OSPF Not load balancing

[Brandon Baker]

We currently have 3 ISPs, the third just being added last week, When I set it up under OSPF in our Core switch it will not load balance, Our current defrault route in the core is going to ISP2 and when I add our new connection in as a default route it stelas all of the traffic from from ISP2 instead of load balancing. Can anyone please assist?

All 3 have same Costs as well

[Brandon Baker]

A little more info that may help, here is the config for each interface going to the 6509:

interface GigabitEthernet0/2 (VXR)
description

ip address X.20.128.1 255.255.255.128
no ip redirects
no ip proxy-arp
ip route-cache same-interface
ip route-cache flow
duplex auto
speed auto
media-type rj45
no negotiation auto
ipv6 address X:480:9::1/64
no cdp enable
hold-queue 100 in

interface GigabitEthernet0/0/0 (ASR Original)
ip address X.20.128.2 255.255.255.128
ip flow ingress
ip ospf 1 area 0.0.0.1
negotiation auto
ipv6 address X:480:9::2/64

interface GigabitEthernet0/0/5
description Xto 6509
ip address X.20.128.139 255.255.255.248
ip flow ingress
ip ospf priority 0
ip ospf 1 area 0.0.0.1
negotiation auto

[Richard Burts]

It seems that my comment about using BGP for outside and OSPF for inside has confused things a bit. So let me try again. I would classify the ASR, the VXR, and the 6509 as providing routing for outside of your own network. As such it is appropriate to run BGP as the primary routing protocol. I would classify any other routers or perhaps layer 3 switches as provide routing services inside your own network. As such it is appropriate to run OSPF as your inside routing protocol. BGP is very good at providing routing to and from the Internet. OSPF is very good at providing routing to and from devices inside your network.

Let me also say a little about EBGP (External BGP) and IBGP (Internal BGP). EBGP is when your router is in its own AS number and it has neighbors which have different AS numbers. So your ASR and your VXR are running EBGP and they have external neighbors. IBGP is when your router is in its own AS number and its neighbors are also within that AS number. So your 6509 is running IBGP (with ASR and VXR as its internal neighbors). The 6509 does not need to have a neighbor relationship with either ISP 1, ISP 2, or ISP 3 and will learn the routes that they advertise from its internal neighbors.

Based on what I believe that you have told me I do not see any need to have static default routes on the 6509. And having them probably interferes with the routing that you are getting via BGP. And I really do not see any need for static default routes on the ASR either.

I would like to try to get back to your original question which was about sending traffic through ISP 2. Your statement was that adding ISP 3 had changed things and had stolen the traffic that used to go through ISP 2. I can see that adding the static default routes might have changed things if the default route for the 6509 previously had been a default route learned from ISP 2. But in one of your posts you showed that there are numerous route entries for prefixes which are routed through ISP 2. So even if the default route changed there should have been a considerable amount of traffic going through ISP 2 anyway.

It is a common way to set up BGP with multiple ISP to learn only a default route from one and to learn more specific routes for a number of prefixes from another. That sounds like what your network was set up to do when there was ISP 1 and ISP 2. Now that there is also ISP 3 someone needs to figure out what is a reasonable strategy for making effective use of all 3 ISP.

HTH

Rick

[Brandon Baker]

Yes thank you. I believe I may have gone a little off topic. I do have traffic passing through isp 3 but my issue is that it is getting there through the interface 128.2. I am trying to activate another gigabitethernet interface so that more than 1gb can get through. I am assuming it would be the same case even if isp 3 did not exist and I had a connection over 1 gb. Is it possible to basically just split the traffic traveling through 128.2 and have the other half travel through the new interface 128.139?

[Richard Burts]

The reason that this traffic is going through 128.2 is that the 6509 has a BGP neighbor relationship with the ASR and both ISP 1 and ISP 3 are reached through that neighbor relationship. So activating another interface is not necessarily going to improve the amount of traffic being forwarded.

One option that you might consider would be to configure Policy Based Routing on the 6509. Using PBR you could forward certain types of traffic (perhaps something like web/HTTP/HTTPS) over the interface for 128.239. Once this traffic gets to the ASR then the ASR can try to split some traffic over ISP 1 and some over ISP 2. I have implemented this type of approach and can say that it can work successfully. But I will also note that running BGP using 2 external router and 3 external ISP, and putting PBR on top of that is a complicated thing to accomplish.

HTH

Rick

[Brandon Baker]

Thank you for the info. I did remove the static router from the 6509 this morning and everyhting still operated as normal. It did change the gateway of last resort to the Neighbor Ip address of 129.1 (ISP2). The thing that I am still trying to figure out is why in the 6509 when I do a show ip bgp it lists routes to the AS of ISP1 and ISP2 but nothing for ISP3. I do temporarily have traffic going through the new connection now, but not in a way that I wanted so still need to do some research. Basically everyhting originally was left alone and I have added a second connection to our CMTS and have it going into vlan3 with the new connection to the ASR so now everything that comes out of the cmts on the new connection is also using the new connection for the ASR. Would it be possible to use a channel group to combine both of the vlans and let the channel group load balance for me or is that a major no no in this scenario?

[Richard Burts]

Thanks for the update. It is good to know that things still worked after you removed the static default route.

As for why you do not see on the 6509 any routes from ISP 3, in an earlier post you seem to indicate that ISP 3 was advertising only a default route (or maybe I did not correctly understand that part of the post). So there is not much to see from ISP 3.

The thought of using a channel group to combine capacity of 2 physical interfaces seems like a good idea. Ether Channel is supported on ASR 1006 as far as I know and it seems that it would give you the greater capacity while still having a single layer 3 connection.

HTH

Rick