>_nexthopreachable

CCNP ENCOR: EIGRP Metrics, Feasible Successors, and Variance

Ali Bahadır Coşkun24 min read

EIGRP is the protocol I've spent the least time with. I learned the basics for my CCNA, and then I moved to Juniper, where EIGRP doesn't exist. So it never came up in any of my five Juniper certifications. When I returned to EIGRP for CCNP ENCOR, I could remember only bits and pieces.

Enhanced Interior Gateway Routing Protocol (EIGRP) is an advanced distance-vector routing protocol. Routers discover neighbors by sending hello packets to 224.0.0.10 using IP protocol 88. When an adjacency forms, they exchange topology information; after that, they send partial updates only when something changes. EIGRP uses the Diffusing Update Algorithm (DUAL) to select the best path and identify loop-free backup paths called feasible successors. If the current path fails and a feasible successor is available, EIGRP can switch to it immediately. EIGRP was originally Cisco proprietary, but Cisco documented the protocol in informational RFC 7868 in 2016. It is still supported and remains in use in Cisco-centric networks, but it is less commonly chosen for new multi-vendor environments, where standards-based protocols such as OSPF are more portable.

In this lab, we'll start with classic EIGRP on three routers, add named-mode EIGRP on the other five, then read the metric and the topology table, break a link to watch DUAL work, and finish with unequal-cost load balancing using variance. Every output in this article comes from the lab itself.

Topology

The topology has eight IOSv routers. R4 is the HQ router where we'll watch the routes, and R8 is the data center router that owns 8.8.8.8/32. R4 has three ways to reach it: through R2, R3, or R6.

The topology ended up looking like a ship, so I’m calling it the EIGRP ship.

Every router has a Loopback0 that matches its number, so R1 is 1.1.1.1/32 and R8 is 8.8.8.8/32.

Every physical interface in IOSv is Gigabit Ethernet, so I adjusted the bandwidth and delay on five links to make them look like Fast Ethernet and Serial links:

LinkSubnetLink typeInterface settings on both ends
R1–R210.0.12.0/24GigabitIOSv default
R1–R510.1.15.0/24GigabitIOSv default
R2–R410.0.24.0/24GigabitIOSv default
R3–R410.0.34.0/24Fast Ethernet-likebandwidth 100000, delay 10
R3–R510.1.35.0/24GigabitIOSv default
R4–R610.2.46.0/24Serial-likebandwidth 64, delay 2000
R5–R810.1.58.0/24Serial-likebandwidth 64, delay 2000
R6–R710.2.67.0/24Fast Ethernet-likebandwidth 100000, delay 10
R7–R810.0.78.0/24Fast Ethernet-likebandwidth 100000, delay 10

For example, R4's link to R6 is configured like this:

ios
interface GigabitEthernet0/2 bandwidth 64 ip address 10.2.46.4 255.255.255.0 delay 2000

Task 1: Configure Classic EIGRP

Objective: Configure classic EIGRP in autonomous system 100 on R1, R2, and R5. Advertise each router's loopback and its connected inter-router links.

Configuration

On R5:

ios
configure terminalrouter eigrp 100 network 5.5.5.5 0.0.0.0 network 10.1.15.0 0.0.0.255 network 10.1.35.0 0.0.0.255 network 10.1.58.0 0.0.0.255end

R1 and R2 follow the same pattern:

Routernetwork statements
R11.1.1.1 0.0.0.0, 10.0.12.0 0.0.0.255, 10.1.15.0 0.0.0.255
R22.2.2.2 0.0.0.0, 10.0.12.0 0.0.0.255, 10.0.24.0 0.0.0.255

The 100 in router eigrp 100 identifies the EIGRP autonomous system. Unlike a BGP autonomous system number, it is locally significant and does not need to be globally unique. However, routers must use the same EIGRP autonomous system number to become neighbors.

The network command determines which local interfaces participate in EIGRP; it is not a list of routes to advertise. IOS compares each interface address against the address and wildcard mask in the command. When an interface matches, EIGRP is enabled on it, allowing the router to send hello packets, form neighbor adjacencies, and advertise the connected subnet. In a wildcard mask, a 0 means the corresponding bit must match, while a 1 means it is ignored. Therefore, 0.0.0.0 matches one exact interface address, while 0.0.0.255 can match any interface address within the specified /24.

