Load balacing in RIP and Ibgp my changing AD

ganpatspatilganpatspatil
in CCIE Routing & Switching Technical

 

 

 


R6--------------------------R5
           |   
           |
           |----------------R22

 

R6 to R5 RIP ( IP SUNBET 9.9.56.0/24)

R6 to R22 IBGP (IP SUNBET 9.9.226.0/24)

R5 advt 55.0.0.0/8 via RIP

R22 advt 55.0.0.0/8 via IBGP

I want to do loan balance the traffic to destination 55.0.0.0/8

I have modified the AD on R6 for IBGP route to 120.

-------R6

Interface              IP-Address      OK? Method Status                Protocol
FastEthernet0/0.46     9.9.46.6        YES NVRAM  up                    up     
FastEthernet0/0.56     9.9.56.6        YES NVRAM  up                    up     
FastEthernet0/0.62     9.9.226.6       YES NVRAM  up                    up     
Loopback0              9.9.0.6         YES NVRAM  up                    up    

!
router rip
 network 9.0.0.0
 no auto-summary
!
router bgp 100
 no synchronization
 bgp log-neighbor-changes
 bgp redistribute-internal
 neighbor 9.9.226.22 remote-as 100
 distance bgp 2 120 10
 no auto-summary
!

----R5

 

R5#sh ip int brief
Interface              IP-Address      OK? Method Status                Protocol
FastEthernet0/0        unassigned      YES NVRAM  up                    up     
FastEthernet0/0.56     9.9.56.5        YES NVRAM  up                    up     
Loopback0              9.9.0.5         YES NVRAM  up                    up     
Loopback55             55.55.55.55     YES NVRAM  up                    up 

router rip
 network 9.0.0.0
 network 55.0.0.0
 no auto-summary
!

----R22

R22#sh ip int brief
Interface              IP-Address      OK? Method Status                Protocol
FastEthernet0/0        unassigned      YES NVRAM  up                    up     
FastEthernet0/0.62     9.9.226.22      YES NVRAM  up                    up     
FastEthernet0/1        unassigned      YES NVRAM  administratively down down   
Serial2/0              unassigned      YES NVRAM  administratively down down   
Serial2/1              unassigned      YES NVRAM  administratively down down   
Serial2/2              unassigned      YES NVRAM  administratively down down   
Serial2/3              unassigned      YES NVRAM  administratively down down   
Loopback0              9.9.0.22        YES NVRAM  up                    up     
Loopback55             55.55.55.55     YES manual up                    up    

 
R22#

router bgp 100
 no synchronization
 bgp log-neighbor-changes
 network 55.0.0.0
 neighbor 9.9.226.6 remote-as 100
 no auto-summary
!

 ON R6 i can see AD value of the ibgp route is set to 120.howvever load balaning not happening.

R6#sh ip route 55.0.0.0
Routing entry for 55.0.0.0/8
  Known via "bgp 100", distance 120, metric 0, type internal
  Last update from 9.9.226.22 00:12:27 ago
  Routing Descriptor Blocks:
  * 9.9.226.22, from 9.9.226.22, 00:12:27 ago
      Route metric is 0, traffic share count is 1
      AS Hops 0
      MPLS label: none

 

If i shut the interface between R6 and R22....same route (55.0.0.0/8) is learned from R5.

R6#sh ip route 55.0.0.0
Routing entry for 55.0.0.0/8
  Known via "rip", distance 120, metric 1
  Redistributing via rip
  Last update from 9.9.56.5 on FastEthernet0/0.56, 00:00:04 ago
  Routing Descriptor Blocks:
  * 9.9.56.5, from 9.9.56.5, 00:00:04 ago, via FastEthernet0/0.56
      Route metric is 1, traffic share count is 1

 

By default R6 is prefering iBGP route to reach 55.0.0.0/8.

Can anyone please suggest why load balancing is not happening on R6 in between RIP and Ibgp route ?

 

Regards

Ganpat

 

 

 

 

Comments

  • markus.wirthmarkus.wirth

    Hi!

    Nice question! Answer is quite straight I think:

     

    Usually you do load balancing with static routes or any other routing protocol, when there are equal costs to the destination (metric). An exception are BGP and EIGRP, those protocols allow unequal-cost load-balancing and you can see this with the "sh ip route x.x.x.x" output when you have a look at the traffic share count. But imagine you were the router and you have one side BGP with ad 100 and one side RIP with AD 100 to the destination. How would you judge your unequal cost load-balancing here? I mean whats better....2 hops or a BGP metric or a AS-path? Or you could even take a different protocol....EIGRP with a cost of 45828424? What would you compare this to in hops?

    Regards!

    Markus

  • ganpatspatilganpatspatil

    Hi Markus,

    As per my underatnding if router is receiving same prefix from two routing protocol and bot having same AD then its should install route in Routing table from both protocol

     

    Can you please explain me in details?

  • markus.wirthmarkus.wirth

    I tested it in the lab an before I paste a lot of stuff here you can read this article here.

     

    http://packetlife.net/blog/2008/dec/6/two-routing-protocols-same-ad/ 

     

    I get the same results as Jeremy!

    HTH!

    Regards!

    Markus

  • RobechoRobecho

    Load balancing is done only for routes learnt with simmilar method i.e IGP/BGP or static mechanism.

