Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
New Member

BGP Routing Question

I have a requirement to support failover processing at two remote data centers where the hosts will both have the same IP assignment. I have MPLS circuits into a Cisco 4331 at each site and using BGP as the routing protocol.  BGP peer is setup to external processors network at both ends and a peer to the MPLS carrier. Can BGP be configured to prefer the external processors production site and in the event of a comm failure to that site, redirect traffic to the backup site?

1 ACCEPTED SOLUTION

Accepted Solutions
New Member

Yes, prepending the AS number

Yes, prepending the AS number to increase AS path length for the prefix advertised from the backup datacentre would be a sensible solution.

3 REPLIES
Bronze

Hi Demiller,

Hi Demiller,

Not sure if I understood the requirement correctly but let me give it a try.

You have two servers located at two different data centers but they have same IP address. You want these servers to be accessed from remote sites (connected via MPLS) in primary and standby mode.

1. I suggest you to advertise specific subnet (into BGP) from preferred data center. Advertise less specific subnet (into BGP) from standby data center. Make sure you must have same route in your routing table first so that BGP can consider it for further advertisement. You may need to advertise an Host route into MPLS from active data center. If it is Nexus, then there are chances of CPU consumed for the traffic to this server since packets will be process switched.

2. Other way of using LISP (Location identifier separation protocol) which is bit complex to implement and may require few more resources.

Please rate if it helped.

HTH

-Amit

New Member

not sure i follow. i thought

not sure i follow. i thought of using a route map at the back up data center and prepend the route. would that work?

New Member

Yes, prepending the AS number

Yes, prepending the AS number to increase AS path length for the prefix advertised from the backup datacentre would be a sensible solution.

9
Views
0
Helpful
3
Replies
CreatePlease to create content