[ovs-dev] [PATCH] bond: Send gratuitous ARPs when load balancing.

[Ethan Jackson]

I didn't check the kernel implementation so I don't know what they
actually do on the wire.

I'm not particularly convinced that this is important, but Sanjay
thinks it is.  It was easy enough to write up so I thought we could
discuss it on the dev list.  I'm not married to the patch either way.
 Though agree I don't really see the value in forcing traffic to
ingress the same link we send it on.

This patch does pave the way towards dropping traffic which comes in
on the wrong slave.  In order to make something like that workable we
would need to update the partner switch whenever load balancing
happens.  There is a small risk of black-hole-ing traffic in some
situations.  However, I think that can be mitigated by only dropping
traffic which we know about in our learning table, or sending
gratuitous arps to update the network in those situations.

Again I'm hesitant on the idea, but I'm also not expert on how these
things look in practice.  I'm fine with doing whatever approach
develops consensus.

Ethan

On Fri, Jul 1, 2011 at 09:15, Ben Pfaff <blp at nicira.com> wrote:
> On Thu, Jun 30, 2011 at 06:49:46PM -0700, Ethan Jackson wrote:
>> According to the following documentation, we should be sending
>> gratuitous ARPs when rebalancing causes us to shift traffic to a
>> different slave.
>>
>> http://support.citrix.com/article/CTX124421
>
> I guess it can't hurt, but I'm not sure of the value.  We only shift
> traffic if there's a significant amount of it, in which case the
> next-hop switch will find out that the MAC has moved at just the right
> time anyhow.  I guess it will make a difference if the traffic has
> actually stopped abruptly and we're just exponentially timing out the
> byte counter, so maybe it's worthwhile for that reason.
>
> Did you look at Citrix's kernel implementation of SLB to see if it
> does that?  If it doesn't, should we follow the implementation or the
> documentation?
>