BGP question when upgrading ISP link

coughlind · ‎09-25-2007

We will be upgrading our ISP link from a DS3 to an OC3 link. We will be keeping our existing IP addresses and AS number. The new interface already has the same configuration as the exsisting and is currently shutdown. Once I shutdown the DS3 link and bring up the OC3, the BGP should reconverge over the new WAN link, correct? Should the the routing table be cleared with a "clear ip route *"?

JORGE RODRIGUEZ · ‎09-25-2007

David, from what you are discribing this implementation involves a physical link cutover, it is obious bgp peering will get disrubted during cutover and routes will not be there until the peer neighbor reconnects with your router, and yes, in theory it should reconverged because you are not adding or deleting any bgp logical configuration in either routers , you need to worry about physical interface and that peer relation is established. What I would suggest prior to the cutover is to observe and note current bgp route table by issuing " show ip route " or " show ip route bgp AS# " , " show ip bgp neighbors ", these commands can also be issue after cutover. If after cutover peer neighbor is there but no routes you can always do " clear ip bgp * " to reset all bgp sessions.

HTH

Jorge

Jorge Rodriguez

Calin C. · ‎09-25-2007

Hello,

I believe that you do not have to do a "clear ip route *", but only a "clear ip bgp PEER_AS_NUMBER" after the new connection is up.Why to clear ip route table? Maybe you have there another routes learned from another IGP protocol.

And another question is how fast the physical connection will be replace.Because if you cut the connection for more the half-hour, than the BGP dead timer will occur and the routes are cleared automatically.If the connection is change in minutes, the my advise is to "clear ip bgp PEER_AS_NUMBER".

Good luck!

Best regards,

Calin

coughlind · ‎09-26-2007

Thank you both. We actually cutover the link this morning. Other than a mismatch on the CRC, it went fine. The BGP neighbor came up as well as the routing. When we tested, they left it at the default of 16, but when the implementation engineer brought it up, he had set it to 32.

Thanks again.