[c-nsp] bgp keepalive/hold timers on ethernet links

[Simon Leinen]

matthew zeier writes:
> BGP's default 60/180 timers are too long

Note that 60/180 is *Cisco's* default for BGP.  RFC 1771 (BGP-4)
suggests 30/90.  This is still long, but already way better for your
situation.

> to go before dropping the peer.  The PHB wanted 1 second keepalives
> and a 3 second hold timer.  However, as soon as I started pulling in
> traffic (and only 50Mbps), I began frequently dropping the the peer.

Interesting.  Maybe BGP is sometimes busy for >2 seconds
sending/receiving or processing BGP updates (from other peers?), so
that the hold timer expires.

We use 10/30 on some important peerings, in particular all iBGP
peerings.  That seems to work nicely for us so far (note that we don't
take full routes; we only have about 30'000 IPv4 BGP routes).

> I'm guessing that these timers are too aggressive - anyone have any
> practical suggestions on how to fix this?

The obvious one is: use a longer hold time.  The other was mentioned
too: try BFD when BGP support comes out on your platform (same as
ours).  I would be curious how well this works.  I'd assume that BFD
will be handled on the MSFC at least for the time being, so it would
still compete with other processing there.  Hopefully BFD will at
least be handled in an interrupt handler.
-- 
Simon.