Routing Protocol EIGRP

EIGRP
Dmitry Figol
CCIE R&S #53592
dmitry@dmfigol.me
August, 2016
Updated: July, 2017

Out of scope
• EIGRP PE-CE for MPLS L3VPN
• EIGRP Over The Top (OTP)
• EIGRP Stub Site (IWAN)
• EIGRP Multi Topology Routing (MTR) in Named mode
2

Quick facts
• Distance-vector routing protocol
• Classless
• Supports unequal cost load-balancing
• Can provide sub-second convergence
• Partial and bounded updates
• Uses Diffusing Update Algorithm (DUAL)
• Arbitrary route summarization
3

Transport
• Uses multicast address 224.0.0.10
• Uses Reliable Transport Protocol (RTP) for transport – IP Protocol
Number 88
• RTP can provide reliable delivery like TCP using sequence numbers
• EIGRP uses both multicast and unicast packets
4

Packets
• Hello
• ACK - hello with no data
• Update
• Query
• Reply
• SIA query/reply
• Goodbye - hello with K values set to 255
5

Timers
• Hello timer
(config-if)# ip hello-interval eigrp as seconds
Default is 5 seconds (on multipoint) or 60 seconds (on p2p)
• Hold timer
(config-if)# ip hold-interval eigrp as seconds
Default is 15 seconds (on multipoint)or 180 seconds (on p2p)
• Active timer
(config-router)# timers active-time minutes
Default is 3 minutes
• SIA retransmit timer
Default is 90 seconds
6

Adjacency
To establish adjacency the following parameters should match:
• AS number
• K-values
• Common subnet
• Authentication type/password
Automatic neighbor discovery is configured using network command.
Network command enables EIGRP on all interfaces where IP addresses match IP + wild card mask
AND advertises the original prefix from the interface
(сonfig)# router eigrp as
(config-router)# network ip wc-mask
Manual neighbor can be configured using:
(config-router)# neighbor ip int-name
Must be configured on both neighbors. The command also turns off sending and receiving multicast
hello
Verification: # show ip eigrp neighbors [detail]
7

Passive interface
You can stop processing and sending any EIGRP packets on the interface
using passive-interface feature:
(config-router)# passive-interface [default | int-name]
Manual neighbors will also stop working on this link, unlike in RIP
8

Administrative distance
• Internal – 90
• External – 170
• Summary – 5
AD for internal and external routes can be changed in general using:
(config-router)# distance eigrp int-ad ext-ad
AD for internal routes can be changed per prefix/neighbor using:
(config-router)# distance ad neighbor-ip wc-mask acl
AD for external routes can’t be changed per prefix
AD for summary routes can be changed per prefix using:
(config-router)# summary-metric prefix/length distance ad
9

Composite metric
EIGRP calculates composite metric using the formula:
where
delay is taken is microseconds, bandwidth – in kbps, load is the highest along the
path, reliability is the lowest. If K5 = 0, the last term equals “1”
Note: load/reliabilityare not updated live, the current values are sent with a new update
Note: composite metric is never sent to the neighbor, only individual components
10

EIGRP terms and definitions
• Computed Distance (CD) – composite metric of the whole path
• Advertised Distance (AD) or Reported Distance (RD) – composite metric of the
best path from neighbor’s perspective
• Feasible Distance (FD) – the lowest value of CD of the best path since the last
transition from Active to Passive
Note: It does not always equal CD of the best path
• Feasible Successor (FS) – the path that meets Feasibility Condition (FC),
guaranteed to be loop-free by DUAL
• Feasibility Condition (FC): RD of the candidate path < FD
• Successor (S) – one of FS with the lowest CD
11

Loop prevention
• Feasibility Condition
• Split horizon – technically not needed as FC already guarantees loop-free path
• Hop count
(config-router)# metric maximum-hops 1..255 ! default is 100
• Router ID for external routes
12

