[j-nsp] MX480 - 10.4R4.5 BGP

[Keith]

Try to make this short. I don't have any gear that I can run in a lab setting to
really get to know how Juniper BGP stuff works so we get some help and
they give me access to an MX to play on once in a while.

Have 3 BGP peers. Have four networks we are announcing.

Peer #1 - all 4 networks are prepended with our AS 5 times:

as-path-prepend "18988 18988 18988 18988 18988"

Longer AS-Path so no traffic from the internet should flow over the link.

Peer #2 - two networks prepended 3 times, two networks not prepended
so all traffic on those two networks comes back via this peer.

Peer #3 - opposite of Peer #2. The two networks not prepended on Peer 2
are prepended 3 times and the other two nets are prepended.

This way I have two networks coming in on one gig link and the other two
networks are coming in over the other gig link.

There should be no traffic incoming in on Peer #1 from the internet. But there is.
It appears we have made it so that the incoming traffic for two networks is split, but not 
evenly,
between Peer #1 and Peer #3.

Route view looking glass sites show that the routes have propagated and appear to be
in the routing tables correctly.

All traceroutes I have done to 2 of the neworks from outside have come in via Peer 3,
which is correct.

The other two networks come in via peer 2, which is also correct. But for some reason
on Peer #1 there is about 150 megs of incoming traffic.

Traceroutes from outside all come back via the correct links.

One thing I am confused about is active vs recieved prefixes.

Peer #1:
Table inet.0
     Active prefixes:              435449
     Received prefixes:            435467
     Accepted prefixes:            435465

Peer #2:
Table inet.0
     Active prefixes:              271
     Received prefixes:            430345
     Accepted prefixes:            430343

Peer #3:
Table inet.0
     Active prefixes:              804
     Received prefixes:            430160
     Accepted prefixes:            430159

One big thing and I am unclear on how all the knobs work on BGP, but on peer 2 and 3 we have
set a local-preference of 50 so all outbound goes out peer #1. We had to do that a while 
ago, I
think because peer 2 was not accepting all of our prefixes at the time, but they are now.

Is this why we have such a low number of active prefixes on peer 2 and 3?

What would be the effects of removing the local-preference 50 from peer 2 and 3 on our 
traffic?

Anyone have some clues they can impart?