Load Sharing with BGP from Single Homed Router to Single ISP...

[boodox]

Hi Everyone,

I'm trying to troubleshoot a dual homed internet connection to a single ISP and I've hit a bit of brick wall with it's load balancing/sharing.

There are two internet circuits on a single router out to the internet. The two circuits terminated at different POP and had the same costing and AS out to the ISP so they both load shared without too much of a problem.

Recently, one of the two circuits was moved over to a new POP which had a worse metric and subsequently, the other of the circuits became the preferred path out to the internet and the changed one sat there dumbly doing very little.

To resolve this, I had to remove the INFERIOR_PATCH route map defined, add separate route-maps for each POP and change the metrics so that both circuits would be utilised again and this brought the "dead" link back to life.

The only problem is now it doesn't seem to be sharing the traffic the way if did prior to moving the connection to the new POP as one of them gets hit more than the other. I believe the fact that the moved connection goes through a POP with the worse metric is what is causing the problem but I'm not sure what I can do to correct it.

Given the opportunity, I would multi home the connections between TWO routers running HSRP but it's something I've inherited so and I've been told that I don't have that option.

Does anyone have any ideas how I can get these to load share properly again? I'd appreciate any help.

Many thanks!

bdx

 

[boodox]

Sorry. Correction to the subject:

Load Sharing with BGP from DUAL Homed Router to Single ISP...

 

[nyy1023]

You sound like you understand what you are doing. My question is, will HSRP give you want you want? I would think the biggest advantage of HSRP is to allow redundancy not really Load Sharing.
In my experience messing with the metrics is the only way to really manipulate the traffic, unless you did some type of QoS and give priority to certain traffic.
Other than what you have already done, and short of calling the ISP and asking for a different POP you have done what you can. But, I am still learning, so perhaps there is another solution.

 

[KiscoKid]

HI

can you confirm if your issue is with inbound (i.e. into your LAN) and/or outbound (into the ISP) load-balancing is affected.

The environment you've just described is one of the more simpler ways of achieving load-balancing outbound.

Normally the ISP should advertise the same set of routes to you with the same MED, i.e. you will see a bunch of routes with a metric (MED) of 0. This is the default MED.

Once you have multiple routes with the same metric, all you need is the 'maximum-paths 2' command under router bgp to facilitate outbound load-balancing.

If the ISP is sending you a set of routes from one router with a MED worse than 1, I would question them as to why they are doing this. They've clearly manually defined some routing policy within their AS to set a poorer metric for those routes and thus it can be easily overcome. Don't forget eBGP doesn't automatically make up a metric based on cost, bandwidth etc that most IGPs do. It's the BGP AS administrators that directly influence what a route metric should be.

If your issue is with inbound load-balancing etc, I wouldn't have expected a change of PoP with routes of worse metric to change this behaviour. This would only modify the behaviour of outbound load-balancing.

To influence inbound load-balancing, you also could try modifying the metric for your LAN subnets (assuming you have multiple LAN subnets). Another popular method is to employ AS prepending but again this works better with multiple LAN subnets.

Let me know how your load-balancing has been affected and, if you need any further assistance of config help, drop a line here.

Hope this helps

 

[boodox]

Thanks CiscoKid/nyy1023

Sorry for getting back on them so late. I've been caught up with various other things recently.

On the question of HSRP, at the moment, the two internet circuits reside on a single router (something I inherited) so if we lost that router, we lose both internet connections. The HSRP is only really desired from a contingency point of view if the router fails which is what we don't have and I'm not allowed to do (for some reason)

Originally both circuits were going to different POP's on the same ISP which by some fluke had the same costing and metrics out to the internet so the maximum-paths worked well enough. When one moved to the new/worse POP, that's when the problems started and have to employ the route maps and metrics to at least balance the outbound.

The way the routing is setup, we seem to "load-balancing" outbound but it's the inbound we don't seem to be getting balanced. We do have multiple subnets but from a business continuity point of view, we can't separate the LANS.

I have spoken to the ISP (both circuits with the same ISP - again something I personally would not have as part of effective design) and they are planning to do to some AS administration on their side and I'm just now waiting for them to do this.

To be quite honest, they haven't been the most responsive with dealing with this problem (seeing as all of this started when we were "forced" to change POP because the old POP building closure) so failing this resolution, I may use this an excuse to re-design the ISP connectivity. That will take time to implement but if I could get this running in the interim, that's help.

I'd appreciate if anyone had any other ideas.

Thanks again for your help.
bdx