I need help figuring out where I am going wrong or being an idiot, if people could point out where...
I have a server running Debian 12 and various docker images (Jellyfin, Home Assistant, etc...) controlled by portainer.
A consumer router assigns static Ip addresses by MAC address. The router lets me define the IP address of a primary/secondary DNS. The router registers itself with DynDNS.
I want to make this remotely accessible.
From what I have read I need to setup a reverse proxy, I have tried to follow various guides to give my server a cert for the reverse proxy but it always fails.
I figure the server needs the dyndns address to point at it but I the scripts pick up the internal IP.
The easiest and quickest way thats still safe is to just use tailscale.
Its a zero config VPN that you can install on all your devices.
I've been using it for quite some time now and I'm still fascinated by how easy to use it is.
I would want to go that appros but it feels very inconvenient having to connect to VPN every time I want to check something, also the battery drain if I stayed connected all the time
I’ve been using Tailscale for about 2 months now. It has a VPN-on-demand setting that I keep enabled. That way, anytime I am not on my local WiFi, it automatically connects the VPN. According to my battery health settings, Tailscale has used 5% of my battery in the last 10 days. And I am even using a Mullvad exit node, which would use even more battery.
Not sure is it same, I don't use tailscale, but using pure wireguard. In my experience battery drain is not even noticable, but staying connected is not smooth as I'd like. I tried to keep active VPN all the time, but then sometimes I just notice my internet is not working ( I have disable or restart VPN connection). It could be issue with my phone (Android), missconfig or something else, but I switched to manually enabling VPN every time I need it. Not amazing, but few clicks every now and then is more than acceptable for my use case
It's easiest to just register a domain name and use Couldflare Tunnels. No need to worry about dynamic DNS, port forwarding etc. Plus, you have the security advantages of DDoS protection and firewall (WAF). Finally, you get portability - you can change your ISP, router or even move your entire lab into the cloud if you wanted to, and you won't need to change a single thing.
I have a lab set up on my mini PC that I often take to work with me, and it works the same regardless of whether it's going thru my work's restricted proxy or the NAT at home. Zero config required on the network side.
Just a reminder that even though the tunnel itself is encrypted, the whole connection is not E2E encrypted between your remote client and the server. Cloudflare as a CDN/PoP provider can see the traffic in plaintext.
In all other aspects, this is a great solution, as we even get to use the edge caching(over top of all others mentioned above) facility - which further reduces the requests to origin server.
I recently went this route after dabbling with other options. I had a wireguard VPN through my Unifi router, with rules to limit access to only the resources I wanted to share, but it can be a struggle for non savvy users, and even more so if they want to use Jellyfin on their TV. Tried Twingate too and would recommend if it fits your usecase, but Cloudflare Tunnels were more applicable to me.
By "remotely accessible", do you mean remotely accessible to everyone or just you? If it's just you, then you don't need to setup a reverse proxy. You can use your router as a vpn gateway (assuming you have a static ip address) or you can use tailscale or zerotier.
If you want to make your services remotely accessible to everyone without using a vpn, then you'll need to expose them to the world somehow. How to do that depends on whether you have a static ip address, or behind a CGNAT. If you have a static ip, you can route port 80 and 443 to your load balancer (e.g. nginx proxy manager), which works best if you have your own domain name so you can map each service to their own subdomain in the load balancer. If you're behind a GCNAT, you're going to need an external server/vps to route traffics to its port 80 and 443 into your home network, essentially granting you a static ip address.
You don't need a static IP to host a VPN. You can do it using a dynamic DNS which updates the DNS records to match your IP when/if it changes. You do need a public IP though, so CGNAT goes straight out.
Im using wireguard VPN. You have to setup VPN server (using your DynDNS address, but duckdns in my case), open wireguard port in your router and configure each device that needs access. Reverse proxy is not needed, but I have it so I can use jellyfin.example.com instead of 192.168.100.40:8096. I use NPM (nginx proxy manager) with awesome GUI that can create lets encrypt certificates. I also use piHole for local DNS server
If you are the only one using the services, then go for a VPN instead of port forwarding or sth. This way, your stuff isn't openly accessible from the internet to anyone poking around
I agree with this, protecting everything behind a VPN is the way to go. I help friends setup their vpn client to my stuff if I want them to access an internal service.
