Re: "Neighbor maximum-prefix" option on routers

[Alexander Koch <efraim () clues de>]

On Mon, 20 November 2006 09:03:17 +0100, Mikael Abrahamsson wrote:
[..]
> Having both max-prefix and route-map and prefix-list makes for good 
> engineering over time.

Ain't that a bit over- engineered? If you have a prefix list
(presumably to a customer) do you want to seriously shutdown
the session when he sends you random prefixes because it is
easy to break when you are new to it? Would create unneeded
tickets I'd say. Prefix lists (exact match) with no max-prefix
(as those are then rejected anyway) work fine here...

But then, I was not wondering what you or Joe do as you know
what you do and why for sure. ;-) But I would really be
interested in how smaller ISPs do it. Given how many ppl
still use prefix-lists outbound (and leak whatever is best
in their routing table when a customer is not announcing a
prefix to them) I wonder what can be done. Let alone the
random appearance of /24 announcements out of a /20 or so,
and when you check and ask and insist on it you find out
there is a redistribute statement and they just added a /24
internally...

Back on track, what is max-prefix good for anything but
peers? A transit session is pretty much 'all' already, and
customer are filtered anyway? (coming back to the initial
question in this thread)

Let alone how many operators monitor sessions down for
prefixes -- some ISPs deserve to be depxxxxx when they find
out after one month (been there, more than once) the
sessions are down, and they ask you why those are down,
cutting and pasting the 'Idle (pfxcount)' in their email
even. That is my personal view though.

So max-prefix is good for peers, maybe for customers, hardly
for transit sessions, and whenever you do it, MONITOR IT ;)=)
A three- line perl / sh script on top of your syslog or during
logrotate or so is advised if only that.

Alexander