gistfile1.txt

#########################################################
# Current layout(moving away from the octavia weirdness)
#########################################################
eth1 -> bond0(interface) -> br-vlan(host ip) -> (vlans defined in neutron unrelated to the base interfaces set up by support)
     -> bond0.EEE        -> br-vxlan         -> [ lxc veth6-0 -> lxc veth6-1 -> LXC Container ]
     -> bond0.AAA        -> br-ironic  -> [ lxc veth1-0 -> lxc veth1-1 -> LXC Container ]
                                       -> veth2-0 -> veth2-1 -> [ Neutron controlled bridge -> neutron AAA vlan int -> VM and/or neutron namespace ]
     -> bond0.BBB        -> br-lbaas   -> [ lxc veth3-0 -> lxc veth3-1 -> LXC Container ]
                                       -> veth4-0 -> veth4-1 -> [ Neutron controlled bridge -> neutron BBB vlan int -> VM and/or neutron namespace ]
     -> bond0.CCC        -> br-mgmt    -> [ lxc veth4-0 -> lxc veth4-1 -> LXC Container ]
     -> bond0.DDD        -> br-storage -> [ lxc veth5-0 -> lxc veth5-1 -> LXC Container ]
     -> bond0.EEE        -> br-ipmi    -> [ lxc veth6-0 -> lxc veth6-1 -> LXC Container ]



# End result to deployer
  - Easier on the deployer to do manually or automated(all configs very similar).
  - Ironic and Octavia need to be set up as a flat network in neutron with the host_bind_override set.
  - For new vlans, the deployer still has to get netsec to trunk down to infra and compute nodes as well as redeploy to update the neutron vlan range.




#########################################################
# Requested layout
#########################################################
eth0
eth1 -> bond0(interface) -> br-vlan(host ip) -> (vlans defined in neutron unrelated to the base interfaces set up by support)
                                             -> br-vlan.EEE -> br-vxlan   -> [ lxc veth6-0 -> lxc veth6-1 -> LXC Container ]
                                             -> br-vlan.AAA -> br-ironic  -> [ lxc veth1-0 -> lxc veth1-1 -> LXC Container ]
                                                                          -> veth2-0 -> veth2-1 -> [ Neutron controlled bridge -> neutron AAA vlan int -> VM and/or neutron namespace ]
                                             -> br-vlan.BBB -> br-lbaas   -> [ lxc veth3-0 -> lxc veth3-1 -> LXC Container ]
                                                                          -> veth4-0 -> veth4-1 -> [ Neutron controlled bridge -> neutron BBB vlan int -> VM and/or neutron namespace ]
                                             -> br-vlan.CCC -> br-mgmt    -> [ lxc veth4-0 -> lxc veth4-1 -> LXC Container ]
                                             -> br-vlan.DDD -> br-storage -> [ lxc veth5-0 -> lxc veth5-1 -> LXC Container ]
                                             -> br-vlan.EEE -> br-ipmi   -> [ lxc veth6-0 -> lxc veth6-1 -> LXC Container ]

# End result to deployer
  - A little more work on the deployer and a single point of failure if something needs happens to br-vlan
  - Harder to troubleshoot as it differs from upstream openstack docs and gives each connection an extra hop to track
  - Ironic and Octavia can be set up as either a vlan or flat network in neutron(You still have to use the host_bind_override as these are flat networks)
  - For new vlans, the deployer still has to get netsec to trunk down to infra and compute nodes as well as redeploy to update the neutron vlan range.
  - Results on multiple vlan interfaces of the same vlan which could cause issues depending on some of the crazy things deployers do.