BaGPipe for Neutron L2

Installation in a devstack test/development environment

  • install devstack (whether stable/kilo or master)

  • enable the devstack plugin by adding this to local.conf:

    • to use branch stable/X (e.g. stable/mitaka):

      enable_plugin networking-bagpipe stable/X
    • to use the development branch:

      enable_plugin networking-bagpipe master
  • enable bagpipe ML2 by adding this to local.conf:

  • note that with devstack, bagpipe-bgp is installed automatically as a git submodule of networking-bagpipe

  • for multinode setups, configure bagpipe-bgp on each compute node, i.e. you need each bagpipe-bgp to peer with a BGP Route Reflector:

    • in local.conf:

      # IP of your route reflector or BGP router, or fakeRR:
    • for two compute nodes, you can use the FakeRR provided in bagpipe-bgp

    • for more than two compute nodes, you can use GoBGP (sample configuration) or a commercial E-VPN implementation (e.g. vendors participating in EANTC interop testing on E-VPN)


On Neutron servers, the following needs to be done, based on an ML2/linuxbridge configuration as a starting point:

  • installing networking-bagpipe package

  • in ML2 configuration (/etc/neutron/plugins/ml2.ini):

    • enabling the route_target type driver (typically keeping flat and vlan type drivers)

    • adding the bagpipe mechanism driver (additionally to the linuxbridge driver which will still handle flat and vlan networks)

    • configuring the use of the route_target type for tenant networks

    • configuring the AS number and range to use to allocate BGP Route Targets for tenant networks

    • example result:

      tenant_network_types = route_target
      type_drivers = flat,vlan,route_target
      mechanism_drivers = bagpipe,linuxbridge
      rt_nn_ranges = 100:319,500:5190
      as_number = 64512

You need to deploy a BGP Route Reflector, that will distribute BGP VPN routes among compute and network nodes. This route reflector will need to support E-VPN and, optionally, RT Constraints. One option, among others is to use GoBGP (sample configuration).

On compute node and network nodes the following needs to be done, based on an ML2/linuxbridge configuration as a starting point:

  • installing networking-bagpipe and bagpipe-bgp python packages

  • configuring Neutron linuxbridge agent for bagpipe /etc/neutron/plugins/ml2.ini:

    • enabling bagpipe agent extension

    • disabling VXLAN

    • result:

      extensions = bagpipe
      enable_vxlan = False
  • configuring bagpipe-bgp

    • setting local_address to the compute node address
    • adding the Route Reflector IP to peers
    • enabling linux_vxlan.LinuxVXLANDataplaneDriver for EVPN

BaGPipe for BGPVPN

Information on how to use bagpipe driver for networking-bgpvpn is provided in BGPVPN bagpipe driver documentation.