Verification

On R1:

text
R1# show ip eigrp neighborsEIGRP-IPv4 Neighbors for AS(100)H   Address                 Interface              Hold Uptime   SRTT   RTO  Q  Seq                                                   (sec)         (ms)       Cnt Num1   10.1.15.5               Gi0/1                    14 00:00:05    9   100  0  50   10.0.12.2               Gi0/0                    14 00:00:06    6   100  0  7

R1 sees both R2 and R5. R2 and R5 only see R1 for now, because their other neighbors don't run EIGRP yet.

text
R1# show ip route eigrp      2.0.0.0/32 is subnetted, 1 subnetsD        2.2.2.2 [90/130816] via 10.0.12.2, 00:00:07, GigabitEthernet0/0      5.0.0.0/32 is subnetted, 1 subnetsD        5.5.5.5 [90/130816] via 10.1.15.5, 00:00:06, GigabitEthernet0/1      10.0.0.0/8 is variably subnetted, 7 subnets, 2 masksD        10.0.24.0/24 [90/3072] via 10.0.12.2, 00:00:07, GigabitEthernet0/0D        10.1.35.0/24 [90/3072] via 10.1.15.5, 00:00:06, GigabitEthernet0/1D        10.1.58.0/24           [90/40512256] via 10.1.15.5, 00:00:06, GigabitEthernet0/1

EIGRP internal routes show up with code D and an administrative distance of 90.

10.1.58.0/24 has a much bigger metric than the others because it sits behind R5's 64 kbps link. With the default settings, the classic EIGRP metric is:

text
metric = 256 × (10,000,000 / lowest bandwidth in kbps + total delay in tens of µs)

For 10.1.58.0/24, the lowest bandwidth on the path is 64 kbps, and the delay is 10 µs on R1 Gi0/1 plus 20,000 µs on R5 Gi0/2:

text
256 × (10,000,000 / 64 + 20,010 / 10)256 × (156,250 + 2,001) = 40,512,256

The delay that counts is the delay of each outgoing interface along the way, ending with the interface that connects the destination subnet.

show ip protocols shows the K-values and a few other details we'll come back to:

text
R1# show ip protocolsRouting Protocol is "eigrp 100"  EIGRP-IPv4 Protocol for AS(100)    Metric weight K1=1, K2=0, K3=1, K4=0, K5=0    Router-ID: 1.1.1.1    Topology : 0 (base)      Active Timer: 3 min      Distance: internal 90 external 170      Maximum path: 4      Maximum hopcount 100      Maximum metric variance 1

The K-values are weights in the metric formula. I really like Kevin Wallace’s mnemonic, “Big Dogs Really Like Me,” for remembering the values EIGRP tracks: bandwidth, delay, reliability, load, and MTU. With the defaults, K1 (bandwidth) and K3 (delay) are 1 and the rest are 0, so only bandwidth and delay are used in the metric. MTU is recorded but never included in the calculation. The K-values must match between neighbors, or they will not form an adjacency.

Task 2: Configure Named EIGRP

Objective: Configure named EIGRP on R3, R4, R6, R7, and R8 using the IPv4 address family in AS 100, and confirm that both configuration styles form one routing domain.

Configuration

On R4:

ios
configure terminalrouter eigrp ENCOR address-family ipv4 unicast autonomous-system 100  network 4.4.4.4 0.0.0.0  network 10.0.24.0 0.0.0.255  network 10.0.34.0 0.0.0.255  network 10.2.46.0 0.0.0.255end

The other four routers get the same structure with their own loopback and links:

Routernetwork statements
R33.3.3.3 0.0.0.0, 10.0.34.0 0.0.0.255, 10.1.35.0 0.0.0.255
R66.6.6.6 0.0.0.0, 10.2.46.0 0.0.0.255, 10.2.67.0 0.0.0.255
R77.7.7.7 0.0.0.0, 10.2.67.0 0.0.0.255, 10.0.78.0 0.0.0.255
R88.8.8.8 0.0.0.0, 10.1.58.0 0.0.0.255, 10.0.78.0 0.0.0.255

Named mode changes how EIGRP is configured, not how it operates on the wire. The router eigrp ENCOR command creates a named configuration container, while the address-family ipv4 unicast autonomous-system 100 command creates the actual IPv4 routing instance. Interface-specific settings such as authentication, timers, and passive interfaces go under af-interface, while settings that affect route processing and path selection go under topology base. This keeps the entire EIGRP configuration in one hierarchy instead of spreading it across router and interface configuration modes.

