Using interface bandwidth to influence EIGRP routing decisions

Now we have our wireless bridge in place, we wanted to configure EIGRP to seamlessly converge over that link in the event of a fiber failure between sites.

We currently have 4 cores (2 at each site) with a full mesh connectivity between them. Each link comprises of 2x 1Gbps fibers bundled together into 2Gbps etherchannels.  There is also layer2 connectivity between cores at each site (allowing cores within a site to be EIGRP neighbors.

Despite the quite redundant looking diagram above, this only provides us redundancy from GBIC/SFP/Line card failures. The fact is that all of these fibers go into 2 multicore fibers running in the same duct underground, so we are still quite vulnerable.

The WIFI Bridge has an AP at both sites, each one connected back to a core.

Once we had confirmed IP connectivity between the two sites via the WIFI bridge, we configured EIGRP to work over the links, the neighbor relationships formed, and BAM utilization on the WIFI link shot to 100%. so i grab a few “show” outputs and disabled EIGRP to return traffic back to normal.

What i discovered was that because we use VLAN’s for alot of routing, the minimum bandwidth EIGRP see’s is usually a gig, the below shows a routing table entry for a subnet that is load sharing over two of the fiber links

D       10.21.10.0/24 [90/3072] via 10.0.0.22, 21:04:21, Port-channel2
                    [90/3072] via 10.0.0.18, 21:04:21, Port-channel3


you can see the minimum bandwidth is always 1000000 Kbit, so its the delay that is used to determine which routes end up in the routing table, as some routes are 2 hops, the delay is incremented per hop, as the below “show ip eigrp topology x.x.x.x” shows;

IP-EIGRP (AS 99): Topology entry for 10.21.10.0/24
  State is Passive, Query origin flag is 1, 2 Successor(s), FD is 3072
  Routing Descriptor Blocks:
  10.0.0.22 (Port-channel2), from 10.0.0.22, Send flag is 0x0
      Composite metric is (3072/2816), Route is Internal
      Vector metric:
        Minimum bandwidth is 1000000 Kbit
        Total delay is 20 microseconds
        Reliability is 255/255
        Load is 1/255
        Minimum MTU is 1500
        Hop count is 1
  10.0.0.18 (Port-channel3), from 10.0.0.18, Send flag is 0x0
      Composite metric is (3072/2816), Route is Internal
      Vector metric:
        Minimum bandwidth is 1000000 Kbit
        Total delay is 20 microseconds
        Reliability is 255/255
        Load is 1/255
        Minimum MTU is 1500
        Hop count is 1
  10.1.1.11 (Vlan1001), from 10.1.1.11, Send flag is 0x0
      Composite metric is (3328/3072), Route is Internal
      Vector metric:
        Minimum bandwidth is 1000000 Kbit
        Total delay is 30 microseconds
        Reliability is 255/255
        Load is 1/255
        Minimum MTU is 1500
        Hop count is 2
  10.0.0.1 (Vlan1000), from 10.0.0.1, Send flag is 0x0
      Composite metric is (3328/3072), Route is Internal
      Vector metric:
        Minimum bandwidth is 1000000 Kbit
        Total delay is 30 microseconds
        Reliability is 255/255
        Load is 1/255
        Minimum MTU is 1500
        Hop count is 2

As the Wireless bridges are connected to Gigabit ports, EIGRP sees the bandwidth as 1000000 Kbit, and the delay as 20microseconds, hence being introduced into the routing table and being seen as an equal cost to the fiber links.

There is a very simple way to correct this issue, and that is to use the interface command “bandwidth” to change the perceived bandwidth of the link (note: only perceived bandwidth, as the link will still transfer packets as fast as it can)

interface Vlan900
 bandwidth 10

Setting the bandwith to 10 means that EIGRP sees the link as only 10k, and therefore pretty much guarantees that it will be the absolute last choice link. The below shows all the available paths to the 10.21.10.0/24 network, Po2 & Po3 are the fiber links, Vlan1000 is the EIGRP VLAN between cores in the same site, and VLAN 900 is used for the WIFI Bridge.

P 10.21.10.0/24, 2 successors, FD is 3072, serno 8990
        via 10.0.0.22 (3072/2816), Port-channel2
        via 10.0.0.18 (3072/2816), Port-channel3
        via 10.0.3.1 (256000512/2816), Vlan900
        via 10.1.1.11 (3328/3072), Vlan1001
        via 10.0.0.1 (3328/3072), Vlan1000

As you can see, the metric is very high, meaning that if variance were to be increased to allow unequal cost load balancing there is still a high probability that this route would not be selected.

  10.0.3.1 (Vlan900), from 10.0.3.1, Send flag is 0x0
      Composite metric is (256000512/2816), Route is Internal
      Vector metric:
        Minimum bandwidth is 10 Kbit
        Total delay is 20 microseconds
        Reliability is 255/255
        Load is 1/255
        Minimum MTU is 1500
        Hop count is 1

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: