Pangolin LAN Test Failed Across Three Devices
I juggle a fascinating world of self-hosted services and secure access, relying on tools like Pangolin for public exposure and Tailscale for private connectivity. It’s an elegant setup, until you start digging into the subtle complexities of network routing.
When I ventured into setting up Headscale as a self-hosted replacement for Tailscale’s cloud control server, the system seemed smooth sailing. However, testing the network traffic between devices on my local home LAN revealed a delightful and frustrating inconsistency.
I ran simple ping tests between these devices, and the results were anything but uniform. Some connections registered incredibly low latency—a quick 5 to 19 milliseconds (direct communication), which is exactly what you’d expect when everything lives on the same network. Other pings, however, jumped to much slower speeds, ranging from 88 to 102 milliseconds. These slower results were clearly the traffic being rerouted through Tailscale’s DERP relay system.
This discrepancy immediately sparked a deep dive into why network traffic was detouring itself, even when all the endpoints were sitting comfortably on the same internal network. It made me wonder about the hidden architecture of VPN and mesh networking—why the detour?
A few days later, while exploring Pangolin’s changelogs, I found a tantalizing line in the 1.21 update: support for same-network detection. This felt like a potential solution to the routing puzzle I was facing. The logical next question arose: does this new feature actually deliver on solving the specific network behavior that Tailscale seemed unable to manage?
The journey into understanding the subtle differences between self-hosted controls and established mesh protocols turned out to be more fascinating than anticipated, proving that even in the world of high-level networking, the details are where the real magic—and the real headaches—reside.
Credit
