Tailscale restart silently cut off k3s node; flannel VXLAN fix in v0.28.11
A k3s node running flannel's VXLAN backend over Tailscale lost connectivity to Pods on other nodes after tailscaled was restarted, while still reporting Ready. Flannel recreated its flannel.1 device, but the kernel had already removed the routes, ARP and FDB entries tied to it, and flannel only rewrites them when a remote lease changes. The node stayed isolated for minutes until k3s-agent was restarted. The root cause is tracked as flannel issue #2617, and a fix replaying known remote leases after device recreation shipped in flannel v0.28.11. Until k3s includes that release, operators are advised to restart k3s after VPN restarts and alert on the flannel.1 route count.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in