EIGRP

Advantages and disadvantages
1. Advantages
  1. Fast convergence
  2. multiple network layer support
  3. works on NBMA and other networks without special configuration
  4. sophisticated metric, including support for unequal cost load-balancing
  5. Reliable Transport Protocol - used by EIGRP to carry all packets
  6. Graceful shutdown
    1. notifies neighbors that eigrp is going down to speed convergence on neighbors
    2. hello packet with all K values set to 255
    3. only sent when 'no router eigrp' command is used, not when interfaces go down or the router is reloaded
2. Disadvantages
  1. cisco proprietary - does not work with non-cisco routers
  2. doesn't scale as well as OSPF and ISIS
EIGRP neighbors
1. hello packets
  1. sent out all interfaces configured for EIGRP
  2. multicast address of 224.0.0.10
  3. timers depend on the medium
    1. 5 seconds on LAN (like Ethernet)
    2. 5 seconds on point to point
    3. 5 seconds for multipoint interfaces greater than T1, including ISDN PRI
    4. 60 second intervals of for slower speed links
  4. adjust with ip hello-interval eigrp as-number seconds interface command
2. hold time
  1. included in hello packet
  2. by default = 15 sec (3 times hello interval)
  3. adjust with ip hold-time eigrp as-number seconds interface command
  4. must be manually adjusted after a manual hello time adjustment
  5. adjacencies will still form with different hello and hold times on neighbors
3. Routers that receive hellos establish adjacency
  1. primary interface address used as source address
  2. neighbors must be on same subnet
  3. neighbors must have same metric calculation constants (K-values)
  4. neighbors must be in the same AS
4. The adjacencies remain as long as the router receives any EIGRP packet (including another hello) from the neighbor before the hold timer expires
Routing
1. Each router forwards a copy of its IP routing table to all adjacent neighbors
2. routers store the routes received from its neighbors in the topology table along with the metric advertised by the neighbor and the metric from this router to the advertising one
3. Diffusing Update Algorithm (DUAL) is run on the topology table to select the routes with the best Feasible Distance (FD). The FD is the distance advertised by the neighbor, or Advertised Distance (AD), plus the distance to reach that neighbor. The FD uses several metrics:
  1. bandwidth = smallest bandwidth on the path
  2. delay = cumulative interface delay on the path
  3. reliability = worst reliability on the path, based on keepalives
  4. loading = highest load on any link along the path
  5. MTU - smallest MTU in the path
  6. By default, only bandwidth and delay are used. Reliability and loading can optionally be used by adjusting the K values
    1. metric = K1*bandwidth+K2*bandwidth/256-load+K3*delay and metric=metric*k5/(reliability+k4) when K5 != 0
    2. by default, k1=k3=1, k2=k4=k5=0
    3. these values are carried in hello packets, and must match a neighbors for adjacencies to form
    4. changing the metrics is not recommended without good reason
4. Successors (routes with the best FD) are offered to the routing table - multiple successors can exist if their FD is the same
5. Feasible successors (FS) are back-up paths that qualify to replace the successors if the path to the successor goes down. To qualify, the AD of the path to the Feasible successor has to be less than the FD of the successor. In other words, the metric the neighbor has to the back-up path has to be less than this router has to the primary path. This implies that the back-up path does not travel through this router and prevents routing loops. FS's are kept in the topology table as backups.
6. Triggered updates are sent when routing changes occur
7. If a path to a subnet goes down, the router uses its FS. If a FS is not available, the router sends a query packet to each of its neighbors. If a neighbor doesn't have a route, it sends a query to its neighbors. The route is in active state during this time, and routing information on this route cannot be changed until all replies are received.
Configuring EIGRP
1. basic setup
  1. router eigrp as-number global command
  2. network network [netmask] router command
  3. For serial links, set the bandwidth for metric calculations
  4. For serial links, you can limit the amount of bandwidth eigrp consumes by using the ip bandwidth-percent eigrp as-num percent interface command
2. optional configuration
  1. global command ip default-network network creates a last resort gateway and will also advertise this default route to other eigrp routers. This route must be reachable for the advertisement to propagate.
  2. interface command no ip split-horizon eigrp AS-number to disable split horizon (useful on frame relay)
3. summarization
  1. no auto-summ router config command
  2. ip summary-add eigrp as-numb address mask [admin-distance] interface command
  3. a more specific route must exist in the routing table for the summary to be advertised
  4. note that an eigrp router will only summarize routes it generates with autosummary. It will not summarize a route that is advertised to it by a neighbor
4. Load balancing
  1. by default, up to 4 equal cost paths will be used
  2. use maximum-paths max router command to set it up to 16
  3. process switching occurs per-packet. fast-switching occurs per-destination
  4. variance multiplier router command enables unequal-cost balancing
  5. only feasible paths are eligible for load-balancing to prevent routing loops
5. MD5 authentication
  1. no auth is used by default
  2. ip authentication mode eigrp as md5 interface command
  3. ip authentication key-chain eigrp as name-of-keychain interface command
  4. key chain name-of-chain global command
  5. key key-id key-chain command
  6. key-string key-string key-id command
  7. optionally specify a lifetime for receiving updates with accept-lifetime start-time {infinite | end-time | duration seconds} key-id command
  8. optionally specify lifetime for sending updates with the send-lifetime command
6. Stub router
  1. eigrp stub [connected | summary | static | receive-only] router command
  2. stub routers do not get queried about active routes
  3. stub routers do not re-advertise routes they learn from one neighbor to another neighbor, which prevents them from being used as a transit router
verifying and troubleshooting
1. Stuck-in-Active - when a router doesn't get a reply to all queries within 3 minutes (by default), the route goes SIA (change the time with timers active-time [time-limit | disabled] router config command. time is in minutes)
  1. the router resets adjacency for neighbors that failed to reply
  2. causes of SIA
    1. high CPU
    2. bad physical link
    3. unidirectional link
  3. preventing SIA
    1. 2 new TLV's in 12.1(5) help - SIA-Query and SIA-Reply
    2. limit query range
      - summarization - routers only propagate a query if they have an exact match in their routing table
      - configure a router as a stub to prevent it from being queried
2. tables
  1. show ip eigrp topology all-links - shows all IP entries in the topology table
  2. sh ip eigrp topo - shows only successors and feasible successors
    1. passive routes are stable
    2. active routes are undergoing re-computation - the router is looking for a new successors because the current one has gone down
3. neighbors
  1. eigrp log-neighbor-changes (router config command)
  2. sh ip eigrp nei
  3. sh ip eigrp topology
  4. sh ip eigrp traffic
4. basic troubleshooting
  1. show ip route
  2. sh ip route eigrp
  3. sh ip protocols
  4. sh ip eigrp interfaces
5. Debug commands
  1. debug eigrp packets - displays types of eigrp packets sent and recieved
  2. debug ip eigrp - displays packets sent and received on an interface
  3. debug ip eigrp summary - displays summarized eigrp activity, including filtering and redistribution
  4. debug eigrp neighbors
6. authentication
  1. debug eigrp packets
  2. sh ip eigrp nei
References
1. Cisco Press BSCI Authorized Self-Study Guide
2. Cisco Press BSCI Officaial Exam Certification Guide
3. Oriley.com

Back to Multicast - Up to index - Next to OSPF

Updated:
11 Aug 07

C
C
N
P

B
S
C
I

E
I
G
R
P

[All Guides]
[Home Page]