Multiple Ethernet WAN Feeds using a Cisco 2811 router

[Tobor1313]

I am in need of some guidance in trying to use two 10 Mbit ethernet feeds using a Cisco 2811 router. The router has two built-in ethernet ports and a HWICD9ESW, 9 port card. I would appreciate either URL, book reference or actual code to allow two ports to act as failover and load balancing for the two 10 Mbit ethernet WAN feeds.

I would appreciate any help that you might proffer.

Thanks.

[Patrick Laidlaw]

Hello,

The only way I can think of for failover using two ethernet ports in the "Same" router would be to bridge them and use a Bridge virtual interface. Example

interface ethernet 0/0

no ip address

bridge group 1

interface ethernet 0/1

no ip address

bridge group 1

interface BVI 1

ip addresss x.x.x.x 255.255.255.0

bridge irb

bridge 1 protocol ieee

Try that if it doesn't work i'll look at it, just doing it from memory.

Patrick

[Tobor1313]

Thank you very much! Your suggestion worked fine for my purposes once I added the line:

bridge 1 route ip

and edited your "bridge group 1" to "bridge-group 1" on the interfaces, all worked as you suggested.

Thanks again for your assistance.

[amit-singh]

Hi,

I addition to the above post, I think we can also achieve the results using a routing protocol that can do the load-balancing. For this, your connection provider should run the same protocol at their end. You can use EIGRP or OSPF and have the interfaces advertised in the routing protocol.

If you ISP cannot run that, you can also use static routes to achieve the above.You can have 2 default routes pointing towards the next hop ip of your ISP.

conf t

ip route 0.0.0.0 0.0.0.0 NEXT_HOP_IP1

ip route 0.0.0.0 0.0.0.0 NEXT_HOP_IP2

HTH,

-amit singh

[Tobor1313]

I received two suggested solutions for this problem and tried the first which worked.

However the bridge solution appears to have a latency when performing failover as there appears a 20-30 second latency when one of the interfaces goes down. I am hoping that your routing solution will not have display this latency. I'll let you know when I have successfully applied your suggestion.

Thanks again for all your help.

[Patrick Laidlaw]

Hi,

The routing solution shouldn't show any delays but sometimes it's very hard to get ISP's to run routing protocols like EIGRP or OSPF just to provide internet access.

Other options that are out there if your just dealing with ethernet, include some firewall products that load balance and do failover over two internet links. One of them off the top of my head is called SonicWall. I'm hopeing that the ASA will be able to do this soon if not already.

Patrick

[Patrick Laidlaw]

Hi,

The routing solution shouldn't show any delays but sometimes it's very hard to get ISP's to run routing protocols like EIGRP or OSPF just to provide internet access.

Other options that are out there if your just dealing with ethernet, include some firewall products that load balance and do failover over two internet links. One of them off the top of my head is called SonicWall. I'm hopeing that the ASA will be able to do this soon if not already.

Patrick

[Tobor1313]

Well, the latency that I speak of occurs whenever one of the ethernet cables are taken offline (i.e. removed). Apparently the router responds by dropping the routing information and re-establishing it on the other bridged port. Thus the eigrp table seems to get reloaded and that is why I think that there appears this latency.

Whenever I remove one of the cables, hopefully simulating a line going down, any file transfers die and this is not what I expect in failover, akin to what happens with our PIX firewall. Evidently the bridging does not setup a stateful mode in failover. I wonder if there is any way to obtain a stateful failover with the 2811 using a bridge group?

Thanks for all your thought on this issue, I greatly appreciate the input.

[clagos]

What are the Ethernet interfaces plugging into? Cable/DSL cpe, fiber mux?

I'm confused how the Bridge-Group example is working. Since you assign an IP address to the group, you can only have one. The separate ethernet links should require two separate addresses.

I would add two static routes to 0.0.0.0, one for each WAN link, and set the floating static weights the same to accomplish the load blancing.

EIGRP convergence is sub-second, there should be virtually no delay for routing table reloads in an enviornment like yours.