On my router
post
On my opnsense router
On my (OpenWrt) router, configured using the OpenWrt interface
I have a vps (hetzner dedicated server auction) as well as my home servers. The vps has a fixed IP so ive setup wireguard endpoints to all point to it with forwarding on so can access every device indirectly through the vps. It allows them to work across DDNS or remotely.
I used this guide (https://www.digitalocean.com/community/tutorials/how-to-set-up-wireguard-on-ubuntu-20-04). Tried different tools gui's and other methods but always came back to this to work the best
Always in the router if it supports it. If it does not support wireguard I would rather (if you are able and allowed to) replace the router instead of using something else.
I have a Raspberry Pi that runs pihole and Wireguard exclusively. My home server is a Kubernetes cluster running on an old desktop PC and 2 Intel NUCs.
The reason for the separate Pi was essentially because I only had the desktop PC initially, and for a while I had a faulty CPU, making the desktop PC crash or become unresponsive, so it helped a lot having DNS and VPN access separated from the instability.
OPNsense
I run one on my firewall, but it's IPv6 only because of CGNAT. The other one is running on a VPS in case I need IPv4 access. I just configured them manually.
Wireguard normally runs with higher than root privileges as part of the kernel, outside of any container namespaces. If you're running some sort of Wireguard administration service you might be able to restrict its capabilities, but that isn't Wireguard. Most of my devices are running Wireguard managed by tailscaled running as root, and some are running additional, fixed Wireguard tunnels without a persistent management service.
I run a wireguard container on my old desktop server and wgeasy in a pi2 as backup.
One end is a local VPS with insanely good peering pretty much round the damn world, other end is my opnsense router. I actually pass a block of ipv6 through the vpn and my router hands it out to devices which is a nice little bonus
Who is your VPS provider, if you don't mind telling?
https://spartanhost.org/ owner is super chill will make custom spec deployments and they actually have a really nice management panels with nice easy custom iso support
I don't get the mesh if everything is behind your router or firewall what is the point.
Im mostly using a self hosted headscale on a remote vps and then tailscale on my clients.
Having the coordination server outside of my network helps quite a bit and things still communicate over the local lan when possible.
For just wireguard itself, I do have a few site to site connections set up at the router level (opnsense).
Started with it on a server but moved it to my Openwrt router. If the router's up the tunnel's up.
One instance runs on the router (Unifi USG) and the other on a Pi3 (as a backup) using PiVPN.
Usually, if I need to set it up, I’ll use PiVPN and either a Pi or Debian/Ubuntu host.
If your router is down how do you get to your pi backup?
On the home server on the host. I couldn't figure out how to make it work in a container and still have ssh access to the host, which was my goal...
Why would you run a WG Client and WG Server on the same host? Am I reading that second mark wrong?
You are, second point means running WG on say, a proxmox root, and using it to access the containers.
Uhhhh...that is...not how you do that. Especially if you're describing routing out from a container to an edge device and back into your host machine instead of using bridged network or another virtual router on the host.
Like if you absolutely had to have a segmented network between hosts a la datacenter/cloud, you'd still create a virtual fabric or SDLAN/WAN to connect them, and that's like going WAY out of your way.
Wireguard for this purpose makes even less sense.
There's no such thing as a client or server with Wireguard. All systems with Wireguard installed are "nodes". Wireguard is peer-to-peer, not client-server.
You can configure nftables rules to route through a particular node, but that doesn't really make it a server. You could configure all nodes to allow routing traffic through them if you wanted to.
If you run Wireguard on every device, you can configure a mesh VPN, where every device can directly reach any other device, without needing to route through an intermediary node. This is essentially what Tailscale does.
Uhhh, nooooo. Why are all these new kids all in these threads saying this crazy uninformed stuff lately? 🤣
https://www.wireguard.com/protocol/ https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/10/html/configuring_and_managing_networking/setting-up-a-wireguard-vpn
And, in fact, for those of us that have been doing this a long time, anything with a control point or protocol always refers to said control point as the server in a PTP connection sense.
In this case, a centralized VPN routing node that connects like a Hub and Spoke is the server. Everything else is a client of that server because they can't independently do much else in this configuration.
Both of those documents agree with me? RedHat are just using the terms "client" and "server" to make it easier for people to understand, but they explicitly say that all hosts are "peers".
Note that all hosts that participate in a WireGuard VPN are peers. This documentation uses the terms client to describe hosts that establish a connection and server to describe the host with the fixed hostname or IP address that the clients connect to and, optionally, route all traffic through this server.
--
Everything else is a client of that server because they can't independently do much else in this configuration.
All you need to do is add an extra peer to the WireGuard config on any one of the "clients", and it's no longer just a client, and can connect directly to that peer without using the "server".
They do no such thing.
The first link explains the protocol.
The second explains WHY one would refer to client and server with regards to Wireguard.
My point ties both together to explain why people would use client and server with regards to the protocol itself, and a common configuration where this would be necessary for clarification. Ties both of them together, and makes my point from my original comment, which also refers to OP's comment.
I'm not digging you, just illustrating a correction so you're not running around misinformed.
It wasn't clear where OP was trying to make a point, just that the same host would be running running Wireguard for some reason, which one would assume means virtualization of some sort, meaning the host machine is the primary hub/server.
top 41 comments