If you are going for a reverse proxy, I highly recommend using Caddy. Issuing TLS certificates is all done automatically and reverse proxy headers are all automatically set.
My advice is to just use Tailscale. It’s a 5 minute setup and you get access to your stuff from anywhere, securely, without opening ports to the public internet. It will give your server a second IP address, which you will be able to access from any other device which is also registered to your Tailscale account.
My personal setup:
Tailscale installed on all devices that need access to my home lab
Custom domain with root A record set to server’s Tailscale IP
caddyserver reverse proxy on server, with DNS https authentication configured (regular http with won’t work because it’s not on the public internet)
I have a wireguard tunnel set up between my home server and the VPS, with persistent keepalive. The public domain name points to the VPS, then I have it set up (simply using iptables) so that any traffic there in port 80 and 443 is sent back to my honeserver and there it’s handled by caddy, and sent to the actual service.
The only ports I need to open are 80 and 443 on my VPS to make this setup work. So, no open ports on my local machine. This does however require you to pay for VPS. Since you aren't doing much on it though, you can get away with a cheap one. I have a $12/year VPS from Rack nerd that I use for this job.
For completely free options, you can do one of three things. (That I can think of. There are probably more ways.)
Either open up some ports on your machine. You'll need to make sure that you aren't behind a CGNat for this. I simply don't like opening ports to the internet, though.
You can use a VPN. Tailscale works great for this. I use it personally for sshing remotely into my machines.
You can use cloudflare-tunnels. Potentially bad privacy-wise since they can technically access the data. So don't use it for sensitive stuff. Also, their policy doesn't allow traffic that's not mostly HTML. So something like a Jellyfin server would violate this. But you do get to use their firewall which is great for protection against DDOS attacks.
P.S. If you need help setting any of these up, lmk.
Your setup sounds great! I hadn’t come across something like that and I’d love to try it out, myself. Do you have a guide or any other resources with more info? I’m currently using a reverse proxy, but I’m not excited about the open ports, even with firewall rules keeping them contained.
I'm afraid that I don't have any guides. But, you're halfway there anyway. Which one of these methods do you prefer? I can maybe give you some pointers.
I am using duckdns.org and let my router ping it when it's public IP changes. Then I use nginx as a reverse proxy with help of https://nginxproxymanager.com/ so I don't need to write config files and it also runs certbot for my so I don't need to deal with https manually.
Actually I also have my own domain so I use a subdomain pointing via CNAME to the duckdns subdomain. This way I can easily change the provider of dyndns.
I do it the simple way. I just stick nginx in front of everything. If I don't want it to be publicly accessible I stick nginx basic auth in front of it.
The advantages is that I can easily access the services from anywhere on any device with just the password. I only need to trust nginx's basic auth to keep me protected, not various different service's authentication.
The downside is that some services don't work great when you have basic auth in the front. This is often due to things like public links or APIs that need to be accessed with other auth.
I just use nginx because I've always used it. I've heard that there are newer reverse proxies that are a bit easier to configure.
I do that, but only allow access to private services from local IP addresses, rather than putting auth in front of them. Then I use IPsec to access my local-only things.
It depends on how much you trust nginx. A HTTP server is probably a bit more complex that your average VPN solution so probably more likely to have vulnerabilities, but it is also the most popular web server on the planet, so if there is a zero day I'm probably not the first target. If you stay up to date you are probably fine.
Headscale is pretty straight forward to set up and easy to use. And there are multiple WebGUIs available to choose from, if you need. If you have any questions, let me know.
I use nginx as a reverse proxy and assign each service either a subdomain or a specific url. SSL is configured once so all services get https. Its not the best though, some services don't like being behind a reverse proxy or don't play nice with the url, subdomain management can get cumbersome and if the service doesn't have a login page, it is open to bad actors.. i was thinking of making a website with login and exposing other web services through an iframe but i don't know how viable that may be.
