Ask the Expert: How to configure and troubleshoot EIGRP, OSPF and BGP.

[Cisco Moderador]

This topic is a chance to discuss more about about the most widely deployed Routing Protocols in today’s Enterprise Networks, such as EIGRP, OSPF and BGP. The use of dynamic Routing Protocols is necessary in order to keep the network operational, provide scalability and reduce complexity. Routing Protocols are easy to set up, however, inner workings and predicting the network reconverge might be a challenge specially in failover scenarios.

To participate in this event, please use the  button below to ask your questions

Ask questions from Monday, August 8 to August 19th, 2016

Featured Expert

Hector Gustavo Serrano is a Customer Support Engineer in the Global Cisco Technical Assistance Center. He is part of the Routing Protocols team, specialized in troubleshooting Routing issues, additionally he educates about best practices of Routing Protocols deployed on customer’s networks. For the last 6 six years he has worked with Cisco platforms for Enterprise Networks, specially focused on US customers. Hector Gustavo also holds a CCIE in Routing & Switching (49102) and a Bachelor’s degree in Telecommunication engineering.

Hector Gustavo might not be able to answer each question due to the volume expected during this event. Remember that you can continue the conversation on the  Contact Center Community 

Find other  https://supportforums.cisco.com/expert-corner/knowledge-sharing.

**Ratings Encourage Participation! **
Please be sure to rate the Answers to Questions

[Sarah Staker]

Hello Hector,

I am trying to understand when do we need to use redistribution and what is the most efficient way to redistribute routes from one routing protocol to another. Do you have any guidelines?

Thank you.

- Sarah

[lespejel · Accepted solution]

Hello Sarah

You need redistribution when your router has learned a route from a Routing Protocol, let's say EIGRP, and you want to advertise the same route over a different protocol, like OSPF.

Some guidelines to consider when planning redistribution are:

• You can only redistribute routes that are installed in the routing table
• Redistribution is not transitive, meaning if you have static route redistributed into EIGRP, and then EIGRP redistributed into OSPF on the same router, that route won't be present in OSPF because originating protocol is 'static', not EIGRP, so in order to inject the same route into OSPF, you must also redistribute static into OSPF.
• When redistributing routes into OSPF you must include the subnets keyword
• When redistributing routes into EIGRP you must include metrics (unless is coming from another EIGRP process)

By default:

• internal BGP routes are not redistributed
• OSPF external routes are not redistributed

router eigrp 1
network 10.0.0.0
redistribute ospf 1 metric 10000 10 255 1 1500 route-map from-OSPF

router ospf 1
network 11.0.0.0
redistribute eigrp 1 subnets route-map from-EIGRP

ip access-list ROUTES1
permit 10.0.1.0 0.0.0.255

ip prefix-list ROUTES2 seq 5 permit 11.0.2.0/24

route-map from-OSPF
match ip address ROUTES1

route-map from-EIGRP
match ip address prefix-list ROUTES2

The best way to configure redistribution is to use route-maps to control what routes you want to inject and have a granular control by using ip prefix-lists or access list to match the desired routes.

There are several documents describing redistribution available from Cisco site:

-Route Redistribution Explained

-Understanding Redistribution of OSPF Routes into BGP

-Redistributing Routing Protocols

-Route-Maps for IP Routing Protocol Redistribution Configuration

[Hector Gustavo Serrano Gutierrez · Accepted solution]

Hello Sarah,

Redistribution is needed to make communication work when more than one Routing Protocol is running on the network. 

Example of a Redistribution Scenario:

As seen on the diagram:

• R1 is EIGRP peer with R2.
• R3 is OSPF peer with R2.

In order R1 can communicate with R3, the Router R2 should do mutual Redistribution between EIGRP and OSPF protocols. This is necessary so R1 and R3 can receive the respective routes using the Routing Protocols they are running.

The configuration for Redistribution on R2 would look similar to:

!
router eigrp 1
 redistribute ospf 1
 default-metric 1000000 1 255 1 1500
!
router ospf 1
 redistribute eigrp 1 subnets
!

Few variances in the configuration exist depending on what exact Routing Protocol the Router is running. However, the configuration is similar as the one shown above.

There are many different scenarios where Redistribution comes in handy in real world deployments. Additionally, Redistribution should be used with caution, specially when it is required to be configured in more than one Router (for instance, in failover scenarios), as Routing Loops might be introduced to the network if doing careless.

Good Cisco documentation can be found via:

Redistributing Routing Protocols
http://www.cisco.com/c/en/us/support/docs/ip/enhanced-interior-gateway-routing-protocol-eigrp/8606-redist.html

I hope this information has been useful for you.

Best Regards.

[1001QA.Net]

Hi Hector

Does BGP Soft reconfiguration work with RIB Adj In and RIB Adj Out databases?

