default route: bgp vs. static

Question

I'm asking why should one receive a redistributed default route from bgp upstreams and not creating an own static default route pointing to the bgp neighbor as next-hop? Are there any advantages in receiving the default route from bgp? Thanks for answering :)

Answer 1

If you have two ISPs (active/active or active/backup) and the BGP session to one of the ISPs drops, all traffic will automatically be re-routed to the other ISP if you use BGP default routes. If you have static default routes, the traffic will be black-holed to the ISP that dropped.

Answer 2

If an ISP is not sending a complete routeing table, sending a default route is (obviously) shorthand for everything else, which reduces the load on BGP. [If you are multi-homed to one ISP it might save effort if the ISP only sent routes where it matters which connection is used. But if the ISP has a network "moment" and (some of) the border routers you connect to can no longer reach the entire Internet, then some traffic will fall into a black hole pointed to by the default route !]

Pointing a static default route at a BGP neighbour strikes me as a bold move. The essence of BGP is that the neighbour advertises routes for the prefixes it is prepared to forward towards -- prefixes it does advertise routes for, it cannot (or does not wish to) forward towards !

If you only have one ISP (and you are using BGP) a default route received by BGP will, of course, be withdrawn if the BGP session drops (for whatever reason) -- which makes more sense than a static default route.

My feeling is that default routes and BGP are not a good fit, except, perhaps, for exotic use cases for AS's with strictly limited BGP connectivity. [But I mostly know BGP from the implementation of a BGP daemon perspective.] Would you be better off asking this question in the Network Engineering forum ?