cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
4498
Views
15
Helpful
10
Replies

route-map with multiple next hops

Collin Clark
VIP Alumni
VIP Alumni

In a route map with multiple set ip next-hop statements, are each of the hops load balanced?

route-map LB permit 10

match ip address 10

set ip next-hop 1.1.1.1 2.2.2.2

10 Replies 10

Edison Ortiz
Hall of Fame
Hall of Fame

Nope, it tries 1.1.1.1 first and only uses 2.2.2.2 if 1.1.1.1 is down.

http://www.cisco.com/en/US/docs/ios/iproute/command/reference/irp_pi2.html#wp1012397

HTH,

__

Edison.

Jon Marshall
Hall of Fame
Hall of Fame

Collin

Not as far as i am aware. The first hop is used unless it is unavailable and then the next hop is used.

Edit - Collin, Edison is a faster typer than me !!

Jon

I should try it before posting, it would have answered my question.

*Jul 2 18:41:04.315: IP: tableid=0, s=1.1.1.1 (local), d=192.168.1.10 (Loopback0), routed via RIB

*Jul 2 18:41:04.315: IP: s=1.1.1.1 (local), d=192.168.1.10 (Loopback0), len 100, sending

*Jul 2 18:41:04.315: IP: tableid=0, s=1.1.1.1 (Loopback0), d=192.168.1.10 (Loopback1), routed via RIB

*Jul 2 18:41:04.315: IP: s=1.1.1.1 (Loopback0), d=192.168.1.10, len 100, rcvd 4.

*Jul 2 18:41:06.315: IP: tableid=0, s=1.1.1.1 (local), d=192.168.1.10 (Loopback0), routed via RIB

*Jul 2 18:41:06.315: IP: s=1.1.1.1 (local), d=192.168.1.10 (Loopback0), len 100, sending

*Jul 2 18:41:06.315: IP: tableid=0, s=1.1.1.1 (Loopback0), d=192.168.1.10 (Loopback1), routed via RIB

*Jul 2 18:41:06.315: IP: s=1.1.1.1 (Loopback0), d=192.168.1.10, len 100, rcvd 4.

Since you guys are here I have a strange design request. I have a hub router and two spokes connected via MPLS and using BGP. The hub router sends the default route to the spokes and there is no IGP at the spokes. Next week the spokes will be connected to each other with a couple of point-to-point T1's. They now want a specific server at a spoke site to load-balance out across BOTH spoke connections. They only way I can think of is with an iBGP session (they have the same AS#) between the spokes, but that's a mess. Any ideas?

Collin

If there is no IGP at the spokes does that mean that you only have one L3 device at each spoke and that this L3 device is responsible for routing the server subnet ?

Also the P2P connection - will they be connected to the same L3 devices in each spoke site ?

Jon

Yes to both, a single device at each location that terminates both the MPLS and P2P T1's.

The hub router sends the default route to the spokes and there is no IGP at the spokes.

How the hub knows to get to the spokes? Static routes for those networks? Or that's the job for this route-map?

Next week the spokes will be connected to each other with a couple of point-to-point T1's.

Any plans for IGP between the spokes?

They now want a specific server at a spoke site to load-balance out across BOTH spoke connections.

Fair enough, but the spoke need to be able to connect to the other spoke via some kind of IGP.

Ideally, run an IGP between the 3 locations and within the IGP you can alter the metrics for a specific route/host so it can be seen as equal cost hence performing some kind of load-balance.

Keep in mind, you can load-balance out of the spokes by altering the metric from the source (server) - but you need to keep in mind, the other side of the traffic can be asymmetrical (the server responding to the destination won't be load-balanced).

What's prompting this business requirement?

Edison-

We advertise the local subnet at each spoke into BGP. At the hub we redistribute BGP into EIGRP.

It's open on using an IGP or static routes between the sites via the T1's

Unfortunately I can not use an IGP between the sites, we must use BGP. The business requirement is that backups are taking to long across the WAN and they want to utilize both circuits. There is no technical staff at either spoke. I was thinking of using EIGRP, but was unsure on how to load balance across two default gateways and multiple locations.

Collin

The problem you have is that both WAN connections terminate on the same spoke router.

S1 = spoke where server is

S2 = other spoke

At the moment S1 router has a default-route received via EBGP. Even if you had an IBGP session between the spoke sites the default-route that is received directly from EBGP would be preferred.

You could introduce EIGRP but that just gets even more complicated because now you are looking at redsitributing BGP into EIGRP and still you have the same issue with paths being chosen altho as Edison says the metrics could be tweaked.

Or on S1 spoke router

ip route

ip route

where destination subnet is the backup server subnet - you could even just use a host route.

Not pretty and as Edison points out it only sorts out traffic from the server.

It's not clear if the T1 is going to be there simply because of this problem but i doubt it. If it is tho you would be much better off increasing S1 bandwidth to MPLS - but i'm sure you know this.

Edit - i'm assuming that the server is backing up to the main site so most traffic will be going from the spokes to the hub.

Jon

The business requirement is that backups are taking to long across the WAN and they want to utilize both circuits.

Can you explain how this backup takes place?

Is the hub running some kind of backup program and it's trying to backup data from that spoke server?

You want to maximize its backup throughput?

I'm sure there are other ways to maximize backup throughput than what's proposed. You should look into Cisco WAAS for instance.

__

Edison.

I agree there are other solutions but of course it's the networks job to address this. That being said I believe you guys have answered my questions thoroughly. Thanks to both of you.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco