Link redundancy between two MPLS sites

[fasih2012]

Hi All,

I have a HQ and a DC site. I've subscribed 2 links going to the same MPLS provider. My local site AS number for this 2 link is the same.

I've intended to route all subnets on link 1 and in case of link 1 goes down would like to automatically failover to link 2.  I understand that I have to run iBGP and create peering between two routers. Also, I'm running OSPF internally between two sites and they are connected via 2x1Gbps Metro link.  I also would like to perform two way redistribution OSPF --- BGP, BGP----OSPF. 

This is my plan of attack

Oubound traffic:

1. change weight on HQ and DC for the ISP BGP neighbor to use 32769
- under router bgp
- neighbor XXXXXX weight 32769
- RACE condition ****

2. Under OSPF on HQ side use the higher metric (100) when redistrbuting BGP
- redistribute bgp 65001 metric 100 metric-type 1 subnets tag

###

2b. Under OSPF on DC side use the higher metric (50) when redistrbuting BGP
- redistribute bgp 65015 metric 50 metric-type 1 subnets tag ###

Incoming traffic:

3. AS PATH prepend on HQ
- Ask ISP to propogate prepended AS
- Under BGP
  - neighbor XXXXXX route-map SET_PATH out

Route-map SET_PATH
  set AS-PATH prepend 65001 65001 65001

I hope someone can give me a recommended solution or configuration advise

Thanks!

[milan.kulik]

Hi,

I'm not sure if understand your topology 100%, a diagram would help a lot here.

I suppose the 2x1Gbps Metro links are terminated on some other devices, not the BGP routers? And there are some other L3 devices running OSPF between the BGP routers and LAN in both sites?

I understand you want to use the DC line as primary?

a) What do you mean by RACE condition ****  ?

b) I suppose on DC side

- redistribute bgp 65015 metric 50 metric-type 1 subnets tag ###

is a typo and should be

- redistribute bgp 65001 metric 50 metric-type 1 subnets tag ###

c) metric-type 2 might be worth if you want to prefer DC as the outgoing gateway for ALL your devices as long as it's BGP peering is OK?

d) If you are planning to redistribute OSPF to BGP, too, you should use a route-map denying the OSPF prefixes tagged  by the tags used when BGP was redistributed to OSPF. That's for sure to prevent possible routing loops.

e) Isn't there any single point of failure in your topology?

Which might cause both Metro links failed at the same time, e.g.?

f) Also think about OSPF neigbourship consequences:

Will all OSPF devices share the same subnet?

Who will become a DR a BDR then?

Your BGP router will recieve the same prefix from the other BGP router via iBGP  and via OSPF probably.

Etc.

HTH,

Milan