dgroscost
asked on
EIGRP and specific traffic routing w/ failover.
Running EIGRP.
RouterA
LAN - 10.10.0.0
LAN - 10.20.0.0
Serial (T1) - 190.132.0.0
Serial (Frame) - 190.93.0.0
RouterB
LAN - 10.30.0.0
Serial (T1) - 190.132.0.0
Serial (Frame) - 190.93.0.0
RouterA and RouterB have connectivity via both serial links.
I want to enforce specific traffic patterns without losing the failover benefits from having two links and using EIGRP.
Specifically, I want to route any traffic from RouterA (from 10.30.0.0) destined to RouterB (190.20.0.0) to go over the T1 line (as well as reverse traffic).
I also want traffic from RouterA (from 10.30.0.0) destined to Router B (190.10.0.0) to go over the Frame.
Now, if one of the links should fail (T1 or Frame), I would like these traffic patterns to be temporarily inactive until the downed link returns to normal condition.
How does one accomplish this? I would use route-maps, however, doesn't that prevent EIGRP from allowing failover connectivity from the other working link?
Thanks.
RouterA
LAN - 10.10.0.0
LAN - 10.20.0.0
Serial (T1) - 190.132.0.0
Serial (Frame) - 190.93.0.0
RouterB
LAN - 10.30.0.0
Serial (T1) - 190.132.0.0
Serial (Frame) - 190.93.0.0
RouterA and RouterB have connectivity via both serial links.
I want to enforce specific traffic patterns without losing the failover benefits from having two links and using EIGRP.
Specifically, I want to route any traffic from RouterA (from 10.30.0.0) destined to RouterB (190.20.0.0) to go over the T1 line (as well as reverse traffic).
I also want traffic from RouterA (from 10.30.0.0) destined to Router B (190.10.0.0) to go over the Frame.
Now, if one of the links should fail (T1 or Frame), I would like these traffic patterns to be temporarily inactive until the downed link returns to normal condition.
How does one accomplish this? I would use route-maps, however, doesn't that prevent EIGRP from allowing failover connectivity from the other working link?
Thanks.
I thought route-maps would not forward traffic to a non-existant next-hop or down interface. You're route-maps should work fine, but I'm not sure how you would make them work. How are you going to identify a new connection from A to B versus a response to B from A?
-Eric
-Eric
ASKER
Let me add one more thing - the T1 is a Multilink group. There are two T1 lines bundled together. (Plus the Frame connection on the other interface)
Right now we've got route-maps in place - to force the traffic over the T1 multilink only (both ways). I'm assuming that if the T1s died (multilink downed), the traffic would fail to route anywhere because the route-map would enforce the traffic policy over the T1s and not want to use the Frame -- is that an accurate statement?? Or does the route-map work with the dynamic routing protocol and realize since the link is down it should route over any traffic over the remaining Frame link?
Right now we've got route-maps in place - to force the traffic over the T1 multilink only (both ways). I'm assuming that if the T1s died (multilink downed), the traffic would fail to route anywhere because the route-map would enforce the traffic policy over the T1s and not want to use the Frame -- is that an accurate statement?? Or does the route-map work with the dynamic routing protocol and realize since the link is down it should route over any traffic over the remaining Frame link?
You are correct in that a route-map will not dynamically change the next hop. That's where the saa probe and the track availability comes in. This way you can have alternate next hops, depending on which one is up.
ASKER
OK. Now that config shows examples of using it with Ethernet interfaces and such (and with DSL/Cable)...
Would you recommend using that method in a T1 Multilink group & Frame connectivity?
Would you recommend using that method in a T1 Multilink group & Frame connectivity?
The concept is the same regardless of the physical connection. I've used it successfully in several scenarios, and I think it will work for yours.
I prefer to use CEF for load balancing over creating a multilink group to start with, but that's just personal opinion and another discussion. It's the difference between being fast-switched and process-switched. Which is faster?
I prefer to use CEF for load balancing over creating a multilink group to start with, but that's just personal opinion and another discussion. It's the difference between being fast-switched and process-switched. Which is faster?
Gack. multilink groups are fast swtiched. I ran a 5xE1 bundle this summer and processor utilization stayed around 5% when the bundle was maxed out. This was on a 3745.
-Eric
-Eric
ASKER
If it matters at all, I'm researching this for my 3745 models as well...
I guess fast-switching in a multilink group is in the newer IOS?
It's been over a year since I had to research/play with bundles...
Thanks Eric, as always...
It's been over a year since I had to research/play with bundles...
Thanks Eric, as always...
ASKER
I just noticed that you mentioned (and the link shows) Enterprise Base IOS is needed. I'm running the regular IP IOS right now... what's the difference? Is this only available in enterprise base??
Yes, unfortunately the "track" command is only available to apply to a probe in Enterprise Base. Upgrade from IP is about $900
for the feature set...
for the feature set...
ASKER
I'm able to download the IOS file. Does that still mean I need to upgrade? Are there any losses going with the Enterprise Base set over regular IP IOS?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
http://www.cisco.com/en/US/tech/tk364/tk871/technologies_configuration_example09186a0080211f5c.shtml
However, note: Enterprise Base IOS Feature set is required...