BGP terminology question

Mailman List Mirror (Read Only)
1

I asked this question on inet-access and it was suggested I try NANOG.

I understand BGP flapping to be announcements followed by withdraws over a short period. I am seeing a peer with a large number of announcements and the normal number of withdraws. Is there a term to describe what I am seeing? I’d like to understand what is happening, but I’ve been looking for more info and can’t seem to find anything. I suspect I am just not using the right words to search.

If there isn’t a term, why would a peer announce thousands of time an hour with very few withdraws?

2

There is a term, it's called "broken".

A peer should never announce a route it has already announced unless that route is withdrawn. (If the session goes down or is reset, that counts as a withdrawal.)

3

At the risk of sounding like a total moron, can anyone explain what is happening here?

This is from RIS, specifically RRC00. Here is some sample output of route_btoa from this file:
http://data.ris.ripe.net/rrc00/2005.11/updates.20051106.0430.gz

BGP4MP|1131251415|STATE|193.0.0.56|3333|1|2
BGP4MP|1131251415|STATE|193.0.0.56|3333|2|4
BGP4MP|1131251415|STATE|193.0.0.56|3333|4|5
BGP4MP|1131251415|STATE|193.0.0.56|3333|5|6
BGP4MP|1131251415|A|193.0.0.56|3333|8.11.252.0/23|3333 3356 11168|IGP|193.0.0.56|0|0||NAG||
BGP4MP|1131251415|A|193.0.0.56|3333|8.11.254.0/23|3333 3356 11168|IGP|193.0.0.56|0|0||NAG||
BGP4MP|1131251415|A|193.0.0.56|3333|8.10.241.0/24|3333 1103 1273 6395 22324 22324|IGP|193.0.0.56|0|0||NAG||
BGP4MP|1131251415|A|193.0.0.56|3333|8.15.2.0/24|3333 6320 8001 6395 26049 26049 26049 26049|IGP|193.0.0.56|0|0||NAG||

I understand AS3333 is RIS itself, is this some kind of misconfig on their end? It seems to be announcing it’s entire table every 5 minutes. This started late Friday and ended a few hours ago.

4

I understand BGP flapping to be announcements followed by withdraws over a short period. I am seeing a peer with a large number of announcements and the normal number of withdraws. Is there a term to describe what I am seeing? I'd like to understand what is happening, but I've been looking for more info and can't seem to find anything. I suspect I am just not using the right words to search.

If there isn't a term, why would a peer announce thousands of time an hour with very few withdraws?

There is a term, it's called "broken".

A peer should never announce a route it has already announced unless that route is withdrawn. (If the session goes down or is reset, that counts as a withdrawal.)

There's another term for that behavior. It's called "compliant".

There are a number of good implementation reasons why it is reasonable for an implementation to announce a route that it has already announced (e.g., peer groups). Admittedly announcing thousands of times an hour does NOT seem reasonable, but 'never' is not a requirement of the BGP spec either.

Tony

5

A peer should never announce a route it has already announced unless
that route is withdrawn.

one of many counterexamples: change in igp will cause change in
med. any attribute changes, and announcement is required.

e.g., an internal igp oscillation could cause what the op
describes.

randy

6

For the OP, http://www.ietf.org/rfc/rfc3345.txt

7

I give good odds this is not the oscillation issue. More likely a flapping IGP link and a lack of pull-up use (or pull-ups not installed such that link flaps would be non external impacting) etc...

I like pull-ups on all core devices personally...