iBGP & eBGP Turn up

[l.rosas]

Hi all,

We are deploying couple new router and setting up eBGP and iBGP, but having issue with routing table size on one the routers.

Here is some problems I need to resolve during migration.

1. R1 have two upstream provider and we are  accepting entire IPv4 routing table. 

2. R1 is connected to SW1 running OSPF for our Internal subnets.

3. R2 have one upstream provider accepting entire IPv4 routing table.

4. The connection between R1 and R2 is establish though OSPF.

5. Try bringing up iBGP between R1 and R2, but session keeping closing and all other peers on R1 flaps. Error message on R1:

 %BGP_SESSION-5-ADJCHANGE: neighbor X.X.X.X IPv4 Unicast topology base removed from session  No memory

 %BGP_SESSION-5-ADJCHANGE: neighbor X.X.X.X IPv4 Unicast topology base removed from session  Peer closed the session

Since R1 probably cant take a 3rd full routing table coming from R2 what would be a better solution for iBGP?

Thx

[Vasilii Mikhailovskii]

Hello.

First of all, I would ask - why do you need 3 full views on your site? Do you really need this? If not - just receive 0.0.0.0/0 from your ISPs and use primary/standby mode.

If you are sure about full views - you may filter out longer prefixes from BGP (filter out all ge 24) and learn 0.0.0.0/0 to route for missing prefixes.

[l.rosas]

Thx Vasilli. In that case iBGP between R1 and R2 can I just receive default-route from R2 and still keep 3 full views? Both R1 and R2 are not in the same Colos.

R1

     Network          Next Hop            Metric LocPrf Weight Path
 r>i 0.0.0.0          x.x.x.x            0    100      0 i

Total number of prefixes 1

R2

ip prefix-list default-route description The Default Route
ip prefix-list default-route seq 5 permit 0.0.0.0/0

route-map DEFAULT_OUT permit 10
 match ip address prefix-list default-route

 neighbor x.x.x.x route-map DEFAULT_OUT out

[Vasilii Mikhailovskii]

Hello,

Sure you may send only 0.0.0.0/0 from R2 to R1 and this should help to conserve memory.

[Jason M.]

What size links do you have between the Colos to the ISP? Just curious as to what the goal of having full routes are for you?

That's really what should determine which routes you should accept between the sites when trying to determine how you can conserve memory. If you save memory and do not achieve the routing goal then you're wasting effort.

[l.rosas]

All 10G links to all 3 ISP's.

Plus Any2 in LA and SFMIX is SF also 10G.

[Jason M.]

From my experience most companies with BGP connections are given up to 1 AS beyond the ISP and a default then advertise their own blocks. That way if the destination is hosted by the ISP directly or they have direct peering you can take advantage of those generally lower latency/high bandwidth connections. Do you really need to know the quickest way to get to China?