Open Stack

Hi everyone,

This is a very basic doubt and I'm trying to understand this
fundamental thing about creating networks in neutron. My ultimate goal
is to have all instances contain just one interface and a public IP on
them. Now, this public IP can only exist in a specific VLAN, lets say,
they'll only exist on eth0.123 (the is a vlan tagged interface on the
host and I can directly bind public IPs to them and they work
perfectly). I'm using linux bridge + ML2 as the plugin for neutron.

So, the ultimate goal will be something like:

[root at compute1 ~]# brctl show
bridge name     bridge id               STP enabled     interfaces
brq732eb7f9-16          8000.002590c6438e       no              eth0.123

                      tap81474f06-29


Now, with my limited knowledge, there are 2 ways of doing this:-

1. Create the vlan tagged interface manually and then setup "flat"
provider networking with physical_interface_mappings set with
eth0.123.
2. Use a "vlan" provider network and mention "123" as the vlan id in that.

I observed that, both of these approaches eventually create the same
kind of bridge configuration. Can anyone explain what's going on and
when to use what? I'm sure I'm missing a key concept and I'll be glad
if someone can clear that for me.


-- 
Cheers,
Abhijeet Rastogi (shadyabhi)