- cross-posted to:
- selfhost@lemmy.ml
- cross-posted to:
- selfhost@lemmy.ml
From the article:
Since Tailscale was founded in 2019, customers have been forced to choose between either Tailscale or Mullvad without the ability for them to co-exist.
Today we announce a partnership with Tailscale that allows you to use both in conjunction through the Tailscale app. This functionality is not available through the Mullvad VPN app. This partnership allows customers of Tailscale to make use of our WireGuard VPN servers as “exit nodes”. This means that whilst connected to Tailscale, you can access your devices across Tailscale’s mesh network, whilst still connecting outbound through Mullvad VPN WireGuard servers in any location.
Announcement also on Tailscale blog.
If Wireguard wasn’t so easy, and if I was just a little less interested in playing with setting up my own VPN, I would have use Tailscale. It’s what I’ll recommend to family members when they eventually come asking for a VPN.
That said, setting up a VPN with multiple nodes is pretty trivial for anyone with basic networking knowledge. IME it’s also extremely low maintenance; I have far more issues with the containerized apps I’m running occasionally barfing and needing tending than my little 7-node Wireguard VPN. I use Mullvad for the exit nodes and bandwidth (so as not to swamp my little VMs), but Tailscale doesn’t offer me much value.
Still, as I said, for non-technical people, Tailscale is pretty cool.
I’m probably a medium-techincal people :-) Wireguard won’t do the NAT traversal right? I can’t do the port forwarding thing because of the CGNAT for my connection.
It will, as long as you have an exit node outside your NAT. I allow access to my home LAN from our phones through Wireguard, but I hop through a VM in the cloud. All you do is set a keep-alive on the node(s) in your LAN.
I had it set up pretty well with OPNSense as the wireguard gateway into my home and the official wireguard client on my lineage phone and it was working great for a year but something changed recently where it’s become really unreliable. The problem is not OPNSense, but my phone. Not sure if it’s the client or lineageos causing problems.
My WG Android client is so stable that I forget I have it on. I’m not running Lineage, though, so maybe that helps? Why are you sure it’s not OPNSense? Also, there have been rumors of some carriers subtly sabotaging VPN connections; have you eliminated that?
It might be OPNSense, but the problem occurs when I leave my house. My network connectivity dies when it switches from wifi to mobile data, only to recover when I disable then re-enable wireguard. This indicates to me that the android client is not properly updating routes or DNS settings during the network change, or lineage OS is doing something wonky, but I could be wrong.
Interesting. Underlying network changes shouldn’t make a difference to WG; TCIP routing is dynamic.
But you may be onto something about DNS. If, when you switch networks, the OS is overwriting the DNS server information that WG set up, that would do what you’re describing. Restarting WG would re-assert the DNS serves that are configured. The one hitch is that normally this would only cause leakage, not failure to resolve… overwriting WG’s DNS servers with public ones should still work.
Still, it’s a good intuition, and if it were me, that’s where I’d look.
Ok, it’s not DNS. I opened a Termux terminal and tried pinging an IP on my network. No luck. Stopped and restarted the wireguard connection, and was able to ping the machine.
Good idea! Weird. I wonder if one of the networking apps could provide some diagnostics.
If you’re using Termux, you could install the whole suite of common Linux networking tools, like traceroute. I haven’t gone down this route before; I don’t know how far you can go.
I’m also unfamiliar with OPSense, and don’t know how the two apps interact. Good luck!
Thanks!
You know what, I think you are right, it is almost certainly a DNS. I have Adguard setup at home, so I route all my DNA requests through wireguard even though I’m using a split tunnel. That would explain why everything dies, even traffic that shouldn’t be going through the tunnel.
I’ll keep pulling on that thread, thanks for the insight.