VPN all the everythings? (all hosts, VMs etc.)

cracauer@

Developer
How do you deal with an increasing zoo of physical machines in random physical ethernets, VMs on many hosts and in many different network configs (NATed, bridged etc)? It's getting very annoying to remember which jumphost to use for what. And that only covers ssh, some things provide HTTP(S) services.

Once upon a time I thought I would properly IP route everything and use a routing demon to keep things current. Not just to get to the machines but also to up-clue myself about routing demons.

But with the ease of Wireguard now (screw OpenVPN and setting up a certification authority) I wonder. If I make a /etc/hosts or ~/.ssh/config entry for anything should I just hook that entity into the VPN, unconditionally?

The major catch I can think about is that "minor" devices such as routers, TV appliances, printers (shudder) are having between no and inconvenient VPN capability (does an Apple TV do wireguard?). And performance would be impacted, some of my CPUs do not fill 10 Gb/s with encryption.

Is anybody of you dealing with this below the level of having an actual networking group at work?
 
What does your many random physical networks look like?

I run Wireguard on OPNSense and have multiple devices connect to it, including a remote office network. All devices, including phones and laptops can connect and access the DNS on this VPN so no having to edit hosts files, just update DNS (and using DNSSEC for funsies).

I think it's silly to set up the VPN on each VM or jail. Setup proper routing.
 
Run Wireguard on one host per physical location and route traffic through that host.

What about VMs splattered across other hosts in the same physical location that are not in bridge mode?

I get the principal argument. Get your address scheme and routing sorted. That's what IP is for.
 
 
Back
Top