Hello,

I found another issue related to VPP bonding and VLAN subinterfaces.

Reconfiguring a single subinterface on a bonded VPP interface causes all subinterfaces on the bond to go down, resulting in loss of connectivity across the entire bonded interface.

This makes the solution unstable in my environments.

vpp {
    bonding vppbond0 {
        hash-policy layer3+4
        member {
            interface eth6
            interface eth7
            interface eth8
            interface eth9
        }
        mode 802.3ad
        mtu 1500
        vif 38 {
            address 198.51.100.2/30
            address 2001:db8:100::2/126
            mtu 1500
        }
        vif 30 {
            address 203.0.113.2/24
            mtu 1500
        }
    }
}

}

Initial state:

vyos@r1# run show interfaces vpp

BondEthernet0 bond 02:fe:d9:33:87:4d 9000 up
BondEthernet0.38 bond 198.51.100.2/30 00:00:00:00:00:00 1500 up
BondEthernet0.30 bond 203.0.113.2/24 00:00:00:00:00:00 1500 up

Removing a single IPv6 address from one VLAN subinterface:

delete interfaces vpp bonding vppbond0 vif 38 address 2001:db8:100::2/126

Immediately causes both VLAN subinterfaces to transition to DOWN state:

vyos@r1# run show interfaces vpp

BondEthernet0 bond 02:fe:3e:39:fa:41 9000 up
BondEthernet0.38 bond 198.51.100.2/30 00:00:00:00:00:00 1500 down
BondEthernet0.30 bond 203.0.113.2/24 00:00:00:00:00:00 1500 down

Additional observations:

BondEthernet0 itself remains UP

All underlying DPDK interfaces remain operational

The issue appears to affect only VLAN subinterfaces on the VPP bonded interface

MAC addresses on subinterfaces are displayed as 00:00:00:00:00:00

VPP logs also report:
Secondary MAC Addresses not supported for interface index 0

This behavior is reproducible and occurs during normal configuration changes such as adding/removing IP addresses from a single VIF.

Environment:

VyOS rolling / 2026.03

VPP dataplane

LACP bond with multiple VLAN subinterfaces

Please let me know if additional debugging information is required.