The name ENCOR is only a local label and does not need to match on neighboring routers. The autonomous system number, K-values, and other adjacency parameters still need to be compatible. Classic and named mode use the same EIGRP packets, so R2 and R4 can become neighbors even though their configurations use different CLI structures.

Verification

text
R4# show ip eigrp neighborsEIGRP-IPv4 VR(ENCOR) Address-Family Neighbors for AS(100)H   Address                 Interface              Hold Uptime   SRTT   RTO  Q  Seq                                                   (sec)         (ms)       Cnt Num2   10.2.46.6               Gi0/2                    10 00:00:06  119  2340  0  81   10.0.34.3               Gi0/1                    10 00:00:06 1961  5000  1  50   10.0.24.2               Gi0/0                    13 00:00:07   20   120  0  18

R4 has three neighbors: R2 (classic), R3, and R6. VR(ENCOR) in the header tells us this is a named-mode process.

show ip protocols shows a few things that classic mode didn't:

text
R4# show ip protocolsRouting Protocol is "eigrp 100"  EIGRP-IPv4 VR(ENCOR) Address-Family Protocol for AS(100)    Metric weight K1=1, K2=0, K3=1, K4=0, K5=0 K6=0    Metric rib-scale 128    Metric version 64bit    Router-ID: 4.4.4.4    Topology : 0 (base)      Active Timer: 3 min      Distance: internal 90 external 170      Maximum path: 4      Maximum hopcount 100      Maximum metric variance 1

Named mode uses 64-bit "wide" metrics. The classic 32-bit metric was designed when 10 Gbps links were science fiction, and it can't tell the difference between very fast interfaces. Wide metrics fix that by measuring delay in picoseconds. Since the routing table can't store a 64-bit value, the wide metric is divided by the rib-scale of 128 before it's installed. We'll see why that matters in a moment.

Now the route to the data center:

text
R4# show ip route 8.8.8.8Routing entry for 8.8.8.8/32  Known via "eigrp 100", distance 90, metric 90256000, type internal  Redistributing via eigrp 100  Last update from 10.0.24.2 on GigabitEthernet0/0, 00:00:05 ago  Routing Descriptor Blocks:  * 10.0.24.2, from 10.0.24.2, 00:00:05 ago, via GigabitEthernet0/0      Route metric is 90256000, traffic share count is 1      Total delay is 20032 microseconds, minimum bandwidth is 64 Kbit      Reliability 255/255, minimum MTU 1500 bytes      Loading 1/255, Hops 4
text
R4# ping 8.8.8.8 source Loopback0Sending 5, 100-byte ICMP Echos to 8.8.8.8, timeout is 2 seconds:Packet sent with a source address of 4.4.4.4!!!!!Success rate is 100 percent (5/5), round-trip min/avg/max = 9/10/12 ms

Everything works, but look at the next hop. R4 sends traffic to R2, and that path is four hops long. The paths through R3 and R6 are only three hops each.

Task 3: Read the Metric and the Topology Table

Objective: Find out why R4 picked the four-hop path, then identify the successor, the feasible distance, each neighbor's reported distance, and any feasible successors for 8.8.8.8/32.

Bandwidth and delay

First, let's check R4's three interfaces:

text
R4# show interfaces GigabitEthernet0/0 | include BW|DLY  MTU 1500 bytes, BW 1000000 Kbit/sec, DLY 10 usec,R4# show interfaces GigabitEthernet0/1 | include BW|DLY  MTU 1500 bytes, BW 100000 Kbit/sec, DLY 100 usec,R4# show interfaces GigabitEthernet0/2 | include BW|DLY  MTU 1500 bytes, BW 64 Kbit/sec, DLY 20000 usec,

Gi0/0 toward R2 has the IOSv defaults. Gi0/1 toward R3 shows the Fast Ethernet-like values, and Gi0/2 toward R6 the Serial-like values. Notice that delay 10 is displayed as DLY 100 usec.