    Since Metric is highly significant only to a protocol, it is illogical for RIB process to compare route metric of different protocols with the same AD.

    When RIB is faced with decision between routes of same AD, it decides as follows:

    1. Select connected route
    2. Select static route
    3. Select BGP route (irrespective of eBGP or iBGP)
    4. Select EIGRP route ( irrespective of external or internal)
    5. Select OSPF route
    6. select RIP route
  • markus.wirthmarkus.wirth

    Nice one!

    Thanks for the post!

    Do you have a source for us where you got that info?

     

    TIA!

    Regards!

    Markus

  • RobechoRobecho

    Actually I couldn't find any written document about this

    I created a lab with all protocols & static routes and found the result. Anyone can do the same.

    http://netoperation.blogspot.com/2011/03/administrative-distance.html

    P.S I didnot test ISIS but from what I have found for other protocols, it shouldn't be any different.

  • ganpatspatilganpatspatil

    Thanks for your replyies guys!!!

     

    /Ganpat

  • RobechoRobecho

    Hi Guys,

    I'm so sorry but the conclusion I have made previously was very incorrect. The correct one is if the AD values are the same for routes b/n different protocols, then the oldest route ( the first one to enter the RIB) will win.

     

    Please find the right information on Alex Zinin's book chapter 4 Routing maintenance:

     


    When a route is passed to the Routing table (RT) process, the router uses the
    following logic.


    1.


    Check whether the routing table already has a route to the
    same network.

    2.


    If no route is found, install the new
    route.

    3.


    If an existing route is found, compare the AD and metric
    values of the old and new routes as follows.
    a. If the AD of the new
    route is different from the AD of the route in the table, the route with the
    lower AD value is preferred. If the new route is chosen, all paths in the old
    route are removed, the new path is installed, and the associated AD and metric
    values of the route are changed accordingly. Otherwise—the already installed
    route has a better AD value—the new route is not installed, but the router saves
    information about the protocol that submitted the new route for installation in
    the prefix descriptor and sends a query to the protocol when a backup route is
    needed.


    b. If the AD values of the routes are equal and both routes were
    submitted by the same process, the new route is processed according to the rules
    of distance-vector protocols (see Chapter
    8    ).


    c. If the AD values are equal but the routes are from different
    sources, the behavior depends on whether the routes are submitted by processes
    of the same type. If they are, the metrics of the routes are compared as if the
    routes were from the same protocol. If metrics are the same, the route from the
    process with the lower process number wins. If the routes are from different
    sources, the new route is ignored. (Note that this check is implemented for
    EIGRP routes only. All other routes are treated as from different protocols, so
    the old route is left in the routing table    .)
  • davidrotheradavidrothera

    I think you might be misunderstanding something there.

    If two prefixes are to be compared that have the same AD but are from different protocols (i.e. one protocol has had its AD increased/decreased) then the router will compare the original AD before manipulation as the tie breaker.

    What the paragraph from the book you have posted is trying to say is that if within say EIGRP a prefix is advertised from two sources with the same AD and same metric then if you have ECLB turned off then it will only use the older of the advertisements.

  • RobechoRobecho

    HI David,

    I must disagree.This chapter discusses how the RIB process works on a cisco router. Route source is the routing information source ( or routing protocol) that injected the routes into the RIB process.

    When routes to the same destinations are submitted to it from the same route source process, different route source process of the same type or from different process of different type, the RIB route selection process will take the step as stated in the book.

    If the routes are submitted have equal AD ( for whatever reason) and if they come from different EIGRP process/AS, then RIB will select them based on lowe metric and lower process number.

    If the routes are submitted have equal AD ( for whatever reason) and if they come from different protocols or the same protocol other than EIGRP but different process, then the oldest route will win the election. 

  • northlandboynorthlandboy

    If the routes are submitted have equal AD ( for whatever reason) and if they come from different protocols or the same protocol other than EIGRP but different process, then the oldest route will win the election. 

    That's not what I've seen in the documentation. I've seen it say that original AD is used. Should be easy to validate with a lab. 

    I've also seen some blog posts this week from @ioshints and Daniel Dib saying that if you have multiple OSPF processes, the process ID is used as a tie breaker, not the age. Daniel showed this with a simple lab exercise. 
  • davidrotheradavidrothera

    Instead of just believing a book published in 2001 I went away just to check it all out:

    http://networkbroadcast.co.uk/2013/03/cisco-rib-operations-ad-comparisons/

    Hopefully that clears it up for you Rob.

  • RobechoRobecho

    @david,

    Thanks for checking that out. Yes the book is very old, but from what I have seen so far most of the fundamental things in the book are still legitimate. But it seems to contradicts from what we see in our "experiments". I'll verify it further

     

    @northlandboy,

    Which documentations have you seen? I couldn't find any "legitimate" documents except Alex Zinin's so far.

  • northlandboynorthlandboy

    @northlandboy,

    Which documentations have you seen? I couldn't find any "legitimate" documents except Alex Zinin's so far.

    It's one of those things I remember reading, but right now I can't find any good sources. I'll have a hunt around, see if I can find it again.

  • HS628HS628

    Thanks to all for clearing this up (for me). So suppose multi-path is disabled, route will be resolved in the following order:

    AD
    Original AD
    Metric
    Process Id
    Age

    It does make a lot of sense since this guarantees that metric is compared only between sources of the same type.

Sign In or Register to comment.