EIGRP configuration is short, but the loop-free logic under it is the point. EIGRP is Cisco’s Enhanced Interior Gateway Routing Protocol, an advanced distance-vector protocol that converges with DUAL, which RFC 7868 says keeps routing loop-free at every instant.
You enable it with router eigrp and an autonomous system number, then add network statements, and the AS number must match on neighbors. For each destination a router keeps a successor, the best next hop, and where one exists a feasible successor, a neighbor already proven loop-free, so a failure switches to the backup instantly instead of recalculating.
- EIGRP is Cisco’s advanced distance-vector protocol, documented in RFC 7868
- It converges using DUAL, which keeps routing loop-free at every instant
- Configuration is router eigrp
then network statements - The AS number must match on neighbors or no adjacency forms
- A feasible successor is a precomputed, loop-free backup path
On this page
ConfigurationConfiguring EIGRP, the two commands
On Cisco IOS the classic EIGRP configuration is deliberately small, and each piece has a consequence worth knowing.
Enable the process. router eigrp <AS> starts EIGRP under an autonomous system number.
router eigrp 100The AS number must match. The number is not just a label. Neighboring routers only form an adjacency if their AS numbers match, so a typo here produces routers that sit next to each other and never exchange a route.
List the networks. One or more network statements tell EIGRP which interfaces to run on and which networks to advertise.
router eigrp 100
network 10.0.0.0
network 192.168.1.0The network statement selects interfaces, not just routes. An interface whose address falls inside a network statement starts sending EIGRP hellos and can form neighbors. This is why an overly broad statement can bring up EIGRP on an interface facing somewhere it should not, which is what a wildcard mask on the statement, or the passive-interface command, is for.
Nothing converges without neighbors. EIGRP builds everything on adjacencies, so the first check after configuration is always the neighbor table, which you view with : no neighbors, no routes, and the usual cause is a mismatched AS number, a network statement that missed the interface, or a Layer 1 or 2 problem underneath.
What it isWhat EIGRP is, and where it sits
EIGRP occupies a middle ground between the two families of routing protocol, which is most of its appeal.
It is advanced distance vector. RFC 7868 describes EIGRP as a routing protocol based on distance-vector technology, using the DUAL algorithm. It shares routes with directly connected neighbors the way a distance-vector protocol does, rather than flooding a full topology the way a link-state protocol like OSPF does.
It behaves like a link-state protocol in the ways that matter. Among interior gateway protocols it forms neighbor adjacencies with hello packets, sends updates only when something changes rather than on a timer, and keeps a topology table of networks with backup paths.
The result converges quickly without the full link-state database every router in an OSPF area has to hold.
It is Cisco's, now openly documented. EIGRP was designed and developed by Cisco Systems, and for most of its life it was proprietary. RFC 7868, published in 2016 as an informational document, describes the protocol design and architecture in full, which is why it can be implemented and understood outside Cisco.
It runs directly on IP. EIGRP is assigned protocol number 88 and is encapsulated directly into the network-layer protocol, IPv4 or IPv6, rather than riding on TCP or UDP. It uses the multicast address 224.0.0.10 for IPv4 and FF02::A for IPv6 to reach EIGRP routers on a segment.
DUALDUAL, and the feasible successor
This is the mechanism that makes EIGRP fast, and the terms are worth getting exactly right because they explain the behavior.
The successor is the best path. For each destination, the successor is the neighboring router on the least-cost path, the one the router forwards to. RFC 7868 defines it as a neighbor with a least-cost path to a destination that is guaranteed not to be part of a routing loop.
Feasible distance and reported distance. The feasible distance is the least-known total metric to a destination since the route last became stable. The reported distance is the metric a neighbor advertises, its own distance to the destination. Both are precise terms in RFC 7868, and the next definition is built from them.
The feasibility condition proves a path is loop-free. RFC 7868 states the source node condition: a neighbor meets the feasibility condition if the neighbor's reported distance is less than this router's feasible distance.
