I’m running k3s 1.31.4 on Debian Bookworm on several bare metal amd64 nodes. I use the builtin Flannel for networking. On one of the nodes, my route table is getting messed up with a bad default route:

# ip route show
0.0.0.0 dev vetha22f9001 scope link
0.0.0.0 dev vethb44f5cec scope link
0.0.0.0 dev veth7d7b382f scope link
0.0.0.0 dev vethd54d4b62 scope link
0.0.0.0 dev vethe05a10d5 scope link
0.0.0.0 dev veth88dd6cab scope link
default dev vetha22f9001 scope link   <-- this is bad
default via 10.0.1.1 dev enp2s0
default via 10.0.1.1 dev enp2s0 proto static metric 100
10.0.1.0/24 dev enp2s0 proto kernel scope link src 10.0.1.236 metric 100
10.0.1.1 dev enp2s0 scope link
10.42.1.0/24 via 10.42.1.0 dev flannel.1 onlink
10.42.3.0/24 dev cni0 proto kernel scope link src 10.42.3.1
10.42.4.0/24 via 10.42.4.0 dev flannel.1 onlink
169.254.0.0/16 dev veth88dd6cab proto kernel scope link src 169.254.94.24
169.254.0.0/16 dev vethe05a10d5 proto kernel scope link src 169.254.134.89
169.254.0.0/16 dev vethd54d4b62 proto kernel scope link src 169.254.213.243
169.254.0.0/16 dev veth7d7b382f proto kernel scope link src 169.254.52.68
169.254.0.0/16 dev vethb44f5cec proto kernel scope link src 169.254.211.83
169.254.0.0/16 dev vetha22f9001 proto kernel scope link src 169.254.37.78

That node cannot talk to anything else until I delete that route; whenever another pod deploys to that node the route comes back. Any idea what’s wrong? These routes don’t appear on another Debian Bookworm node in the same cluster:

# ip route show
default via 10.0.1.1 dev enp3s0 onlink
10.0.1.0/24 dev enp3s0 proto kernel scope link src 10.0.1.235
10.42.1.0/24 dev cni0 proto kernel scope link src 10.42.1.1
10.42.3.0/24 via 10.42.3.0 dev flannel.1 onlink
10.42.4.0/24 via 10.42.4.0 dev flannel.1 onlink
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1