kube-vip remove all ip address on the eth device

[maitredede]

Describe the bug
I have a small k3s cluster : 3 master Rpi4 (raspberry pi OS bookworm 64), connected on ethernet (ip are DHCP static). When installing kube-vip as daemonset, the vip is set, and after a few minutes, there is no longer any ip on the device.
It is hard to get kube-vip logs since when a node is isolated (no ip address), k3s drops its sockets, and crictl no longer works.
If I add manually the ip (ip addr...) and the default route, I can restart k3s, but when corum is available (2 nodes on 3), ip are removed again (a few moment later). All the ip are 192.168.20.xxx/24 : vip is .200, dhcp nodes are 201, 202, 203

I tried many options, but I don't know why either the DHCP address has netmask updated from 24 to 32, why VIP has netmask of 32, why all addresses are removed... I don't understand how options acts on final addresses and netmasks...

To Reproduce
Steps to reproduce the behavior:

1. Install kube-vip as daemonset
2. Wait a few minutes
3. All nodes have lost their DHCP address

Expected behavior
IP given by DHCP should be kept intact and VIP should have same netmask to work properly. Also routes should not be altered since they are on same network.

Environment (please complete the following information):

• OS/Distro: Raspberry Pi OS lite bookworm arm64
• Kubernetes Version: k3s stable v1.28.8+k3s1
• Kube-vip Version: v0.7.2

Kube-vip.yaml:

docker run --network host --rm ghcr.io/kube-vip/kube-vip:$KVVERSION manifest daemonset \
    --interface eth0 \
    --vip 192.168.20.200 \
    --vipSubnet "/24" \
    --cidr 24 \
    --inCluster \
    --taint \
    --controlplane \
    --services \
    --servicesElection \
    --arp \
    --leaderElection
    ```

[MartinLoeper]

Observed exactly the same on a tinkerbell k3s setup in my homelab (since tinkerbell uses kube-vip). Extremely difficult to debug...