Each label must be mapped to a unique bridge. A given bridge is only good for a single flat network, since it is untagged traffic. Thus, if you want to have two flat networks you would need two different label/bridge mappings, and two different server interfaces - one in each bridge.

I just answered your other thread, but the answer should be the same:

Per host you can have:

• 1x integration bridge where VMs connect
• 1x tunnel bridge used for overlay (vxlan /gre) traffic
• 1x or more provider bridges that connect to physical interfaces used for tagged (vlan) and untagged (flat) traffic

bridge_mappings defines the mapping of a label (or name) to an actual bridge. The label name is arbitrary, but must be mapped to an actual bridge on the host. The bridge should contain a physical server interface that likely connects to a physical switch, such as eth1. The OVS agent will connect the provider bridge to the integration bridge. When you create a provider or external network using the net-create command, you can specify the provider label to ensure traffic on that network utilizes the mapped bridge and respective interface.

You can have more than one provider bridge, each with its own physical interface or bonded interface. This is useful when you have, say, a 1G network and a 10G network:

bridge_mappings = slownet:br-eth1,fastnet:br-eth4

The br-eth1 bridge would contain eth1, a 1G interface. The br-eth4 bridge would contain eth4, a 10G interface. The labels must be consistent across all hosts, but the bridges they map to can be unique to that host. As a result, I would make the labels fairly generic and not identify some interface name that may not be consistent across hosts. Same goes for the bridge, too, if it may not always contain eth1 or eth4.