Category: EIGRP

EIGRP: Over the Top

Also referred to as OTP, this feature allows EIGRP to create overlay multi-point VPNs between edge routers without any coordination with the service provider connecting the edge routers, removing the need for a service provider to assist in distributing the routing protocol messages and exchanges.

OTP functions using the Locator/Identifier Separation Protocol or LISP.  LISP separates a host machines address and the location of the host itself, allowing the host to retain its identity regardless of its location in a network.  The location and identity are considered two separate independent entities, both represented by a complete address, and provides a way to define and map so that the address representing the identity of a host can be resolved into the address that represents the location.  Then a tunneling mechanism is used to encapsulate the packets between end hosts addressed using the end host identities into new packets that are destined to the addresses representing the end host locations.  This allows a host to change its location while retaining its identity without losing connectivity.

Host machines with LISP have an Endpoint ID or EID, that identifies its identity which never changes.   This EID can be an IPv4 address an IPv6 address or any other address format as needed.  The outside address of  the edge router represents the location of the EID of the hosts, this is referred to as  the Routing Locator or RLOC.  Multiple EIDs can exist behind a single RLOC.

LISP.PNG

LISP also has a control and data plane, where the control plane comprises of the registration protocol and procedures by which the tunnel routers register the EIDs they are responsible for along with their RLOCs, then use those registrations they map EIDs into the RLOCs.  The data plane defines the actual tunnel encapsulation used between RLOC routers.

Configuration of OTP involves the use of statically configured EIGRP neighbors to exchange EIDs and RLOCs, however this does not scale if there are multiple and growing number of sites.  OTP uses a route-reflector feature very similar to BGP to exchange routes in a hub and spoke model fashion.  The only neighbor routers are configured with in this scenario is the hub router which exchanges all OTP EID and RLOC information to the spoke sites.

EIGRP: Split Horizon

In hub and spoke networks where multiple sites would exist and be routed through the same interface at the hub, it may be necessary to disable Split Horizon.  EIGRP uses Split Horizon with poison reverse however for this type of implementation, it is necessary to disable this feature.

To disable split horizon you would use the no {ip | ipv6} split-horizon interface command in classic mode or just the no split-horizon command in af-interface section of named mode.

EIGRP: Default Routing

EIGRP has no command to inject a default route into the routing table so there are only two methods by which you can install one into EIGRP:

  1. Redistribute the default route from another routing source
  2. Manually summarize a default route to all neighbors

If a static default route is configured without a next hop interface and you configure a network command with 0.0.0.0 IOS will think the route is directly connected and advertise it out to neighbors.  However, this will also cause EIGRP to advertise out all directly connected interfaces, this could be a non desirable behavior and as such is never a good idea to advertise the default route in such a way.

EIGRP: Authentication

EIGRP supports both MD5 and SHA-2 hashing algorithms, where MD5 is supported in classic and named mode and SHA-2 is only supported in named mode.

Configuration of authentication includes the use of Keys and Keychains, where you can optionally set the time frame where certain keys are valid and others are not.  If multiple keys are eligible at a certain time of day the key with the lowest Key ID is selected.

EIGRP: Graceful Shutdown

When an EIGRP router goes offline for instance when the power goes out, other EIGRP neighbors must wait for the holddown timer to begin any sort of convergence on the network event.

EIGRP routers configured for graceful shutdown however, when they’re shutdown will send a Goodbye message to let other neighbors know it’s going down so they don’t have to wait on the holddown timer.

A Goodbye message is simply a standard Hello message with all K values set to 255

EIGRP: Passive Interfaces

The intended use of EIGRP interfaces to advertise and accept and process EIGRP packets is meant only for interfaces actually participating in the EIGRP network.  To prevent interfaces from participating in EIGRP and no longer accept/process EIGRP packets you can configure the passive-interface command for those interfaces individually or globally, and enabling only  the interfaces you want to participate.

EIGRP: Route Summarization

Route Summarization is the act of advertising a larger supernet to the network that encompasses multiple contiguous subnets.  Doing this allows for more efficient and leaner routing tables, and benefits to Query propagation.

Should a Query be sent to a router advertising a summary route, if that router did not have the more specific route in its routing table it will immediately Reply back with an infinite metric.

Whenever summary routes are advertised the router performing the summarization also installs a ‘discard route’, this discard route is identical to the network and netmask of the advertised route, and its interface is set to Null0.

The purpose for this route is for any traffic traveling to a subnet encompassed by  the summary route, where the destination network is not reachable or does not exist, it will cause that traffic to be dropped to Null0 instead of following the default route…so it’s an effective tool to prevent routing loops from occurring.

EIGRP also takes into account all component routes of the summary and chooses the best metric to advertise the summary.  This also means that if those component route metrics change, the summary route metric will also change.  This can be CPU intensive for networks who have thousands of component routes for a single summary.  To get around this you can set the summary metric statically.

EIGRP: Stub Routers and Query Packets

Stub routers handle Query packets the following ways:

  • Originating Query Packets are not modified in any way, entering active and sending Queries are the same.
  • Processing received Query packets, depends on the network that was queried.  If the network queried is allowed to be advertised by the receiving router, the router will process the Query normally and send back an appropriate Reply.   If the queried network is one that the router knows about but is not allowed to advertise, it will be processed in the usual way but the reply will always indicate infinite distance, regardless of what the Stub router truly knows about the network.  Receiving a query for an unknown network will immediately cause the router to Reply with an infinite distance.

EIGRP: Stub Routing

Stub routing is used to improve network stability and scalability.  It’s primarily used in hub and spoke networks and would only be configured on spoke routers.  routers configured as stub announce it in additional TLVs in EIGRP Hello messages.

Here are the results of configuring stub routers:

  • Stub routers do not advertise routes learned via EIGRP to its neighbors except those specified in a leak-map.  This prevents stubs from being considered Feasible Successors for remote networks.
  • Stub routers advertise only its own EIGRP enabled networks to its neighbors.  These prefixes are denoted as eigrp stub in  the configuration.
  • Neighbors of a stub router are aware of the stub status, and will never send Query packets to a stub router, this prevents neighbors from converging/diffusing through a stub router to a remote network.

 

EIGRP: Add-Path

In some scenarios in which you have a hub router that connects to multiple paths to a specific site, only the hub router is aware of the redundant paths to the same destination. When it advertises routes for that destination to other neighbors it only advertises the single best path to that site.

Add Path support allows for the hub router to advertise all possible paths to the same destination so other neighbors can send traffic to it on both links, or have the redundant path in the routing table for immediate failover.  The only requisite needed by the hub router is to turn off split horizon on the multipoint tunnel interfaces toward individual spokes.

The Variance and Add-Path features are not compatible with each other.  when using the Add-Path set the variance to 1.

no-ecmp-mode is recommended if the hub router uses multiple tunnel interfaces to reach the spoke sites.