[c-nsp] Routing failover

[Jee Kay]

I have a slightly odd situation with a customer at the moment, and I
was wondering if anyone here might be able to suggest a solution..

We have two connections to the customer, currently overe BGP. The two
peerings are configured such that one is 'primary' and one is
'secondary'. If they are both up, we expect all traffic to traverse
the primary link. In the event of the primary link failing, all
traffic will fail over to the secondary link. So far so good.

Now the problem is that due to the way they firewall the connections
(stateful firewalling isn't always a good thing), their internal
clients are disconnected from our services when the lines fail over.
When this happens because the backup line is kicking in that is fine,
but... When the primary line comes back up, due to the configuration,
all traffic will switch back to that and cause yet another disconnect.

It's that second disconnect I'm trying to get rid of. Ideally traffic
should stay on the backup line until a manual intervention sends it
back onto the primary line.

At the moment I'm toying with BGP dampening and seeing if we can
simply implement a 'long' dampening time such that it would
effectively be down until a manual 'clear ip bgp dampening', but it
seems a little.. untidy.

Are there any better solutions available?

Thanks,
Ras