Assuming the Harvester nodes' management interfaces are attached to the VLAN 1 network. The user creates an additional cluster network called provider (this implies using a secondary network interface other than the management one on the nodes) and then creates three VM networks, net-1, net-100, and net-200, with VLAN 1, 100, and 200, respectively. The net-1 and net-100 networks are associated with the default mgmt cluster network. The remaining one, net-200, is associated with the provider cluster network. The user then creates three LB IP pools for the three VM networks, called pool-1, pool-100, and pool-200.

Case 1: The VM is attached to the net-1 network, and an LB is created from the pool-1 IP pool. This configuration is straightforward, and it works out of the box.

Case 2: The VM is attached to the net-1 network, and an LB is created from the pool-100 IP pool. This configuration doesn't work because the LB IP address is currently always bound to the mgmt-br interface. Since the management interface is attached to the VLAN 1 network, VLAN 100 traffic won't reach the mgmt-br interface.

Case 3: The VM is attached to the net-1 network, and an LB is created from the pool-200 IP pool. This configuration doesn't work for a similar reason as case 2.

Case 4: The VM is attached to the net-100 network, and the LB is created from the pool-1 IP pool. This configuration works as the LB IP address, which is bound to the mgmt-br interface, is in the same VLAN network as the node's management interface. Traffic will be DNAT'd and routed by the default gateway to the backend VM once it reaches the mgmt-br interface.

Case 5: The VM is attached to the net-100 network, and the LB is created from the pool-100 IP pool. This configuration doesn't work for a similar reason as case 2.

Case 6: The VM is attached to the net-100 network, and the LB is created from the pool-200 IP pool. This configuration doesn't work for a similar reason as case 2.

Case 7: The VM is attached to the net-200 network, and the LB is created from the pool-1 IP pool. This configuration works for a similar reason as case 4. The only difference is the DNAT'd and routed traffic goes through the provider-br bridge but not mgmt-br.

Case 8: The VM is attached to the net-200 network, and the LB is created from the pool-100 IP pool. This configuration doesn't work for a similar reason as case 2.

Case 9: The VM is attached to the net-200 network, and the LB is created from the pool-200 IP pool. This configuration doesn't work for a similar reason as case 2.

5,6c5,6
< Currently Active Slave: None
< MII Status: down
---
> Currently Active Slave: enp59s0f0np0
> MII Status: up
10a11,18
>
> Slave Interface: enp59s0f0np0
> MII Status: up
> Speed: 25000 Mbps
> Duplex: full
> Link Failure Count: 0
> Permanent HW addr: 7c:fe:90:cb:73:62
> Slave queue ID: 0

ou26l-compute:
----------------
mgmt-bo

ou26r-compute:
----------------
cnet1-bo
cnet2-bo
mgmt-bo

ou31c-compute:
----------------
mgmt-bo