HSRP and BGP

[cisco8887]

Hi Guys,

I have looked around the net for a solution to allow BGP with HSRP.

The problem is you cannot source BGP update from HSRP however I have managed to fix this as explained below using LAB.

I can only use /30 between ISP and my routers so here is how I fixed it by using /30 on ISP side and /29 on my side

What is your view on this ?

ISP1 10.0.0.2

R1 and R2 -

HSRP:10.0.0.1/29

R1 real: 10.0.0.3/29

R2 real :10.0.0.4/29

I am not sure if using .3 causes any problem as it is the broadcast of ISP1 but cannot see it causing any issues.

Connection is made from higher IP thus ISP will initiate a connection to 10.0.0.1 which is on the same /30 subnet , .1 will then reply sourcing packet from .1 back to ISP . So essentially to get around the limitation of the R1-R2 not being able to source packets from their HSRP , I am relying on ISP making the connection .

This works in GNS3, do you see any issues with real world implementation ?

[Mark Malone]

Hi

I wouldn't do that in prod network ,you would usually just use the BGP path attributes for redundancy than HSRP , BGP is very flexible, maybe it would work if you got the ISP involved but from my experiences with ISP they wouldn't work with you on something like that in real scenario 

so if you had 2 links , and wanted to prefer one,  set say local pref or med to alter the preferred paths in  our out

This doc shows multiple ways to load balance or have redundant links in BGP that will be active after primary fails if that's what your trying to achieve

http://www.cisco.com/c/en/us/support/docs/ip/border-gateway-protocol-bgp/13762-40.html

[cisco8887]

Problem I have is , they have /30 which is a point to point so there is no way of doing this unless you use HSRP.

how would I connect both R1 and R2 using a /30? I can't so expectation is only one of the routers will be active at anyone time so BGP will be established to R1 if it is standby active

R2 in meantime will keep trying to establish a connectioni using .4 to .2 and nothing will happen.

You won't even need the ISP involvement to make this work, would you ?

to add to complexity of this my connections are both L2 from my routers to the BGP neighbor

[Mark Malone]

but BGP would do that doesn't matter there p-p most circuits are what matters is there both advertised locally same BGP table both subnets out each pipe , you could have an IBGP link between them as well with 2 EBGP links then if you want to the ISP have an active/standby link setup or go full active/active with BGP , this requires no ISP involvement and you can load balance whatever and as its BGP everything knows about everything same BGP table

[cisco8887]

thanks for the reply. I am not sure if I fully understood.

Are you agreeing with my earlier design ?

[Mark Malone]

Hi not really I dont think you should use HSRP with BGP , it can be used fine on your vlan interfaces to allow physical switch failover on lan side but for ISP redundancy I think you should use BGP as its more flexible but maybe im miss understanding you , do you have screenshot of any type what it looks like the setup

this is what I think you have in place from my head something like below and that's what I was talking about in my last post use IBGP between them then EBGP to the ISP

[cisco8887]

exactly as you said, I am trying to achive physical redundancy with HSRP

see diagram attached. You have S1 and S2 connecting to ISP using VPLS.

Quick questions outside of the topic, the red link is not blocking traffic between the Switches .

One of the VPLS links should be blocking else you will have aloop right ? Do we have STP in VPLS over MPLS?

[Mark Malone]

ok so you using vpls that's a bit different from what I thought that's a layer 2 setup , I have been describing a pure layer 3 setup with BGP

your extending your layer 2 through the mpls cloud as if it wasn't there basically and I don't use vpls but it does send bpdu traffic across the cloud  unless the ISP is blocking it that you want to check with them or check each end in spanning-tree details and see if the bpdus are incrementing

If its pure ;layer 2 end to end bypassing mpls cloud could you not use stp cost on 1 link each end make it less preferred ? that's will give you redundancy at layer 2 , the higher cost should only kick in if the lower cost link fails

There seems to be n issue with Sup2t incase you have them

With a Supervisor Engine 2T, Layer 2 protocol tunneling is not supported with VPLS, which prevents use of the Cisco Discovery Protocol (CDP), the VLAN Trunking Protocol (VTP), and the Spanning-Tree Protocol (STP) over VPLS (CSCue45974).