The documentation says that:

-software outbound reconfiguration is always there and it is done by reseting BGP RIB version to 0. Then there will be a difference between RIB Adj Out version and the BGP RIB and the route will generate updates for the neighbors that have out of sync records

So for the above there is no need of another data structure

-software inbound reconfiguration - for this the documentations says that the router keeps an unfiltered copy of the routes received from its neighbors, which is used to rerun the filtering in order to see what routes are allowed with the new policy.

My question is this: Is this copy of unfiltered routes the same as RIB ADJ In table? If not why can't this be used instead of having a copy of this. The documentation seems to indicate RBB Adj in as a permament data structure not as a structure that is created when the software reconfiguration inbound feature is configured.

[Vinayaka Raman]

Hi Hector,

 Thanks for the opportunity.

1) When should we consider using EIGRP named mode instead of classic mode?

2) In OSPF we have 2 layer hierarchy that non-zero areas should connect to area 0 backbone directly.

If this rule doesn't exist how would it cause loop?

[Hector Gustavo Serrano Gutierrez]

Hello Vinayaka,

1) EIGRP named mode brings advantages over EIGRP classic mode.

For example:

• The use of EIGRP Wide Metrics allows EIGRP to calculate an appropriate metric which can differentiate the use of high-speed interfaces, such as 10 Gigabit Ethernet or Ether channels on the network, therefore computing the best path accordingly.

This is possible as Wide Metrics feature supports 64-bit metric, the composite cost metric formula was modified and also, the delay is now measured in picoseconds.

- EIGRP Wide Metrics

- Enhanced Interior Gateway Routing Protocol (EIGRP) Wide Metrics White Paper

• Named mode has been extended to support hierarchical configuration. The advantage is that all the EIGRP related commands can now be found under the EIGRP process section, which facilitates identifying EIGRP related commands.

• EIGRP Security enhancements with the use of SHA2-256 Algorithm for stronger authentication between EIGRP peers.

• New products which require the use of EIGRP are developed with Named mode in mind, like the iWAN Cisco's solution.

Additionally, a command that might come in handy is: 'eigrp upgrade-cli', which can be used to upgrade from classic mode to named mode smoothly.

- EIGRP Classic to Named Mode Conversion

As a conclusion, on new network deployments, I would consider the use of EIGRP named mode. There is nothing wrong with keep running EIGRP classic in stable network deployments though.

2) A Router that is not part of Area 0 is not considered an ABR (Area Border Router). That means the device will not generate the OSPF LSA (Link State Advertisements) Type 3 - Summary necessary to make inter-area communication possible:

In the above topology, R3 will not generate LSAs type 3 - Summary (it is not an ABR) between Area 1 and Area 2, making R4 unreachable from Area 1 and Area 0.

The solution in such scenario would be the use of OSPF Virtual-links.

Now, R4 should be receiving the appropriate LSAs to know the path to communicate with Area 1 and Area 0.

More information about how to configure Virtual Links can be found via:

- OSPF Virtual Link

You would ask, "well, why not configuring a loopback0 in R3 on Area 0, that should make it an ABR, and should start creating LSA type 3 - Summary packets, right?"

Technically yes, however, the real Area 0, (in our example, between R1 and R2) would not accept the LSAs type 3 generated on Area 2, since a Backbone Area (aka Area 0) should only export and not import Type 3 LSAs, therefore breaking full internetwork communication.

That is also an example of a discontiguous OSPF network design.

As per RFC2328, all the OSPF areas should connect to Area 0.

I hope this information has been useful for you.

Best Regards.

[vasanth77]

Hi Hector,

Can I redistribute routes between OSPF and BGP, EIGRP and BGP,

EIGRP and IS-IS?

Thanks in advance

[Hector Gustavo Serrano Gutierrez]

Hi vasanth77,

1) Yes, you can redistribute between OSPF and BGP. Some quick observations:

• When redistributing OSPF into BGP, the OSPF External type routes are not redistributed by default.
• When redistributing any Routing Protocol into OSPF, you should include the 'subnet' parameter at the end of the 'redistribute' statement.

2) Yes, you can redistribute between EIGRP and BGP. Some quick observations:

• When redistributing from any other Routing Protocol into EIGRP, you should manually set the parameters bandwidth delay reliability load MTU, so the metric can be calculated for the new EIGRP prefixes. Actually, It is always a best practice to do so.

3) IS-IS is unlikely to be found in Enterprise Networks, it is more frequently used in the core of Service Provider environments.

Yes, you can redistribute between EIGRP and IS-IS. Some quick observations:

• When redistributing IS-IS into EIGRP, by default, only Level-2 routes are redistributed. However, you can still specify to redistribute Level-1 or Level-1-2 too.
• When redistributing into IS-IS, a default metric value of 0 is used by default.

