Open Stack

On Thu, Sep 17, 2015 at 12:35 PM, Kevin Benton <blak111 at gmail.com> wrote:
>>Also I believe that (c) is already true for Neutron external networks -
>> i.e. it doesn't make sense to assign a floating IP to an instance that is
>> directly on an external network. Is that correct?
>
> Well not floating IPs from the same external network, but you could
> conceivably have layers where one external network has an internal Neutron
> router interface that leads to another external network via a Neutron
> router.

Today, a floating IP implies NAT to the instance's private IP. Without it,
the instance won't understand why its getting traffic destined for some
random public address. Also, today's floating ip implementation in Neutron
requires a router between the external network and the private network with
the instance.

Kris Lindgren described to me something that they do with floating ips that
doesn't use NAT. They inject routes to route traffic straight to the
instance, adjust allowed-address-pairs to allow the traffic in to the port
and then do something to inject the address in to the VM instance (iiuc)
for the instance to accept the traffic directly.

We've also thrown around the idea of doing may on the compute host.  I
don't know.  The point is that I think there may me some room to improve
our expand on the idea of floating ip.

Carl
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20150917/874271ad/attachment.html>