Why the four-hop path wins

EIGRP doesn't count hops. With the default K-values, it only cares about two things: the lowest bandwidth anywhere on the path and the sum of the delays along it.

Every one of R4's three paths crosses a 64 kbps link somewhere, so the bandwidth part of the metric is identical for all of them. The only thing left to compare is the total delay:

The R2 path has one more hop, but those extra hops are cheap Gigabit links with 10 µs each. The R3 path saves a hop but pays 100 µs on the Fast Ethernet-like link. The R6 path pays for two Fast Ethernet-like links on top of its Serial-like link.

The extra 1.25 µs on every path comes from R8's Loopback0. show interfaces Loopback0 says DLY 5000 usec, but named EIGRP counts a loopback as 1,250,000 picoseconds. A classic router uses the full 5,000 µs, which is why R1's loopback has a much higher metric than R3's when you look at them from R4.

The named-mode metric uses the same idea as the classic formula, with a bigger multiplier:

text
metric = 65,536 × (10,000,000 / lowest bandwidth in kbps + total delay in µs)
PathHopsLowest bandwidthTotal delayWide metric
via R2464 kbps20,031.25 µs65,536 × (156,250 + 20,031.25) = 11,552,768,000
via R3364 kbps20,111.25 µs65,536 × (156,250 + 20,111.25) = 11,558,010,880
via R6364 kbps20,201.25 µs65,536 × (156,250 + 20,201.25) = 11,563,909,120

Divide the winning metric by the rib-scale of 128, and we get the 90256000 from show ip route.

The same thing happens on a smaller scale with R3's loopback. R3 is directly connected to R4, but R4 doesn't use that link:

text
R4# show ip route 3.3.3.3Routing entry for 3.3.3.3/32  Known via "eigrp 100", distance 90, metric 26240, type internal  Routing Descriptor Blocks:  * 10.0.24.2, from 10.0.24.2, 00:00:21 ago, via GigabitEthernet0/0      Route metric is 26240, traffic share count is 1      Total delay is 42 microseconds, minimum bandwidth is 1000000 Kbit      Loading 1/255, Hops 4

R4 reaches 3.3.3.3 through R2, R1, and R5, four hops away, instead of using the one-hop link to R3. The direct link is only 100 Mbps, and the bandwidth part of the metric alone (65,536 × 100 = 6,553,600) is already larger than the entire four-hop Gigabit path (3,358,720).

The topology table

The routing table only shows the winner. The topology table shows everything R4 knows:

text
R4# show ip eigrp topology 8.8.8.8/32EIGRP-IPv4 VR(ENCOR) Topology Entry for AS(100)/ID(4.4.4.4) for 8.8.8.8/32  State is Passive, Query origin flag is 1, 1 Successor(s), FD is 11552768000, RIB is 90256000  Descriptor Blocks:  10.0.24.2 (GigabitEthernet0/0), from 10.0.24.2, Send flag is 0x0      Composite metric is (11552768000/11552112640), route is Internal      Vector metric:        Minimum bandwidth is 64 Kbit        Total delay is 20031250000 picoseconds        Reliability is 255/255        Load is 1/255        Minimum MTU is 1500        Hop count is 4        Originating router is 8.8.8.8  10.0.34.3 (GigabitEthernet0/1), from 10.0.34.3, Send flag is 0x0      Composite metric is (11558010880/11551457280), route is Internal      Vector metric:        Minimum bandwidth is 64 Kbit        Total delay is 20111250000 picoseconds        Reliability is 255/255        Load is 1/255        Minimum MTU is 1500        Hop count is 3        Originating router is 8.8.8.8  10.2.46.6 (GigabitEthernet0/2), from 10.2.46.6, Send flag is 0x0      Composite metric is (11563909120/19742720), route is Internal      Vector metric:        Minimum bandwidth is 64 Kbit        Total delay is 20201250000 picoseconds        Reliability is 255/255        Load is 1/255        Minimum MTU is 1500        Hop count is 3        Originating router is 8.8.8.8

Each Composite metric is (X/Y) pair has two numbers:

  • X is R4's total metric to the destination through that neighbor.
  • Y is the reported distance (RD), the neighbor's own metric to the destination, as advertised to R4.

