Greetings! I currently host a number of services on an old pc in my basement. I have ports 80 and 443 forwarded and am running Nginx Proxy Manager as well as Authelia to protect most of them. I have set up a lemmy instance that I am using as my main point of access to the fediverse. I guess I have two questions. I am assuming that hiding lemmy behind Authelia would break Federation (although maybe only one way?), is that correct? And secondly, would it be objectively safer for me to pay for a VPS, run Nginx Proxy Manager there and then forward all of the traffic to the services hosted in my basement server using Tailscale? Thanks!
Not sure about the Lemmy part of the question… but it’ll definitely be safer to run nginx on a VPS and then communicate with your basement box over tailscale. That doesn’t expose your home ip for your services. The other thing you can look into is using a wildcard cert so the specific services you’re ruining aren’t enumerated in your DNS, might not be necessary since you have nginx sitting in front.
Thank you! I have a followup question if you don’t mind. If I host npm on a vps that allows me, say 500GB of data transfer a month, and I am streaming media through a jellyfin instance that I have forwarded through the VPS, will the media I’m streaming count against that transfer amount? This seems like basic information that I should have learned a long time ago.
I think that depends on if the VPS charges incoming and outgoing external connections, some only charge outgoing. From VPS to browser would be outgoing, from your home instance to VPS would be incoming.
Thank you! It does seem like there are afforadable VPS options that don’t limit traffic anyway, so I might as well start there.
Mind sharing which VPs those are?
For sure, I think Oracle has unlimited bandwidth even in it’s free tier. Ionos is one that I was looking at, fairly cheap but also offers unlimited transfer
Sure… i already have an oracle free tier that i use for NPM (though switching it to caddy) … i’d be wary of trusting either of them for long term usage though…
I’ve definitely heard bad things about Oracle deleting people’s machines without warning. Unfortunately I don’t know anything about Ionos. I may go with Linode, their cheapest VPS at $5/month has a 1TB transfer limit, which I might stay under. But please let me know if you end up finding a decent option from a reliable company!
OVH US has unlimited traffic but limits the bandwidth. I think they start at 100mbps and go over 1gbps depending on which configuration you pick.
Like the other poster said, the traffic will go through the VPS. But since you mentioned tailscale, why not just connect to your network over tailscale. You could even use DNS to point to your services (nicer names than IPs) but then the clients would connect directly and you wouldn’t need to route through VPS.
I’ve definitely thought about it, and I would if it were just me. I share some services with family and I don’t want to be responsible for troubleshooting if they accidentally uninstall tailscale or reset their phones
Yeah worth the reduced headache. Good luck!
NPM also looks abandoned right now. There’s some security patches that are not being addressed, and certificate renewal is hit or miss due to the age of NPM’s certbot vs the mainline.
If you’re deploying something new, you might want to consider caddy or nginx by itself or some other reverse proxy at this point since it really looks like the dev has vanished and nobody is taking over maintenance yet.
Are you sure it’s abandoned? The docker page was last updated three days ago and there is activity on the github page. Are they just updating the software without pushing any security fixes? I’ve definitely not had issues renewing certificates
There are nearly 1000 open issues and a couple of them are about potential vulnerabilities where the repeated refrain is ‘we tried to contact the developer, but there’s no response’ which makes me… uncomfortable, especially given that NPM was the gatekeeper to a lot of services on my local network.
The cert error is related to outdated python code in the latest shipping version, https://github.com/NginxProxyManager/nginx-proxy-manager/issues/2912 and 2921.
Again, you CAN work around it but the whole radio silence on issues and ongoing issues just makes me uncomfortable with the project and exceedingly reluctant to continue using it because it’s unclear what’s going on, and why.
Thank you for explaining and for the examples, it’s definitely time to start learning caddy. I liked npm for the simple gui, but security and reliability are more important
the problems mentioned by @[email protected] are what lead me to try Caddy… there’s no fancy gui but it does “just work” once you figure out the caddyfile…