It is important to remember the rule of thumb that only routes installed in the Routing Table are redistributed.

- Redistributing Routing Protocols

I hope this information has been useful for you.

Best Regards.

[vasanth77]

Thanks Hector.

[Hector Gustavo Serrano Gutierrez]

No problem vasanth77.

Thanks for participating.

Best Regards.

[mpbhargav]

Hi Experts,

We have a requirement where in we need to place two Linux servers in active/standby mode and the boxes will be geographically separated and each box will be in a separate subnet.

What we want is to provide a VIP (Virtual IP which needs to be floating between two boxes. The VIP address can be in a separate subnet and there are no real hosts in that subnet ) so that the client which needs to  connect to the server will use the VIP address instead of real IP address. The VIP will always be owned by the active box and upon failover we need to move the VIP to new active.

Problem: Can we move a VIP  from one host to another.? if possible,  How the routing is handled here since the VIP is now moved from one host to other.? Whether we need any routing protocol here to indicate the routers  when the subnet got moved..?

Please suggest any other alternate solution(if any).

[Hector Gustavo Serrano Gutierrez]

Hello mpbhargav@gmail.com,

Thanks for participating, regarding your questions:

1. Can we move a VIP  from one host to another.?

I wouldn't be able to tell you if nowadays Operating Systems for Servers have that feature incorporated or not. At least in theory, it should be possible. For that to happen, there should be some kind of 'heartbeat' mechanism to make the Servers aware the peer is up or not.

2. How the routing is handled here since the VIP is now moved from one host to other.?

From the Networking/Routing point, a workaround that I can think is that the Router on the Active Server side can advertise a more specific prefix for the VIP of the Active Server, I am referring to a /32 advertisement. If that Router senses the Server is unreachable (maybe monitoring it with an IP SLA + Track configuration pointing to the Server), the Router will stop advertising the /32. The Router on the Standby Server network, should always be advertising a less specific route for the VIP server's network.

That way, as long as the VIP for the Active Server network is advertised, the rest of the network will prefer the path to the Active Server.

Advertising both VIPs at the same with the same prefix-length time will make the network prefer one Server or the other depending on the metric the remaining Routers are calculating to reach the VIP network.

3. Whether we need any routing protocol here to indicate the routers that subnet when the subnet get moved..?

A workaround would be using the one described above using any Routing Protocol, it depends in the detailed network implementation.

I am not sure if LISP technology may be a better option for the described scenario. I still need to go deeper into it.

I hope this helps.

Best Regards.

[mpbhargav]

Hi Hector,

Thank you so much for your inputs.

Yes, with the pacemaker and corosync (which is a cluster monitoring software that implements heartbeat) we can detect  if the peer is up or not.

Regarding my second question,

How the routing is handled here since the VIP is now moved from one host to other.?

Whether the Server needs to advertise a specific prefix towards the router when the server becomes active(by using routing protocols like RIP or OSPF).? or a router itself can be configured monitor a VIP network.? 

Also can you explain more on the following:

"The Router on the Standby Server network, should always be advertising a less specific route for the VIP server's network."

Whether it means the router has to indicate the other routers this is not the preferred route to the destination(in BGP terms insert extra AS numbers in the ASPATH, since its a backup route).? 

[Hector Gustavo Serrano Gutierrez]

Hello mpbhargav@gmail.com,

Let's take a look at the following diagram:

Whether the Server itself advertises its VIP network using a Routing Protocol (I know for sure there is software that can make Servers run Routing Protocols) or the the Router is who advertises the VIP, the idea is the Active side should advertise a more specific prefix-length for the VIP network as seen above.

That way, the rest of the network should prefer the path towards the Server at left side (configured as Primary) as long as its more specific prefix-length is advertised.

Upon a Server failure, the VIP will stop being advertised, causing the network to failover to the next best path towards the VIP network, which resides on the Standby Server at the right side of the topology.

A Router/L3 Switch bases its Routing decision on:

1. Prefix length.
2. Administrative Distance
3. Metrics

About point #1:

  1. Prefix length. The longest subnet mask is preferred.

For instance:

• 1.1.1.1/32 entry will be preferred over
• 1.1.1.1/24 entry, due to the prefix length.

- Route Selection in Cisco Routers

The above approach should help you as a starting point.

Details about the implementation should be carefully reviewed as it depends in the existing network deployment. (Is the WAN the Internet, MPLS/L3 VPN, VPLS, etc?)

Adding an AS number in the AS_PATH should for sure make the route less desirable, however, in certain scenarios, some parts of the network can still elect that route to forward the traffic.

I hope this helps.

Best Regards.