|
Thread Index Re: [isp-bgp] How to properly utilize backup link in multi-homed setup?
Colo Host wrote: You lost me. For some random Internet destination, I would think that your IGP would not have a route for this, only BGP. Did you mean IBGP vs. EBGP? Even that doesn't make sense to me. If two adjacent routers prefer "the other one" for egress, traffic might never leave your network.It is a border router, but there's a mesh and between the core and the border so if the right combination of cables or interfaces were to drop, traffic to the border could flow through router #2 first. I'm going to have to contact Foundry TAC on the route choice thing. We're using Foundry routers and their algorithm is to prefer IGP over EGP, opposite of Cisco's "administrative distance" factor which treats IGP as a much lower preference to EGP. So in my case, once the router determines that both hops are usable, local preference is the same and AS path length is the same, it should then choose IGP according to their docs but seems to be going with EGP instead. First, let me ask this: why a backup link to L3, especially if you're trying NOT to use it? Personally, I'd just aim for N+1 upstreams with reasonably consistent sizes (i.e. fire up several DS3s before you begin moving to any 100M or OC3 links, for easiest load sharing adjustments). Alternatively, I'd consider flip-flopping the GBLX and L3 links to put both L3 links on the same router and GBLX on a different router. You'll be able to make the backup L3 link work the way you want a lot easier (I believe).Do you have a recommendation or best practices on how a provider with multiple links can best utilize their links across multiple border routers? We're pushing a lot of data through router #1 and just brought up #2 with the second Level 3 link for redundancy and fortunately haven't had any outages on the primary Level 3 link or we probably would have seen the other issue you mentioned. :-) But we're bringing up two more upstreams and will be connecting them to router #2 as it has the ports we need on it, so we'll have four total neighbors split between the two routers but with VRRP and static routes at the border, traffic will normally start with router #1 as it's leaving the network and can then take iBGP routes as appropriate. For path length differences, BGP will still hit the shortest AS path provider across border routers so I guess I can just do traffic engineering on the rest as needed but if you have any suggestions that would be much appreciated too. Second, I'd try to grow to a point where your border (let's clarify terms: it seems your "border" is the upstream egress point) routers don't have any customer links on them, and the routers with your customer links/ports are speaking IBGP with your border routers. On the border routers (if they're Cisco), configure a non-zero weight on all routes learned from upstream, so there's less chance of a packet arriving on R1 only to get punted over to R2 (which will make your exit decisions easier). Your customer routers will choose the best upstream/border and send the traffic that way intelligently. If you can only rely on VRRP/OSPF/etc. for a default route, you will probably have big challenges getting your traffic to balance easily unless you revert to a purely hot-potato exit strategy and split your VRRP points, etc. Or at least that's my $0.02... pt To unsubscribe via postal mail, please contact us at: Jupitermedia Corp. Attn: Discussion List Management 475 Park Avenue South New York, NY 10016 Please include the email address which you have been contacted with.
Thread Index |
|
