Description
In general, it's very difficult for users to learn on their own why direct connections aren't working. We provide tools like coder netcheck and coder ping which seem designed for our engineers than end users. For example, these are the first few lines of coder ping -v:
I ~ coder ping c0 -v
2024-08-12 17:06:44.261 [debu] net.wgengine: [v1] using fake (no-op) tun device
2024-08-12 17:06:44.261 [debu] net.wgengine: [v1] using fake (no-op) OS network configurator
2024-08-12 17:06:44.261 [debu] net.wgengine: [v1] using fake (no-op) DNS configurator
2024-08-12 17:06:44.261 [debu] net.wgengine: dns: using dns.noopManager
2024-08-12 17:06:44.262 [debu] net.wgengine: link state: interfaces.State{defaultRoute=en0 ifs={en0:[10.0.0.21/24]} v4=true v6=false}
2024-08-12 17:06:44.262 [debu] net.wgengine: [v1] couldn't create raw v4 disco listener, using regular listener instead: raw disco listening not supported on this OS
2024-08-12 17:06:44.262 [debu] net.wgengine: [v1] couldn't create raw v6 disco listener, using regular listener instead: raw disco listening not supported on this OS
2024-08-12 17:06:44.262 [debu] net.wgengine: magicsock: disco key = d:fb319b0d612d3eb5
2024-08-12 17:06:44.262 [debu] net.wgengine: Creating WireGuard device...
2024-08-12 17:06:44.264 [debu] net.wgengine: Bringing WireGuard device up...
2024-08-12 17:06:44.264 [debu] net.wgengine: wg: [v2] UDP bind has been updated
2024-08-12 17:06:44.264 [debu] net.wgengine: wg: [v2] Interface state was Down, requested Up, now Up
The result is frustration for our customers that have to bring Coder engineers into support calls. Today I'm experiencing that frustration as my development environment is much slower than normal. I even have the privileged ability to ping @coadler on Slack that our customers lack.
The most common question is "why isn't p2p working" and "what can I do to make p2p work". Can we design our various networking UIs in the product to explain that?
Here's coder ping today:
I ~ coder ping c0
pong from c0 proxied via DERP(Council Bluffs, Iowa) in 73ms
pong from c0 proxied via DERP(Council Bluffs, Iowa) in 73ms
pong from c0 proxied via DERP(Council Bluffs, Iowa) in 76ms
pong from c0 proxied via DERP(Council Bluffs, Iowa) in 73ms
✔ preferred DERP region: 999 (Council Bluffs, Iowa)
✔ sent local data to Coder networking coodinator
✔ received remote agent data from Coder networking coordinator
preferred DERP region: 999 (Council Bluffs, Iowa)
endpoints: 204.16.241.141:46660, 172.17.0.1:46660, 172.20.0.13:46660
✔ Wireguard handshake 10s ago
One could imagine it explaining that P2P doesn't work because
- it's disabled as a deployment setting
- both sides of the connection are hard nat
- ???