Operation
• EIGRP puts all received routes into Topology Table
• We can then run local computation and find FS/S
• The best path (or several) are then presented to RIB
• Converged routes are marked as Passive
• If there is reconvergence event, it is possible that route will go to Active state
• Verification:
# show ip eigrp topology [all-links]
# show ip eigrp topology prefix/len
13

Scenario – rules and task
Rules:
• If some EIGRP setting is not mentioned, assume default.
• K3=1, K1=K2=K4=K5=0
• Delay is written above each link. Please note that you don’t need to divide microseconds by 10 and then multiply by 256, you can
add metrics blindly just for the sake of this exercise
• All questions are asked from perspective of R1 towards destination prefix6.6.6.6/32
Task:
1. Which router is Successor?
2. What is the value of FD?
3. Fill in the table with ✓ or ✗ and write down the number of ticks
# ✓ = ?
15
path
via ..
in topology
table?
is FS/S?(yesor
no)
R2
R3
R4
R5

Scenario – answers
1. Which router is Successor? R3
2. What is the value of FD? 110
3. Fill in the table with ✓ or ✗ and write down the number of ticks
Pitfall: R4 will choose the path to the destination via R1 due to 115 < 120. Because the best path is via
R1, R4 will send an update with infinite metric to R1 due to split-horizon on R4 link to R1
16
path
via ..
in topology
table?
is FS/S?(yesor
no)
R2 ✓ ✗
R3 ✓ ✓
R4 ✗ ✗
R5 ✓ ✓
# ✓ = 5

Unequal cost load balancing
• EIGRP supports unequal cost load balancing
• For the path to be eligible for load balancing, the path must be FS
• Also the metric of the path must follow this inequality:
CD of FS <= CD of S x Variance
• Variance is configured using:
(config-router)# variance number ! default is 1
• Number of maximum paths for load balancing is configured using:
(config-router)# maximum-paths number ! default is 4, maximum is 32
17

Reconvergence
If we lose Successor, two scenarios are possible:
1. If there is no FS:
• The route goes to Active state
• Router sends QUERYto all neighbors
• During QUERY the route is frozen in RIB/topology table
• Local computation of FS/S is done after we receive REPLY for all queries
2. If there is FS:
• FS with the lowest CD becomes Successor*
• The route stays passive
• Results in sub-secondconvergence
*Note: If we lose successorand the path with the lowest CD is not in FS list, the route goes
to Active state regardless if we have other FS or not.
18

Reconvergence – Query/Reply
• Query checks if neighbors have FS/S
• Query also informs neighbors about the lost path (poisons with infinite
metric) and they remove this path from the topology table
• Conditions to send a REPLY to a received QUERY:
• If we have a Successor, reply with the metric of the Successor
• If the route is already in Active state, reply with infinite metric
• If this route is NOT in the topology table, reply with infinite metric
• Otherwise the route goes into ACTIVE state, queries are sent to all non-stub neighbors
except of the neighbor from which QUERY was received. Upon receiving all replies,
local computation is performed (FS/S is selected) and REPLY to the original QUERY is
sent.
• Stub router can send query to another stub router
19

Reconvergence - Stuck in Active
• If all replies are not heard within Active timer (3 minutes) the route will go to
SIA state and the neighbor from which we didn’t receive reply will be declared
down
• This behavior was changed with introduction of new packet types: SIA
query/reply
• After half of Active timer (SIA retransmit timer - 90 seconds), SIA query is sent
to check the status of the query/neighbor and Active timer is reset
• Neighbor must respond with Reply (if the route is not in topology table or if
Successor has been found) or with SIA reply (if the route is still in Active)
• If Reply or SIA reply is not heard, adjacency will be torn down
• If Reply is not heard after three SIA query attempts, adjacency will be torn
down
20

Reconvergence (cont.)
• Reconvergence is the function of a query domain: more routers have to be
queried, slower the process is. Also, more routers increase the chance of SIA.
• To speed up convergence we should limit the query domain using:
• Stub feature
• Summarization
• Filtering
21

