True, although in this scenario would soft reconfig not be an answer?
As each router has two copies of the full table, one via the eBGP peer
and another from the iBGP peer. If the eBGP peer dropped all the iBGP
learned routes would remain and be used. When the eBGP peer came back up
soft reconfig would allow for a seemless move back to the prefered peer?
Ideally what is needed is BFD to detect the link failure between the
host and the external peer, that way the BGP timers could be set to
something more conservative. Also some means of reliable flap control
would be good to save restoring a session to an unreliable host.
Good point well taken though.
Stuart Henderson wrote:
On 2008-10-08, Simon Slaytor <[EMAIL PROTECTED]> wrote:
It's also important to tune the BGP dead timers as low as you can
if you do this, do it with care, it's a double-edged sword.
sure you pick up a dead session sooner, but, it greatly increases
the chance of killing a session when your or more likely your peer's
routers are working ok, forwarding ok, but a bit busy to handle
control plane traffic in a timely fashion.
when that happens, dropping the session and forcing them to feed
you full table is about the last thing you want to do...
.