Managed Routes use incorrect interface when used over multiple networks

[theminecoder]

I have 2 different ZeroTier network that use managed routing for access into the physical network remotely. When attempting to connect to one of the network it wouldn't respond to any traffic & would show "no route to host" on ping attempts from my MacBook.

Investigating further it seems that the ZeroTier service (at least on Mac) seems to reuse the interface when registering managed routes, so all the traffic for network A (10.150.1.x) was being sent to the interface for network B (10.0.1.x). This does not reflect what is shown in the UI for the interfaces (see below screenshots).

Further debugging shows that it seems to be whatever network gets connected first that has a managed route seems to take priority and is used in all route registrations.

Screenshots

Network A - UI Info

Network B - UI Info

Route Info - Showing double up of interfaces

[laduke]

Thanks for writing it up. Surprised we haven't seen it before.

Does each network have a 10.242.0.0/16 route defined on it? I would guess this is what triggers the bug. Try changing it to 10.242.0.0/17 or 10.242.0.0/15 on one of the networks to work around the issue.