Solved: How can containers be forced to use a VPN?
I'd basically like to run some containers within a VPN and some outside of it. The containers running within the VPN should not be able to send or receive any traffic from outside the VPN (except localhost maybe).
The container could be docker, podman, or even a qemu VM or some other solution if need be.
Is that possible? Dunno if this is the right place to ask.
Gluetun, is overkill if you already have a working setup. Your system is able to handle this in a much simple way with built in tools.
You can use systemd to restrict some daemon to your your VPN IP. For instance here's an example of doing that with transmission: override of the default unit by using the following command:
systemctl edit transmission-daemon.service
Then type what you need to override:
[Service]
IPAddressDeny=any
IPAddressAllow=10.0.0.1 # --> your VPN IP here
Another option, might be to restrict it to a single network interface:
[Service]
RestrictNetworkInterfaces=wg0 # --> your VPN interface
Save the file and run systemctl daemon-reload followed by systemctl restart transmission-daemon.service and it should be applied.
This is a simple and effective solution that doesn't require more stuff.
Do you have a link at hand on how start a process within a specific veth by chance?
Own name spaces are easy enough and a lot of tutorials but I don't want my programs to ever be not in the vpn space, not at startup not as fail over etc.
That's the reason why I stuck with the container setup, only for gluetun plus vpned services.
Processes run in network namespaces (netns), and that's exactly what ip netns exec does.
A newly created netns via ip netns add has no network connectivity at all. Even (private) localhost is down and you have to run ip link set lo up to bring it up.
You use veth pairs to connect a virtual device in a network namespace, with a virtual device in the default namespace (or another namespace with internet connectivity).
You route the VPN server address via the netns veth device and nothing else. Then you run wireguard/OpenVPN inside netns.
Avoid using systemd since it runs in the default netns by default, even if called from a process running in another netns.
The way I do it is:
A script for all the network setup:
ns_con AA
A script to run a process in a netns (basically a wrapper around ip netns exec):
ns_run AA <cmd>
Run a termnal app using 2.
Run a tmux session on a separate socket inside terminal app. e.g.
export DISPLAY=:0 # for X11
export XDG_RUNTIME_DIR=/run/user/1000 # to connect to already running pipewire...
# double check this is running in AA ns
tmux -f -f <alternative_config_file_if_needed> -L NS_AA