The best total metric becomes the feasible distance (FD): 11,552,768,000 through R2. The neighbor that provides it is the successor, and it's the one installed in the routing table.

The other two paths are not just "worse routes". DUAL checks each of them against the feasibility condition: a neighbor's RD must be strictly lower than the local FD. A neighbor that passes is a feasible successor, a backup path that's guaranteed to be loop-free.

NeighborReported distanceRD < FD?Role
R211,552,112,640YesSuccessor
R311,551,457,280YesFeasible successor
R619,742,720YesFeasible successor

The logic behind it: if a neighbor is closer to the destination than I am, its path can't be going through me. Otherwise its distance would include mine and would be higher. So I can use it without creating a loop.

R6's RD is tiny because R6 reaches R8 through R7 over two Fast Ethernet-like links. What makes the R6 path expensive is R4's own 64 kbps link to R6, and that cost is only added on R4's side.

When a path is not feasible

All three paths to 8.8.8.8 passed. To find paths that don't, use all-links. The regular topology table only lists successors and feasible successors, while all-links also shows the paths that failed the check:

text
R4# show ip eigrp topology all-links | section 7.7.7.7|1.1.1.1P 7.7.7.7/32, 1 successors, FD is 11557355520, serno 20        via 10.2.46.6 (11557355520/13189120), GigabitEthernet0/2        via 10.0.34.3 (11564564480/11558010880), GigabitEthernet0/1P 1.1.1.1/32, 1 successors, FD is 329646080, serno 5        via 10.0.24.2 (329646080/328990720), GigabitEthernet0/0        via 10.0.34.3 (342097920/329646080), GigabitEthernet0/1

For 7.7.7.7, R3 reports 11,558,010,880, which is higher than R4's FD of 11,557,355,520. R3 isn't a feasible successor.

For 1.1.1.1, R3 reports 329,646,080, exactly the same as R4's FD. The condition says strictly lower, so R3 fails here too. R3's path to R1 goes through R5 and doesn't actually loop through R4. It just happens to have the same delay and bandwidth as R4's own path through R2. DUAL can't prove that from the numbers alone, though, so it doesn't take the risk.

These two paths only appear with all-links. The plain show ip eigrp topology lists a single path for both prefixes.

Classic routers show different numbers

R2 runs classic EIGRP, so it shows the same destination with 32-bit metrics:

text
R2# show ip eigrp topology 8.8.8.8/32EIGRP-IPv4 Topology Entry for AS(100)/ID(2.2.2.2) for 8.8.8.8/32  State is Passive, Query origin flag is 1, 1 Successor(s), FD is 40512544  Descriptor Blocks:  10.0.12.1 (GigabitEthernet0/0), from 10.0.12.1, Send flag is 0x0      Composite metric is (40512544/40512288), route is Internal      Vector metric:        Minimum bandwidth is 64 Kbit        Total delay is 20021 microseconds        Reliability is 255/255        Load is 1/255        Minimum MTU is 1500        Hop count is 3        Originating router is 8.8.8.8  ...

On a classic router, the topology metric and the routing table metric are the same number. On R4, the routing table shows the topology metric divided by 128. When comparing FD and RD, always use the numbers from the same topology output, and don't mix them with what you see in show ip route.

Task 4: Fail the Successor

Objective: Shut R4's link to R2 and compare what happens to a route that has a feasible successor with one that doesn't.

On R4:

ios
configure terminalinterface GigabitEthernet0/0 shutdownend

The route to 8.8.8.8 moves to R3 immediately:

text
R4# show ip route 8.8.8.8Routing entry for 8.8.8.8/32  Known via "eigrp 100", distance 90, metric 90296960, type internal  Last update from 10.0.34.3 on GigabitEthernet0/1, 00:00:01 ago  Routing Descriptor Blocks:  * 10.0.34.3, from 10.0.34.3, 00:00:01 ago, via GigabitEthernet0/1      Route metric is 90296960, traffic share count is 1      Total delay is 20112 microseconds, minimum bandwidth is 64 Kbit      Loading 1/255, Hops 3

