[c-nsp] IPv6 first-hop redundancy: short-lived RA or FHRP?

[Phil Mayers]

On 10/24/2012 08:12 AM, Peter Rathlev wrote:
> On Tue, 2012-10-23 at 19:15 +0100, Phil Mayers wrote:
>> HSRPv6 gives you access to the "track" stuff which can be useful to have
>> more explicit control over how you failover and when.
>
> Right, that's a rather good point.

I should have added: we have used dual RAs in a couple of places where 
HSRPv2 isn't available (old-skool 3750s). We found it unsatisfactory 
primarily due to the lack of control. But we didn't use the kind of 
aggressive timers you describe, to be fair.

It's interesting that the IPv6 spec specifies that stacks will perform 
router reachablity detection using upper-layer feedback (if you sent a 
TCP data packet via gateway X and got an ACK, gateway X is up). This 
ought in theory to give gateway failover quickly without any upstream 
intervention or short timers. Sadly, that wasn't the case when we tested 
it :o(