Administrative distrace

tecnodog7

Been reading up on Administrative distance and this questions is been in my mind

Why would one ever want to set a administrative distance which is not the default one? When running RIP or OSPF, isn't the router smart enough to always use the better one in this case OSPF since it had an AD of 110 so why would i even want to put a something else?

When using route redistribution, occasionally there may be a need to modify the administrative distance of a protocol so that it takes precedence. For example, if you want the router to select RIP-learned routes (default value 120) rather than IGRP-learned routes (default value 100) to the same destination, you must increase the administrative distance for IGRP to 120+, or decrease the administrative distance of RIP to a value less than 100. <<<< Why would you want to do this?

Thanks

gorebrush

Trying to give you a simple explanation - but generally you would change these in events where you are having to redistribute at *multiple* points in a network - you can end up with routing loops/feedback when the routes come back to the second point in the network after being redistributed from the first, because the second router can choose "external" routes, because the AD is "lower"

Codeman6669

someone can correct me here if im wrong,

but you can set the AD just above lets say OSPF that way if the link goes down you have a backup route. Since it will chose the path with the best AD then the next best thing after that first link fails, is the one you configured with the higher AD

gorebrush

That's not really the intended use, but I guess you can.

It's really about managing routes from several sources (i.e. more than one routing protocol in your network)

A better way in OSPF is to use LFA (Loop Free Alternative)

OSPFv2 Loop-Free Alternate Fast Reroute - Cisco

Priston

We use static routes to null0 with an administrative distance of 250 so the addresses allocated to that area of the network show up in the routing table even if they aren't in use. It wasn't my idea and I'm still kind of iffy about the design.

gorebrush

Interesting, so if the route goes down from a dynamic routing protocol the traffic just gets black holed on the nearest router? Not sure I'd like that myself..

Priston

The static routes to null0 are on the GWs. This part of the network is all Layer 2 ( vlans ) except for the uplinks which are BGP.

So if the vlan is up, the vlans have priority over the static routes since the vlans are directly connected with an AD of 0. If the vlan is down, we'll I don't think there will be any traffic to worry about.

I'm not 100% sure how the BGP part works, with the static routes to null0. All I know is if the uplinks go down. I've got bigger issues to worry about

OfWolfAndMan

gorebrush wrote: »

Trying to give you a simple explanation - but generally you would change these in events where you are having to redistribute at *multiple* points in a network - you can end up with routing loops/feedback when the routes come back to the second point in the network after being redistributed from the first, because the second router can choose "external" routes, because the AD is "lower"

This. Also, in a situation where you're migrating routing protocols, a method of moving from OSPF to EIGRP example, is another situation. Raising AD of EIGRP until your topology is ready to move over to the EIGRP routes, and then implementing a moving boundary (redistribution)

gorebrush

OfWolfAndMan wrote: »

This. Also, in a situation where you're migrating routing protocols, a method of moving from OSPF to EIGRP example, is another situation. Raising AD of EIGRP until your topology is ready to move over to the EIGRP routes, and then implementing a moving boundary (redistribution)

I like this.

theodoxa

Chris Bryant's CCNP Videos include a scenario where internal RIP routes would lose [based on AD] to external/redistributed routes, resulting in the redistributed (inferior) route being used instead of the RIP route. He resolved this by altering the AD of one of the protocols so that redistributed routes would only be placed in the routing table if the internal routes went down.

tecnodog7

Oh ok.
Thank you guys.

Would this also work as redistribution of traffic, i.e load balancing technique? What if you have two parallel routes to the same place but you want the traffic redistributed so that it doesn't crash that one side would this work for that?

theodoxa

tecnodog7 wrote: »

Would this also work as redistribution of traffic, i.e load balancing technique? What if you have two parallel routes to the same place but you want the traffic redistributed so that it doesn't crash that one side would this work for that?

That's unnecessary. If you advertise two parallel routes in a routing protocol such as EIGRP, they should have the same AD and metric [assuming the same bandwidth for both links] and be load-balanced by default. EIGRP supports unequal cost load-balancing if the metrics are slightly different.

OfWolfAndMan

I think you're getting the term redistribution mixed up with traffic distribution. Two different things.

If you're asking what I think you're asking though, OSPF and EIGRP both support equal cost multipathing, up to four links I believe. BGP has its own method of multipathing as well. The routing metric must be equal though, whether it be through physical interfaces or logical modification of the metric (link bandwidth, delay, offset lists, etc). As theodoxa said though, EIGRP supports unequal cost load balancing (But is not configured for it by default).

In regards to links crashing, you obviously have never heard of the queue. If the router experiences congestion, unless a queueing feature is configured, packets on the tail end behind the full queue will drop.

tecnodog7

Thank you guys for your feed back and helping me understand. As i read thing my mind forms questions and sometimes even I don't know what I'm asking. Thank you OfWolfAndMan and theodoxa and the rest for helping with this topic

Hondabuff

Here is the best example for adjusting AD of a route. It comes when you setup backup routes for a DR site on a VPN. Both of these routes use different vpn tunnels. One is to the primary firewall and the other is to the DR site incase the primary fails. All I'm doing is changing the AD of the second route so if the primary tunnel drops then all packets will route across the second tunnel. Tunnel 0 goes to primary site and Tunnel 1 goes to the DR site. Most of the time in a production environment your not adjusting OSPF or EIGRP.

R1(config)#ip route 10.0.0.0 255.255.255.0 tunnel 0
R1(config)#ip route 10.0.0.0 255.255.255.0 tunnel 1 100

Heero

AD is a major consideration whenever you are doing redistribution. It can also be used for traffic engineering. Each routing protocol has it's own method of determining the best route within the protocol, but there needs to be some tie breaker rule that applies when you are learning the same route from multiple different protocols. That is the function of AD.

Typically you won't be changing AD in your production environment for any of the routing protocols, but it is fairly common to use floating static routes where the AD of a static route is set higher than the AD of the dynamic protocol you normally learn the route from.