The EIGRP event log shows how it got there. The log is printed newest first, so read it from the bottom up:

text
R4# show eigrp address-family ipv4 100 events...174  08:18:40.752 Route installed: 8.8.8.8/32 10.0.34.3175  08:18:40.751 Route installing: 8.8.8.8/32 10.0.34.3176  08:18:40.751 RDB delete: 8.8.8.8/32 10.0.24.2177  08:18:40.751 FC sat rdbmet/succmet: metric(11558010880) metric(11551457280)178  08:18:40.751 FC sat nh/ndbmet: 10.0.34.3 metric(11552768000)179  08:18:40.751 Find FS: 8.8.8.8/32 metric(11552768000)

R4 looked for a feasible successor (Find FS), found that R3 satisfies the feasibility condition (FC sat), deleted the R2 path, and installed R3. All of this happened within the same millisecond, without sending a single packet to another router. The route stayed passive the whole time.

These are the lines for 1.1.1.1/32 from the same log, with unrelated lines removed:

text
59   08:18:40.855 Route installed: 1.1.1.1/32 10.0.34.363   08:18:40.854 RDB delete: 1.1.1.1/32 10.0.24.264   08:18:40.853 Find FS: 1.1.1.1/32 metric(Infinity)69   08:18:40.853 Rcv reply dest/nh: 1.1.1.1/32 10.2.46.6109  08:18:40.798 Rcv reply dest/nh: 1.1.1.1/32 10.0.34.3125  08:18:40.755 Metric set: 1.1.1.1/32 metric(Infinity)126  08:18:40.755 Active net/peers: 1.1.1.1/32 2127  08:18:40.755 FC not sat Dmin/met: metric(342097920) metric(329646080)128  08:18:40.755 Find FS: 1.1.1.1/32 metric(329646080)

This time, the feasibility condition is not satisfied (FC not sat), so R4 moves the route into the active state and queries its two remaining neighbors. R3 replies after about 43 ms and R6 after about 100 ms. Only after both replies arrive does R4 install the new path through R3.

The 100 ms convergence time is insignificant in this eight-router lab, but the same process can take much longer in a large network. When EIGRP loses its successor and has no feasible successor, it marks the route as active and queries its neighbors for another loop-free path. A neighbor that cannot answer immediately queries its own neighbors and waits for their replies before responding. This diffusing calculation can continue across multiple routers, and the original router must wait for every query to return. If the process does not complete before the active timer expires the route is declared stuck-in-active. A feasible successor avoids the query process because the router already has a verified loop-free backup path it can install immediately.

After the test, bring the link back:

ios
configure terminalinterface GigabitEthernet0/0 no shutdownend

R2 becomes R4's neighbor again and the successor for 8.8.8.8 once more.

Task 5: Unequal-Cost Load Balancing with Variance

Objective: Configure variance 2 on R4 so that feasible successors can be installed alongside the successor, and verify the result with the routing table and traceroute.

Configuration

On R4:

ios
configure terminalrouter eigrp ENCOR address-family ipv4 unicast autonomous-system 100  topology base   variance 2end

In named mode, variance is configured under topology base; in classic mode, it goes directly under router eigrp 100.

EIGRP first selects the path with the lowest metric as the successor. With the default variance of 1, it installs only equal-cost paths, up to the configured maximum of four. The variance command expands the range of metrics that can be installed, but it does not change any metric or bypass EIGRP’s loop-prevention rules.

An alternate path must pass two separate tests:

  1. Feasibility: Its reported distance must be lower than the local feasible distance (RD < FD). This proves that the neighbor is closer to the destination and that the path is loop-free.

  2. Variance: Its total metric must fall within the best path’s metric multiplied by the configured variance.

For example, if the successor’s metric is 100 and the variance is 2, a feasible path with a metric of 150 can be installed, while one with a metric of 220 cannot. A path that fails the feasibility condition is never installed, regardless of how high the variance is set. Once the paths enter the routing table, the forwarding process shares traffic between them, giving more traffic to the path with the lower metric.

