On Thu, May 26, 2016 at 9:14 AM, Kyle Mestery <mest...@mestery.com> wrote:
> On Thu, May 26, 2016 at 10:46 AM, Guru Shetty <g...@ovn.org> wrote: > > > >> > >> > >> OK, I'm going to play with this a bit. I was mostly interested in > >> feedback from others. We're interested in this approach and so if we > >> take the patch up does it stand a chance of merging? > > > > > > If I remember correctly, the main problem was around ARP handling when > the > > destination's mac address is not known. Who would do the job that > > ovn-controller currently does for OVN? If a clean solution is proposed, > then > > I can't think of any reasons why it can't be merged. > > > Your last comments said you also didn't appreciate the overall > approach, which is what I was referring to. > > Either way, yes, the ARP handling needs to be solved, and VTEP schema > likely needs some work around NAT as you say. One advantage of exploring the VTEP schema for L3 gateway here is it is one formal way to remove some assumptive constraints associated with the existing OVN datapaths. You don't even have to stay within the bounds of the VTEP schema. > I'm going to revive this > patch and see what can be done here. > > Thanks! > Kyle > > >> > >> > >> >> > >> >> > >> >> >> > >> >> >> > >> >> >> Thanks! > >> >> >> Kyle > >> >> >> > >> >> >> [1] https://patchwork.ozlabs.org/patch/553435/ > >> >> >> _______________________________________________ > >> >> >> dev mailing list > >> >> >> dev@openvswitch.org > >> >> >> http://openvswitch.org/mailman/listinfo/dev > >> >> > > >> >> > > >> > > >> > > > > > > _______________________________________________ > dev mailing list > dev@openvswitch.org > http://openvswitch.org/mailman/listinfo/dev > _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
