Warning message

  • Cisco Support Forums is in Read Only mode while the site is being migrated.
  • Cisco Support Forums is in Read Only mode while the site is being migrated.

Dual ISP BGP Routing Issue

Unanswered Question
Jun 19th, 2006
User Badges:

I am connected to two tier 1 ISP's and take full routes from both. I have a T3 to one and dual T1's to the other. The dual T1's are intended to be a backup to the T3. I have local-pref set to 200 for all routes learned via the T3 ISP. I also am prepending my AS# several times to the T1 ISP. For 99% of routes this has the desired affect of routing traffic via the T3. For some reason the T1 ISP is learning more specific routes for a small number of IP blocks. This causes a small number of /24 blocks to use the T1's.

I want to ensure all traffic utilizes the T3 because the asymmetric paths seem to fail.

I was thinking of modifying BGP to only look at /16 or larger IP blocks.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
mheusinger Mon, 06/19/2006 - 06:54
User Badges:
  • Green, 3000 points or more


seems like a standard BGP problem in the internet. someone aggregating routes of someone else or improper aggregation.

Blocking everything below /16 would surely cut you off large portions of the internet unless you install a default route somewhere. Probably also not what you want, as you lose part of your redundancy, which seems a rather high price to pay to solve this problem.

The first approach would probably be to contact the ISP announcing the /24s asking to aggregate them for you the way your other ISP does.

One other way to tackle the problem is to find out who aggregates and who doesn?t by looking at the BGP attribute atomic aggregate. Maybe you can identify the AS where things go wrong and contact them (->whois AS#).

Hope this helps! Please rate all posts.

Regards, Martin

olorunloba Mon, 06/19/2006 - 10:39
User Badges:
  • Silver, 250 points or more

To make it simpler. Ask both of your ISPs to send you a default route via BGP. Do without the full internet table, and use local-preference as you have use previously to prefer the T3 connection to the T1.

Since you are not looking at any form load-balancing, and the T1 link is there in case of failure of the T3, practically you do not need the full internet table. BGP will still detect link failure and route traffic to the secondary link, and traffic will only use one link at a time.

Hope this helps


This Discussion