On 11/09/2018 04:16 PM, Guru Shetty wrote:
> Ankur's proposal:
> ==============
>
> Though the complete patches do not exist, Ankur wants to solve
the problem
> 1 by having a chassis specific MAC. So when packet leaves a
hypervisor for
> east-west routing, it uses a unique mac. The disadvantage with this
> proposal is that the VM (i.e logical port) will see the mac of
its first
> hop router change continuously which may have some yet to be clearly
> defined side-effects (leads to more ARP requests from the VM).
I'm having trouble understanding this. Can you elaborate a little bit?
Consider the following topology
VM-A (Hypervisor-A) <-----> Switch-A <------>Router <--------> Switch-B
<----------------> VM-B (Hypervisor-B)
When VM-A wants to send traffic to VM-B, it will send a packet with
source mac of VM-A and destination mac of Router.
When VM-B responds to that packet and when VM-A eventually receives that
packet, the source mac address of packet will be chassis specific mac of
hypervisor-B. So VM-A sends packet to one mac address, but receives
packet from another mac address.
Got it. Now it makes sense.
Is this not a problem with the current implementation too?
With the current implementation, since router mac address is the same
across the cluster, it is consistent.
>
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