Put plainly, if a neighbor is already closer to the destination than you have ever been, it cannot be routing through you, so using it cannot form a loop.
A feasible successor is a proven backup. A feasible successor is a neighbor that meets the feasibility condition: a downstream path that may not be the best but is guaranteed loop-free. When the successor fails and a feasible successor exists, the router switches to it immediately, with no recalculation, because the backup was already proven safe.
With no feasible successor, the route goes active. When the successor fails and no feasible successor exists, the route enters the active state and the router queries its neighbors to compute a new path, coordinating with query and reply messages.
A route in the active state is unusable until that finishes, which is the slow case EIGRP works hard to avoid.
The metricThe metric, and why bandwidth dominates
EIGRP's metric is often described as complicated; the useful version is short.
A composite metric from K-values. In its default configuration EIGRP computes a composite metric from several inputs, scaled by coefficients called K-values. RFC 7868 gives the defaults as K1 = K3 = 1 and K2 = K4 = K5 = 0, which means that by default only two inputs count: K1 selects on bandwidth and K3 on delay.
Bandwidth and delay, in practice. With the default K-values, EIGRP picks paths on the bandwidth of the slowest link in the path and the cumulative delay along it. Load and reliability are available as inputs but are off by default, because turning them on makes the metric change as traffic changes, which destabilizes routing.
The interface bandwidth value matters. EIGRP uses the bandwidth configured on an interface, which is not always the real speed.
RFC 7868 notes an architect may set an artificially high or low bandwidth to influence routing, and warns that a bandwidth value that does not match the physical link can make EIGRP behave badly, including using too much of the link for its own traffic. Set interface bandwidth to reflect reality unless you are deliberately steering.
Do not change K-values lightly. The K-values must match between neighbors, like the AS number, or no adjacency forms, and displays the values in use, and RFC 7868 says the defaults were chosen for good performance in most networks. Changing them is a decision for a specific, understood reason, applied everywhere at once.
PacketsWhat EIGRP sends, and how reliably
EIGRP's packet types and its transport explain both its speed and one common surprise.
Five packet types. RFC 7868 lists them: hello (which also carries acknowledgements), update, query, reply, and request. Hellos discover and maintain neighbors, updates carry routing information, and query and reply drive the active-state recomputation described above.
Reliable only when it must be. EIGRP has its own reliable transport that guarantees ordered delivery, but RFC 7868 is explicit that reliability is provided only when necessary. A hello on a multicast-capable Ethernet does not need acknowledging by every neighbor individually, while an update does, so EIGRP mixes reliable and unreliable delivery for efficiency.
It caps its own bandwidth use. By default EIGRP limits itself to no more than 50 percent of the bandwidth reported by an interface for its own packet pacing. This is another reason the configured interface bandwidth matters: set it wrong and EIGRP either starves itself or floods a slow link.
Updates are triggered, not periodic. Unlike RIP, which re-sends its table on a timer, EIGRP sends routing information when something changes. A stable network is quiet, which is part of why it scales better than the distance-vector protocols it descends from.
PitfallsWhere people go wrong
Mismatched AS numbers. Two routers with different EIGRP AS numbers never become neighbors. It is the first thing to check when a configured link exchanges no routes.
A network statement that is too broad or too narrow. Too broad brings EIGRP up on an interface it should not touch; too narrow leaves an interface out and its network unadvertised. The neighbor table and the interface list show which happened.
Leaving interface bandwidth wrong. A default or inherited bandwidth value that does not match the real link skews path selection and EIGRP's own pacing. Set it to reflect the link, or set it deliberately to steer, but never leave it by accident.
Changing K-values on one router. They must match on every neighbor. A K-value change on one device drops its adjacencies, and the symptom looks like a link failure rather than a configuration change.
Expecting a feasible successor to always exist. The instant failover only happens when a loop-free backup was already known. Where none exists, the route goes active and convergence takes longer, which is normal, not a fault.
Treating EIGRP as multi-vendor by default. RFC 7868 makes the protocol documentable and implementable, but in practice EIGRP is overwhelmingly a Cisco deployment. A mixed-vendor network usually reaches for OSPF or BGP instead.
