A quick plain-English guide to the interface names you'll see on an EVE-OS device.

When EVE-OS boots, every physical NIC on the box gets discovered and given a name based on PCI location (like eth0, eth1, etc). These are your physical ports, the actual copper/fiber jacks on the hardware.

• The raw physical network port as EVE-OS first sees it.
• Shows up in the Hardware Model / device model as an adapter.
• If it's not used by a Network Instance, it just stays as ethX.

This is the one that confuses people.

• Only appears when a physical NIC is assigned to a Switch Network Instance.
• Switch NI is pure L2 bridging, meant to act like the physical NIC is directly wired to your VMs/containers.
• The bridge needs to be named eth0 so everything else in the system keeps referring to a stable, familiar name.
• The physical NIC is already using that name, so the kernel can't have two things called eth0.
• The physical NIC gets renamed to kethX (kernel eth) and demoted to being a port/slave on the new bridge.
• The bridge takes over the eth0 name and becomes the interface everyone talks to.

In short: eth0 stops being the physical NIC's name and becomes the bridge's name. The physical NIC gets relabeled keth0 and sits underneath as one port on that bridge.

• Created for Switch NI.
• Takes over the original interface name/identity.
• kethX (physical NIC) is a port on it.
• App VIFs (container/VM virtual NICs) are also ports on it.
• Physical NIC and app traffic share L2 like a real switch.

• Local NI is NAT plus DHCP plus DNS, not pure passthrough.
• It builds its own bridge but doesn't need to preserve the physical NIC's original name the same way.
• No kethX renaming here.

On tenant offboarding/re-onboarding, a stale kethX left over from a previous Switch NI config is a real failure mode. If the bridge/keth pairing didn't tear down cleanly, the rename can conflict with the next NI setup. Worth checking during migration troubleshooting.