Path manipulation
• The metric can be influenced by changing individual metric components
• Recommended to use delay to change the metric as it is the only cumulative
parameter
• Changing the bandwidth does not always change the metric
• Changing the bandwidth can also influence other features like QoS
22

Path manipulation – offset-list
You can add specific value to the composite metric inbound or outbound per
prefix using:
(config-router)# offset-list [0 | acl] [in | out] [int-name]
Note: offset-list does “reverse engineering” in order to find a delay which will
cause required change in composite metric. It means, that offset-list affects only
cumulative delay. If K3 = 0, offset-list will not work
Note: with offset-list you can increase metric to such a high value, that the route
will not be installed in RIB (232)
23

Filtering – distribute-list
• You can filter an update inbound or outbound using distribute-list:
(config-router)# distribute-list [route-map map | prefix prefix-list [gateway
prefix-list] | acl] [in | out] [int-name]
• You can specify allowed list of routes and neighbors using gateway keyword
with prefix-list.
• You can also use extended ACL for ”per-neighbor” filtering:
• Source field in ACL is responsible for the source of the route (neighbor’s IP address)
• Destination field in ACL is responsible for the prefix itself (subnet mask is not matched)
24

Filtering – tagging
• Using route-map in redistribution or in distribute-list you can tag internal and
external routes and then filter them based on tag
• Tags are 32-bit integer and can be represented in dotted-decimal format using:
(config)# route-tag notation dotted-decimal
• You can also match tag using wildcard:
(config)# route-tag list tag-list permit ip wildcard
(config)# route-map map permit 10
(config-route-map)# match tag list tag-list
• Refer to the following article for more information EIGRP Route Tag
Enhancements
25

Summarization
• You can summarize networks on any router in EIGRP domain using:
(config-if)# ip summary-address eigrp as prefix subnet-mask [leak-map route-map]
• By default specifics are suppressed. You can unsuppress them using leak-map
keyword. By doing so you can achieve traffic engineering
• EIGRP looks for the lowest metric from specific routes and uses it as a metric
for the summary
• Discard route (via Null0 with AD 5) is automatically created locally to prevent
routing loops
• You can set maximum AD 255 using summary-metric command, which will
remove discard route from RIB and as the result neither summary nor specific
routes will be sent to the neighbor
26

Stub router
• You can mark non-transit routers as “stub”, so queries are not sent to them.
• Configured using:
(config-router)# eigrp stub [connected | summary | redistributed | static |
leak-map route-map | receive-only]
• Default is connected + summary
• An argument indicates which routes a stub router will send to its neighbors
• Using leak-map keyword you can leak any route in topology table
• Actually, queries are sent to stub routers by another stub routers.
27

Verification
# show ip eigrp interfaces
# show ip eigrp neighbors [detail]
# show ip eigrp topology [all-links]
# show ip eigrp topology prefix/len
# show ip eigrp events
# show eigrp protocols
# debug eigrp packet
# debug eigrp fsm
28

Authentication
• Classic mode supports only MD5 authentication, which is configured using:
(config-if)# ip authentication mode eigrp as-num md5
• You also need to specify key-chain for authentication:
(config-if)# ip authentication key-chain eigrp as-num key-chain-name
• If needed, you can configure automatic key rotation using accept-lifetime
and send-lifetimeunder key chain configuration
29

Named mode
• The main benefit of named mode is that the entire EIGRP configuration is
located in a single place
• Activated by:
(config)# router eigrp process-name
30

Named mode – Address family
• To enter address family configuration:
(config-router)# address-family AFI SAFI [vrf vrf-name] autonomous-
system as-num
For example:
(config-router)# address-family ipv4 unicast autonomous-system 100
• This is where any configuration specific to EIGRP process itself is applied, like
network and neighbor commands
31

Named mode – Address family commands
32

Named mode – AF-interface
• To enter af-interface configuration:
(config-router-af)# af-interface if-name | default
• This is where interface specific configuration is applied, like passive-
interface or summary-address
33

