[ovs-dev] [PATCH 2/2] support L3 logical router in vtep emulator

[Guru Shetty]

I did spend quite some time looking at this code and VTEP schema. The goal
for me has been not only to review this piece only for VTEP emulator but
also to see the future viability of using this in a L3 edge device for OVN.
I feel that it may make sense to step back a bit and rethink the whole
architecture of the VTEP emulator.

If we go the current path, I think it will become increasingly difficult
adding more feature rich code, reviewing and testing it just for VTEP
emulator (for e.g. L3 with ARP, stateful security, NAT etc). We will end up
having to implement every new feature added to OVN again at the L3 edge
with a complete different implementation. Even for L2 gateway, if the idea
is to use a dpdk gateway, someone may have to write new code.

So instead of the current approach, consider the following new approach.
Instead of the VTEP emulator programming multiple OVS bridges via openflow
(as currently done), it indirectly uses the intelligence built in OVN code
base that converts OVN SB database to openflow by:

1.  It instantiates a OVN NB and OVN SB schema that is only local.
2. OVN NB currently has vlan+vif = logical port logic, so we get the
physical port + vlan = logical switch binding in VTEP schema.
2.  For every record created in VTEP schema by controller, a corresponding
record is created in local only OVN NB schema and that percolates to OVN SB.
3. ovn-controller or something similar will convert this information to
openflow flows.

We still need something that populates locally learned mac address back to
VTEP schema, so that will be a new enhancement.

We also need to propose enhancements for VTEP schema for NAT.

What do you guys think about the overall approach?



On 23 December 2015 at 08:57, Justin Pettit <jpettit at ovn.org> wrote:

>
> > On Dec 23, 2015, at 12:49 AM, Ben Pfaff <blp at ovn.org> wrote:
> >
> >> On Mon, Dec 07, 2015 at 09:58:53AM -0800, Justin Pettit wrote:
> >>
> >>> On Dec 7, 2015, at 9:15 AM, Ben Pfaff <blp at ovn.org> wrote:
> >>>
> >>> On Tue, Dec 08, 2015 at 07:35:44AM +0800, Shuangmin Zhang wrote:
> >>>> This patch is to support L3 DLR E-W traffic functionality in
> ovs-vtep. So it is focus on supporting switch-binding field in
> Logical_Router Table schema.
> >>>>
> >>>> Like what we do for logical switch, there are 2 parts of work
> >>>
> >>> Is there a patch 1/2 somewhere?
> >>
> >> I'd asked him (off-list) to resubmit the patch with a couple of
> >> superficial fixes.  I think he meant this to be "v2", not "2/2".
> >
> > OK.
> >
> > Justin, are you going to review this, then?
>
> Guru is reviewing this. He's getting an understanding of L3 in VTEP right
> now.
>
> --Justin
>
>
> _______________________________________________
> dev mailing list
> dev at openvswitch.org
> http://openvswitch.org/mailman/listinfo/dev
>