From: DougAtHome (dcalton@fuse.net)
Date: Thu Apr 10 2003 - 15:43:34 GMT-3
I was working on Solie's "Unnamed" lab and struggling with synchronization,
etc. To understand the problem, you have to know that in this lab, a
"backbone" router is set up to run both RIP and BGP to my lab router, and RIP
also has network statements for the networks being advertized by BGP (all
originate with loopbacks).
My problem seemed to be that the BGP routes learned on my lab router from this
"backbone" router would turn to "possibly down", entering holddown after a few
minutes. After struggling with several scenarios, searching CCO and
newsgroups, on a lark, I turned on debug ip routing, and cleared out bgp.
Voila, I see the route first gets installed via RIP, and then gets replaced
with the BGP entry later when the neighbors are established. Apparently, this
somehow confuses the route table process, because - although RIP continues to
receive RIP updates every 30 seconds - they bounce off the routing table
update, since the entries are now BGP. Unfortunately, the routing table entry
seems to be a bit confused, and after the holddown period, the route is
deleted :(
My "fix" was to filter out the RIP route coming in from the "backbone" with a
distribute-list. Of course, this flies in the face of the published
solutions, but then, what else is new, eh?
Anyway, anyone have a clear explanation as to what might be happening here,
and if it is a bug, or just a rational behavior that I don't understand?
This archive was generated by hypermail 2.1.4 : Thu May 01 2003 - 13:35:50 GMT-3