ComparisonEIGRP, OSPF and RIP side by side
| Criterion | EIGRP | OSPF | RIP |
|---|---|---|---|
| Family | Advanced distance vector | Link state | Distance vector |
| Algorithm | DUAL | Dijkstra (SPF) | Bellman-Ford |
| Convergence | Fast, with feasible successors | Fast | Slow |
| Updates | Triggered | Triggered | Periodic |
| Metric | Bandwidth and delay | Cost from bandwidth | Hop count |
| Vendor | Mainly Cisco | Multi-vendor | Multi-vendor |
| Holds full topology | Neighbors and backups | Full area database | No |
The convergence and updates rows are why EIGRP and OSPF both replaced RIP; the vendor row is usually what decides between the two.
FAQFrequently asked questions
What is EIGRP?
Cisco's Enhanced Interior Gateway Routing Protocol, an advanced distance-vector routing protocol that converges using the DUAL algorithm. It is documented in RFC 7868 and assigned IP protocol number 88.
How do I configure EIGRP?
Enable the process with router eigrp <AS>, then add network statements for the networks to advertise. The AS number must be the same on all routers that should become neighbors.
Why are my EIGRP routers not becoming neighbors?
The most common cause is a mismatched autonomous system number, since neighbors only form when the AS numbers match. Other causes are a network statement that missed the interface, mismatched K-values, or a Layer 1 or 2 problem on the link.
What is a feasible successor?
A neighbor that meets the feasibility condition, meaning its reported distance is less than the router's feasible distance, so it offers a guaranteed loop-free path. It is a precomputed backup that EIGRP switches to instantly if the successor fails.
What is the difference between a successor and a feasible successor?
The successor is the current least-cost next hop, the one in use. A feasible successor is a backup that is already proven loop-free, ready to take over without a recalculation.
What is the feasibility condition?
A test for loop freedom: a neighbor passes it if the neighbor's reported distance to a destination is less than this router's feasible distance. A neighbor already closer than you have ever been cannot be routing through you.
What is DUAL?
The Diffusing Update Algorithm, the algorithm EIGRP uses to keep routing loop-free at every instant during a route computation. Routers affected by a change coordinate; routers not affected are left out of the recalculation.
What metric does EIGRP use?
A composite metric whose default K-values, K1 = K3 = 1, make it select on bandwidth and delay. Load and reliability can be added but are off by default because they make routing react to traffic.
Is EIGRP a Cisco proprietary protocol?
It was designed by Cisco and was proprietary for most of its history. RFC 7868, published in 2016 as an informational document, describes it in full, so it can now be implemented by others, though it remains mainly a Cisco deployment.
What does the active state mean?
A route is active when its successor has failed and no feasible successor exists, so the router is querying neighbors to compute a new path. The route is unusable until that completes, which is the slow convergence case.
Does EIGRP use TCP or UDP?
Neither. It runs directly on IP as protocol 88, with its own reliable transport that acknowledges only the packets that need it, and multicasts to 224.0.0.10 on IPv4.
When should I choose EIGRP over OSPF?
Mainly when the network is all Cisco and you want fast convergence with simpler configuration than OSPF areas. For a multi-vendor network, OSPF or BGP is the more portable choice.
Keep readingRelated concepts
Read next · Routing OSPF Explained The link-state protocol EIGRP is most often weighed against. Open this next12 min- Routing · 10 min The RIP Protocol, and Why You Still Meet It The older distance-vector protocol EIGRP and OSPF both replaced.
- Routing · 12 min Static vs Dynamic Routing, and the Route That Never Fails When a dynamic protocol like EIGRP is worth running at all.
- Routing · 9 min BFD, the Fast Way to Notice a Link Has Failed A protocol that subscribes to a BFD session instead of its own timers.
- Routing · 9 min Protocol Independent Multicast, and How PIM Builds the Tree The kind of routing table PIM reuses instead of building its own.