How to configure backup WAN links on a 2800????

[ba.mittleman]

Hi,

I currently admin a WAN link between two international sites.  I have a 2800 at each end. The primary connection between the two sites is 20MB Ethernet PIP MPLS BGP circuit.  I also have a pair of T1's in a PPP multilink that I switch over to in the event the primary circuit goes down.  I currently am doing the switch over manually.  It takes me about 15 minutes to complete it, but I know there must be a way to make this an automatic process.

So a couple questions:

1) Is there a way to create an "enhanced multilink" with the 2 T1's and the Ethernet circuit in the same bundle, in effect giving me 23MBps bandwidth while still being able to maintain connectivity if one of the 3 links goes down?  This is my preferred solution

and if 1 is not possible then

2) Is there a way to configure the router so if the Ethernet circuit goes down, all traffic will be automatically directed over the T1's, and then come back to the Ethernet when it's back online?

Many thanks.

[cadet alain]

Hi,

as far as option 2 is concerned and if I understand correctly what you want then over the ethernet you receive the prefix via iBGP with an AD of 200 and if on your MPP you configure  a static route with  an AD of 201 for same prefix then when the route is known by BGP it wins and when the BGP adjacency fails or the route is no more propagated by BGP then the static route will get installed in the routing table. even if it was eBGP then the AD concept with the static route would still hold true.

Regards.

Alain.

[ba.mittleman]

Hi Alain,

I added an AD of 250 on the T1 routes, will test tonight.  I think that will do it though.

Would still like to be able to bundle them all thogether though.

Thanks

[Vaibhava Varma · Accepted solution]

Hi Bruce

As Alain has suggested the perfect way to achive the Active:Stanby mode of operation for your two links is manipulating the AD and will help you out.

Unfortunately there is no way to bundle different media type links together. But you can still use them as an aggregate BW by Routing Manipulation . What we need here is to have equal cost routing across both the links to do the default per-destination load-balancing.

But this would certainly enforce some major changes to your existing routing design across both links which will help you to attain an aggregate BW of 23 Mbps from traffic flow perspective but not from bundling perspective.

Hope this helps you.

Regards

Varma

[ba.mittleman]

Hi,

I've configured the backup T1 multililink as stated using Administrative Distance.  We had the primary Ethernet WAN link go down today.  The switch over to the T1 did occur, but users were kicked out of applications that communicated over the WAN when the switch over occurred, probably because the route was suddenly changed.

It was a bad scenario because the primary kept coming up and going back down every 15 minutes today, each time knocking everyone off of our database and disconnecting VOIP calls.

I have since disabled the primary Ethernet circuit and am keeping everyone onto the backup T1's until V* can troubleshoot their Ethernet circuit.

Is there any way to prevent the disconnects from applications when the route changes because the primary Ethernet goes down and the backup T1 kicks in?

Many thanks.

[Vaibhava Varma]

Hi Bruce

Good that we are able to reduce the 15 mins outage time to a couple of mins owing to the routing table convergence.

Since we are switching over from a BGP Learned route to Static Route, the Routing Table update is subject to detection of BGP Peering Failure (BGP Support for next-hop tracking subject to your IOS Support+ Removal of BGP Learned route from RT owing to BGP Scan Interval of 60 seconds and then installing the Static Route.

Using the default BGP Timers there would be delay of somewhere around 90 seconds in this process of switchover assuming T0 as 0 when the primary ethernet goes down.

My suggestion would be if BFD is supported between the CE and PE we can try using BFD for BGP as well as BFD for Still. Still I am not sure for much faster switchover as the withdrawl of the BGP route from RT would depend on the BGP Scan Time of 60 seconds and we might need to look at fine tuning the BGP scan interval to a lower value but again it will impact the c2800 CPU much.

Hope this helps you on this issue.

Regards

Varma