Arie Vayner \(avayner\) writes: > Danny, > With iBGP the timers for BGP are not really important... You > actually need to worry about the IGP convergence. > The reason for that is that usually when a link fails, you don't > really expect the BGP session to the RR to go down, but just use the > redundant IGP path.
Yes, but what about when a router fails, in particular a border (eBGP+iBGP) router? In such a case, iBGP timers (or the configuration of a mechanism such as BFD) will determine how long it takes for other routers that the eBGP routes from the dead router have to be dropped. This can be very important, because using the dead router's eBGP routes can mean blackholing traffic. On our network (iBGP mesh with 3 RRs and 38 clients, mostly 7600) we have been using 10/30 seconds keepalive/hold on all iBGP sessions for a few years. That has been working so well that we could consider lowering the timers some more. Of course someone from Cisco will now step in and tell us that changing routing protocol timers sucks, and how much better it is to use shiny new BFD. (Damn, I forgot this is a Cisco list! Of course Cisco routers never fail, so you only have to worry about link failures. Just forget everything I just said. :-) -- Simon. _______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/