A vpn would probably be the best way to go from a security standpoint but accessing services may be a pain on remote devices where a vpn isn't supported - like how would a TV on a remote network access tour jellyfin server if the service is only accessible through a vpn tunnel and the tv has no way of connecting to it? Not sure.
If it's only you (or your household) that is accessing the services then something like hosting a tailscale VPN is a relatively user friendly and safe way to set-up remote access.
If not, then you'd probably want to either use the aforementioned Cloudflare tunnels, or set up a reverse proxy container (nginx proxy manager is quite nice for this as it also handles certs and stuff for you). Then port forward ports 80 and 443 to the server (or container if you give it a separate IP). This can be done in your router.
In terms of domain set-up. I've always found subdomains (homeassistant.domain.com) to be way less of a hassle compared to directories (domain.com/homeassistant) since the latter may need additional config on the application end.
Get a cheap domain at like Cloudflare and use CNAME records that point domain.com and *.domain.com to your dyndns host.
Iirc there's also some routers/containers that can do ddns with Cloudflare directly, so that might be worth a quick check too.
Mhh I don't know if I can help you too much.
I initially followed spaceinvader ones tutorials for my unraid machine. But with time I changed from swag to nginx proxy manager. And I changed from using a duckdns docker to a router based dyndns tracker.
But honestly I don't remember too much from the process I currently try to switch domain but just can't get them to work :D so I am in a smilar spot like you.
I assume you want to access a self hosted service on your local server from the Internet.
To make the service accessible from the Internet multiple things are required:
the router can be accessed from the outside. Find your public IP in the router or use a find-my-ip website. Better: do both.
This is the address you can use to access your router (or whatever service you choose to expose through it).
Side note: If the Ip-adresses of your router and the one of the find- my- ip- site are different it could mean that your provider uses CG-NAT (because ipv4- addresses are scarce, the provider doesn't give you a real publicly accessible address). This means you can't access your router from the Internet. Try IPv6 or contact your provider to get a publicly accessible ipv4- address.
because the above mentioned IP- address of your router might change, dyndns is used. Configure it in your router and test it. Test if the DNS- name you have set up resolves to your ip- address (nslookup or ping it).
to make your service available to the Internet you need to configure port forwarding in your router (or add your server as exposed host - means all ports are forwarded to the Internet). This means the router passes request to itself on to your internal server.
Careful: everybody can access whatever services you expose. Advice: it's a good idea to use a VPN. Setup a VPN-server in your Lan and only port-forward its port in the router. Connect to the VPN from the outside - Afterwards use the internal services through the vpn- connection.
scripts and the internal ip: the dyndns name needs to be used instead of the IP. Find a way to make the scripts use that name to resolve it to your external IP.
Wireguard for network access, istio gateway for exposing services, and keycloak for SSO. I want to experiment with Teleport for more fine grained access to my services.
If I had more exposed services I would mess with crowdsec for some another firewall rule set and maybe even exposing it through a TOR service proxy.
Another response for "VPN". I don't have any of my self-hosted services exposed publicly & use WireGuard with the on-demand settings so that whenever I'm not at my house, I am automatically connected.
Some users did mention that things like TV's, etc outside your network wouldn't be able to connect, but that hasn't been an issue for me, since I don't use my services like that.
My IPv4 connection uses CGNAT, so I use a VPN to access my server. I also have IPv6, so I have a couple of things directly accessible over it in case the VPN drops for some reason. I do have dynamic DNS set up, although it's not really necessary. My IPv6 prefix doesn't seem to change unless I change the DUID on my firewall.
I have two servers (and different other machines) on two different LANs joined by a wireguard tunnel between the routers, with DynDNS running on both, and wireguard on all mobile devices that need access to the LAN.
If your router can natively run wireguard, I'd highly recommended. It just works. Or just use tailscale, it's wireguard for lazy people