80 and 443 |
| Native units | npm.service, openresty.service |
| Persistent state | /data, /etc/letsencrypt |
For another homelab, replace the addresses, domains, certificates, and network access policy. Keep direct administration available on a trusted management network as a recovery path when a hostname or proxy route fails.
Initialize the application
The guest setup installs the native application through the Proxmox VE Community Scripts installer. Confirm both services are running, open the direct administration URL, and complete the administrator setup required by the installed release. Set a unique password and store it in your password manager. Change any installer-provided initial credentials immediately.
sh
systemctl is-active npm openresty
curl --fail --silent --show-error --output /dev/null http://10.120.40.101:81/Configure DNS
Cloudflare domain and private service names
This lab uses rwk.me, with its authoritative DNS managed by Cloudflare. The home.rwk.me subdomain groups the homelab's internal services, such as npm.home.rwk.me and up.home.rwk.me.
The DNS records direct these names to Nginx Proxy Manager's private address, 10.120.40.101, rather than to the home's public internet address or each application's backend. For example:
text
npm.home.rwk.me → home.rwk.me → 10.120.40.101 → NPM → application backendInternal network access
The *.home.rwk.me hosts are intended for access only through this homelab's network. A client must have a permitted route to the private proxy address, either on an allowed local network or through approved Tailscale remote access. Public DNS may resolve the names from elsewhere, but knowing the hostname does not provide a route into the lab. VLAN firewall rules, proxy access lists, and application authentication still determine which services each client can use.
Configure the records
In Cloudflare, select the rwk.me zone and open DNS → Records. The base record for home points to 10.120.40.101; service names can use CNAME records pointing to home.rwk.me, or a wildcard record can cover the common suffix. For another homelab, substitute your own domain and private proxy address.
Keep these records DNS only (the gray cloud). Cloudflare provides DNS resolution, while clients connect directly to the private NPM address; Cloudflare's HTTP proxy is not the access path for these services. See Cloudflare's proxy-status documentation.
Make the records resolvable from the client networks and monitoring host. This setup does not require WAN port forwarding for the service names. HTTPS certificates can be obtained using DNS validation without publishing the internal web servers to the internet.
Obtain a certificate
In SSL Certificates, request or import a certificate covering the intended names. This lab uses a Let's Encrypt certificate for home.rwk.me and *.home.rwk.me.
A wildcard certificate requires DNS validation: select your DNS provider and enter a token with the permissions needed for the relevant zone. Keep the token and certificate private key in protected application state and your secret manager. Follow the certificate dialog's validation and renewal requirements; do not remove a required DNS credential after issuance without replacing the renewal mechanism.
For another lab, use your own domain and certificate. A wildcard covers one label beneath its suffix; check certificate coverage before assigning it to a new name.
Add a proxy host
Select Hosts → Proxy Hosts → Add Proxy Host and enter:
| Field | What to choose |
|---|---|
| Domain Names | The hostname clients will visit |
| Scheme | The protocol actually served by the backend, HTTP or HTTPS |
| Forward Hostname / IP | The application's internal host address |
| Forward Port | The application's published or native listener port |
| WebSockets | Enable when the application needs a live WebSocket connection |
| Access List | A deliberate policy appropriate to that service |
| SSL certificate | A certificate covering the client-facing hostname |
| Force SSL / HTTP/2 | Enable for normal HTTPS services |
For example, Uptime Kuma uses up.home.rwk.me, HTTP backend 10.120.40.100:3001, and WebSocket support. The NPM dashboard can proxy to its own HTTP backend 10.120.40.101:81 under npm.home.rwk.me.
Check direct backend reachability from the proxy before blaming DNS or TLS. An HTTPS backend with a self-signed certificate needs deliberate upstream trust handling; the client-facing certificate does not validate the upstream by itself. Avoid enabling HSTS subdomains until all affected names can use HTTPS.
Configure access lists and special cases
Create LAN/VPN access lists for private dashboards and assign them to the appropriate proxy hosts. Check which source address the proxy actually sees when a VPN router performs SNAT. A proxy access list applies only to proxied requests; restrict direct backend ports separately when needed.
Some applications require extra settings. LibreSpeed needs upload limits and buffering configured for throughput tests, and RackPad requires explicit proxy trust and overwritten forwarding headers. Follow each service guide instead of applying one Advanced snippet globally.
This lab also proxies a guest captive portal on the shared HTTPS listener. Preserve its global guest-isolation includes before allowing guest-network access. Its full configuration, listener ports, firewall policy, and recovery procedure stay in the private runbook.
Verify and maintain
sh
sudo /usr/local/openresty/nginx/sbin/nginx -t
systemctl is-active npm openresty
curl --fail --silent --show-error --output /dev/null https://npm.home.rwk.me/Verify a representative HTTPS application, browser login, WebSocket behavior where used, and certificate renewal status. To apply reviewed snippet changes, validate first and gracefully reload with sudo /usr/local/openresty/nginx/sbin/nginx -s reload.
An NPM application update can interrupt every proxied site. Take a protected backup of the guest and application state, then run the native installer's updater, sudo update, during maintenance. Review failures before rebooting. Use normal Debian package updates for the guest OS; do not replace this native deployment with Docker commands.
Backups and recovery
Preserve /data and /etc/letsencrypt from the same recovery point, including custom proxy snippets, ownership, and permissions. Protect backups: they contain database credentials, DNS-provider tokens, and private certificates. A tested Proxmox backup of the whole LXC is the preferred recovery path when the application and database are damaged together.
After recovery, verify the direct dashboard, DNS, HTTPS routes, access-list denials, and any captive-portal isolation before returning traffic. Keep executable assets in the homelab repository; use these guides as the setup and operations documentation.
See the NPM documentation and Community Scripts installer for upstream behavior.