[c-nsp] BGP update group leader

[Oliver Boehmer (oboehmer)]

> I'm trying to optimize the route reflectors for faster convergence;
the
> network has a wide range of PE types from 2800 to GSRs. I have two
> questions about BGP dynamic update-group that I haven't got
documented:
>
> 1. How it's the group leader elected? From experience it seems that
> highest IP address / router-id wins and it can preempt an already
> established leader.

not entirely sure, but it seems that a leader is elected during an
update run based on who is furthest behind..

> 2. The update queue size displayed in show ip bgp replication / show
ip
> bgp vpnv4 all replication is referring to the whole group or just to
the
> leader. In other words, does a slow peer elected as group leader
> influences the convergence of whole group?

A slow peer definitly has negative impact on convergence of the whole
group, and 12.2SRE/12.2(31)SB16 and XNE has enhancements to deal with
this dynamically ("bgp slow-peer detection")

	oli