7-2 Interior Gateway Routing Protocol (IGRP)

• IGRP is a classful routing protocol.

• IGRP is a distance vector routing protocol that computes a metric from a combination of delay, bandwidth, reliability, load, and mtu (although mtu is not used in the computation).

• Redistribution to and from EIGRP is automatic, provided that the autonomous system number is identical for both IGRP and EIGRP on the router.

• IGRP advertises three route types: interior (routes between subnets in an AS), system (routes to networks within an AS: directly connected networks, and system routes from other IGRP routers), and exterior (routes to networks outside an AS).

• IGRP uses a gateway of last resort chosen from the list of exterior IGRP routes.

• Updates are broadcast every 90 seconds. Routes are declared inaccessible if an update is not received within three update periods, and they are removed from the routing table if no update occurs in seven update periods.

• Flash updates are sent before the regular 90-second update if there is a metric change.

• Poison reverse updates are sent to place a route in hold-down state if large metric changes are detected . This prevents routing loops .

NOTE

IGRP broadcasts its updates to neighbors using IP protocol 9. If specific neighbors are defined, their unicast IP addresses are used instead.

Configuration

1. Enable the IGRP routing process:

    (global)  router igrp   autonomous-system  

   The autonomous-system (AS) is a number that associates IGRP routers with the same routing domain. Routers running IGRP with the same AS number can exchange routes.

2. Associate the network with the IGRP AS:

    (router)  network   network-number  

   Updates are sent out these interfaces, and these interfaces are advertised. (Network numbers are reduced to classful networks in the configuration.)

   NOTE

   If you are transitioning to EIGRP, run both IGRP and EIGRP on the transition routers. The AS or process number must be the same in both IGRP and EIGRP to automatically redistribute routes. As soon as redistribution is complete, IGRP can be turned off.

3. (Optional) Define specific unicast neighbors:

    (router)  neighbor   ip-address  

   Unicast updates can be sent to IGRP neighbors on nonbroadcast networks.

4. (Optional) Specify unequal -cost load balancing:

    (router)  variance   multiplier  

   The default variance is 1 (equal-cost load balancing). The multiplier argument specifies the limit a route metric can vary from the lowest -cost metric and still be included in unequal-cost load balancing. Load balancing can exist over up to four paths to a destination.

5. (Optional) Adjust metric weights:

    (router)  metric weights   tos k1 k2 k3 k4 k5  

   The default is a 32-bit metricthe sum of segment delays and lowest segment bandwidth (scaled and inverted). For homogeneous networks, this reduces to hop count. Refer to Section 8-3 for more information about the metric computation and weight values.

   NOTE

   The bandwidth command can be used to alter the IGRP route metric. However, this affects both IGRP and OSPF, because both use the bandwidth value for metric computation. To affect only IGRP, modify the metric using the delay interface configuration command.

6. (Optional) Adjust IGRP timers:

    (router)  timers basic   update invalid holddown flush  [  sleeptime  ] 

   Various timer values can be adjusted: update (the number of seconds between updates; the default is 90), invalid (the number of seconds after a route is declared invalid and enters holddown; the default is 270), holddown (the number of seconds that a route is marked inaccessible; this should be at least three times the update period; the default is 280), flush (the number of seconds until a route is removed from the routing table; this must be at least the sum of invalid and holddown; the default is 630), sleeptime (the number of milliseconds to postpone an update after a Flash update; this should be less than update; the default is 0 milliseconds).

7. (Optional) Disable holddown state:

    (router)  no metric holddown  

   Holddown is used to prevent routing loops from forming. This mechanism can be disabled. Use caution. All IGRP routers in an AS must have a consistent use of holddown.

8. (Optional) Disable split horizon:

    (interface)  no ip split-horizon  

   If this is enabled, updates and queries are not sent to destinations for which the interface is the next hop. Split horizon is enabled by default (except for Frame Relay and SMDS networks).

9. See the following sections for information on route-processing features:

   • 8-3: Redistributing Routing Information

   • 8-4: Filtering Routing Information

Example

An IGRP routing process is configured to provide routing information for both the 10.0.0.0 and 192.168.6.0 networks. The router advertises and listens for IGRP updates on the 10.0.0.0 network, but it only listens (and doesn't advertise) on the 192.168.6.0 network. Figure 7-2 shows a network diagram.

Figure 7-2. Network Diagram for the IGRP Example

  interface ethernet 0   ip address 10.0.0.1 255.0.0.0   interface serial 0   ip address 192.168.6.2 255.255.255.0   router igrp 101   network 10.0.0.0   network 192.168.6.0   passive-interface ethernet 0