Verification

text
R4# show ip protocols | include variance      Maximum metric variance 2
text
R4# show ip route 8.8.8.8Routing entry for 8.8.8.8/32  Known via "eigrp 100", distance 90, metric 90256000, type internal  Redistributing via eigrp 100  Last update from 10.2.46.6 on GigabitEthernet0/2, 00:00:04 ago  Routing Descriptor Blocks:    10.2.46.6, from 10.2.46.6, 00:00:04 ago, via GigabitEthernet0/2      Route metric is 90343040, traffic share count is 1      Total delay is 20202 microseconds, minimum bandwidth is 64 Kbit      Reliability 255/255, minimum MTU 1500 bytes      Loading 1/255, Hops 3    10.0.34.3, from 10.0.34.3, 00:00:04 ago, via GigabitEthernet0/1      Route metric is 90296960, traffic share count is 1      Total delay is 20112 microseconds, minimum bandwidth is 64 Kbit      Reliability 255/255, minimum MTU 1500 bytes      Loading 1/255, Hops 3  * 10.0.24.2, from 10.0.24.2, 00:00:04 ago, via GigabitEthernet0/0      Route metric is 90256000, traffic share count is 1      Total delay is 20032 microseconds, minimum bandwidth is 64 Kbit      Reliability 255/255, minimum MTU 1500 bytes      Loading 1/255, Hops 4

R4 now has three next hops for 8.8.8.8. The topology entry agrees:

text
R4# show ip eigrp topology 8.8.8.8/32 | include State  State is Passive, Query origin flag is 1, 3 Successor(s), FD is 11552768000, RIB is 90256000

Variance didn't discover any new paths. R3 and R6 were already in the topology table in Task 3, and the feasibility check had already been done. Variance only allowed those existing feasible successors into the routing table.

Every path has a traffic share count of 1. EIGRP can share traffic in proportion to each path's metric, but these three metrics are within 0.1% of each other, so they all get the same share.

Traceroute shows all three paths:

text
R4# traceroute 8.8.8.8 source Loopback0 numericTracing the route to 8.8.8.8VRF info: (vrf in name/id, vrf out name/id)  1 10.0.24.2 8 msec    10.0.34.3 9 msec    10.2.46.6 4 msec  2 10.0.12.1 9 msec    10.1.35.5 7 msec    10.2.67.7 5 msec  3 10.1.15.5 8 msec    10.1.58.8 6 msec    10.0.78.8 7 msec

Each of the three probes per hop took a different path: R2 → R1 → R5, R3 → R5 → R8, and R6 → R7 → R8. The trace stops at hop 3 because two of the probes already reached R8, so we never see the fourth hop of the R2 path.

Variance still requires the feasibility condition

Variance only works with feasible successors. Two other prefixes in this lab show both conditions failing:

text
R4# show ip eigrp topology | section 7.7.7.7|3.3.3.3P 7.7.7.7/32, 1 successors, FD is 11557355520        via 10.2.46.6 (11557355520/13189120), GigabitEthernet0/2P 3.3.3.3/32, 1 successors, FD is 3358720        via 10.0.24.2 (3358720/2703360), GigabitEthernet0/0        via 10.0.34.3 (13189120/163840), GigabitEthernet0/1
DestinationAlternateFeasible successor?Metric / FDInstalled with variance 2?
8.8.8.8R3Yes1.0005Yes
8.8.8.8R6Yes1.001Yes
7.7.7.7R3No, RD is higher than FD1.0006No
3.3.3.3R3Yes3.93No

The R3 path to 7.7.7.7 is only 0.06% worse than the successor, which would easily fit within a variance of 2. Since it fails the feasibility condition, it never even makes it into the regular topology table. Variance can't install it.

The R3 path to 3.3.3.3 is a valid feasible successor, but its metric is almost four times the FD because of the 100 Mbps link. It would need variance 4 to be installed.

Both routes keep a single next hop:

text
R4# show ip route 7.7.7.7 | include via|metric  Known via "eigrp 100", distance 90, metric 90291840, type internal  * 10.2.46.6, from 10.2.46.6, 00:00:15 ago, via GigabitEthernet0/2      Route metric is 90291840, traffic share count is 1R4# show ip route 3.3.3.3 | include via|metric  Known via "eigrp 100", distance 90, metric 26240, type internal  * 10.0.24.2, from 10.0.24.2, 00:00:16 ago, via GigabitEthernet0/0      Route metric is 26240, traffic share count is 1