Named mode – AF-interface commands
34

Named mode – AF topology
• This section is related to Multi Topology Routing (MTR), which gives an ability
to create different logical topologies over the same physical topology.
• MTR is not supported in 15.x M/T code train
• If MTR is not used, relevant commands should be entered under topology
base
• To enter topology base configuration:
(config-router-af)# topology base
• This is where commands like redistribute, distribute-list, offset-list,
variance are applied
35

Named mode – AF topology commands
36

Named mode – Exclusive features
• Wide metrics
• HMAC-SHA authentication
• Add-path
• Disabling EIGRP on specific interface
• Default interface settings (af-interface default)
• Unique IPv6 behavior
• Default taggingall internal and external routes
• Over the Top (OTP) – not covered in this presentation
• Stub site (IWAN) – not covered in this presentation
37

Named mode – Wide metrics
38

Named mode – Wide metrics (cont.)
• After calculation the metric is scaled down by EIGRP_RIB_SCALE variable,
which is 128 by default. This is done because RIB supports only 32-bit values
for metric
• EIGRP_RIB_SCALE variable can be changed using:
(config-router-af)# metric rib-scale 1..255
39

Named mode – Wide metrics (cont.)
• Wide metric introduces new K coefficient – K6, which controls two new metric
components: jitter and energy. However, you can’t configure them in current
IOS train
• Type of the metric (wide/normal) is NOT sent, so neighbor does not actually
know how metric components were calculated previously, so we can
potentially have inconsistent path selection if both wide and normal metrics are
used in the network because of different delay calculation
40

Named mode – HMAC-SHA authentication
• Named mode supports HMAC-SHA authentication which is stronger than MD5.
Also it provides anti-replay protection
• HMAC-SHA authentication can be enabled on the link using:
(config-router-af-interface)# authentication mode hmac-sha-256 password
Note: the neighbor must use named mode as well to form an adjacency
41

Named mode – Add-path
• Add-path feature allows to send several best equal-cost routes to neighbors
• Note: Useful in DMVPN Phase 2 scenarios where Hub will “reflect” best routes
from spoke without changing next-hop
• Note: Unequal cost load balancing (variance) should be disabled
• Activated using the following command:
(config-router-af-interface)# add-paths number
42

Named mode – Disabling EIGRP on interface
EIGRP process on a specific interface can be disabled using:
(config-router-af-interface)# shutdown
Note: this will stop advertising an associated prefix into EIGRP, unlike with
passive-interface command
43

Named mode – Unique IPv6 behavior
• When IPv6 unicast address family is entered, EIGRP is automatically enabled
on all interfaces with IPv6 address
• It is the easiest way to get IPv6 connectivity in your network
44

Named mode – Default tagging
You can automatically tag all routes (to filter them later) using:
(config-router-af)# eigrp default-route-tag tag
Refer to slide “Filtering – tagging”
45

Named mode – Conversion from classic
Old classic mode configuration can be converted to the new syntax using:
(config-router)# eigrp upgrade-cli process-name
*Note: appeared only in 15.4(2)T
46

Named mode – Verification commands
Verification commands have new syntax:
• show eigrp protocols
• show eigrp address-family [ipv4 | ipv6] as-num interfaces [detail]
• show eigrp address-family [ipv4 | ipv6] as-num neighbors [detail]
• show eigrp address-family [ipv4 | ipv6] as-num topology [all-links]
• show eigrp address-family [ipv4 | ipv6] as-num events
*Note: Old syntax is still supported
47

Additional Resources
• RFC 7868 - EIGRP
• Routing TCP/IP Volume 1, 2nd Edition (Jeff Doyle)
• CCIE Routing and Switching v5.0, Volume 1 (Narbik
Kocharians, Peter Paluch)
• EIGRP Frequently Asked Questions
• Glimpse of EIGRP named mode
• Advances in EIGRP
• EIGRP Route Tag Enhancements
48

Routing Protocol EIGRP

More Related Content

What's hot (20)

Similar to Routing Protocol EIGRP (20)

Recently uploaded (20)

Routing Protocol EIGRP