While using tailscale, uninstall should clean routes in table 52 that was added on startup.

[ShylajaDevadiga]

Environmental Info:
K3s Version:
k3s version v1.27.2+k3s-b66a1183

Node(s) CPU architecture, OS, and Version:
SLES15 SP3

Cluster Configuration:
Single-node or multi-node

Describe the bug:
After installing k3s with tailscale, pods are in CrashLoopBackOff state due to leaving old obsolete routes in the table.

Steps To Reproduce:
Install multiple clusters using default route

Expected behavior:
routes in table 52 should be removed while uninstalling to avoid network issues.

Actual behavior:
routes in table 52 are present after uninstall

Additional context / logs:
After uninstalling k3s

ip route show table 52
10.42.0.0/24 dev tailscale0 
10.42.1.0/24 dev tailscale0 
10.42.2.0/24 dev tailscale0 
10.42.3.0/24 dev tailscale0 
10.50.0.0/24 dev tailscale0 
1.2.3.4 dev tailscale0 
1.2.3.5 dev tailscale0 
1.2.3.6 dev tailscale0 
1.2.3.7 dev tailscale0 
1.2.3.8 dev tailscale0 
1.2.3.9 dev tailscale0 
1.2.3.10 dev tailscale0 

[ShylajaDevadiga]

While uninstalling on k3s version v1.27.3-rc1+k3s1, routes in table 52 seem to persist and tailscale interface continues to have an IP

$ k3s -v
-bash: /usr/local/bin/k3s: No such file or directory

$ ip route show table 52
10.42.0.0/24 dev tailscale0 
1.2.3.4 dev tailscale0 
1.2.3.5 dev tailscale0 
1.2.3.6 dev tailscale0 
1.2.3.7 dev tailscale0 

$ ip a|grep tailscale
3: tailscale0: <POINTOPOINT,MULTICAST,NOARP,UP,LOWER_UP> mtu 1280 qdisc fq_codel state UNKNOWN group default qlen 500
    inet <REDACTED>/32 scope global tailscale0

$ tailscale status --json
{
  "Version": "1.42.0-t3a83d61ec-g6702f39bf",
...

[manuelbuil]

This is related to an incorrect scope of the tailscale key which makes all nodes share the same tailscale network, even if nodes belong to different k3s clusters. As a consequence, their subnets get pushed in table 52 as soon as the tailscale client logs in