cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
605
Views
0
Helpful
2
Replies

BGP routes Insertion

12.png

Hi everybdy,

I was practising this simple BGP lab (above fig.) on GNS nothing much fancy now and encountered some issues. The details are:-

on R1 the output of sh ip bgp command is:

Network          Next Hop            Metric LocPrf Weight Path

*>i30.1.1.1/32      1.1.1.2                  0    100      0 200 300 i

*>i30.1.2.1/32      1.1.1.2                  0    100      0 200 300 i

*> 50.1.1.1/32      0.0.0.0                  0         32768 i

*>i100.1.1.1/32     2.2.2.3                  0    100      0 200 400 ?

*>i100.1.2.1/32     2.2.2.3                  0    100      0 200 400 ?

on R2 this is output of sh ip bgp command

Network          Next Hop            Metric LocPrf Weight Path

*> 30.1.1.1/32      10.1.1.2                               0 200 300 i

*> 30.1.2.1/32      10.1.1.2                               0 200 300 i

*>i50.1.1.1/32      1.1.1.1                  0    100      0 i

*>i100.1.1.1/32     3.3.3.4                  0    100      0 200 400 ?

*                   10.1.1.2                               0 200 300 400 ?

*>i100.1.2.1/32     3.3.3.4                  0    100      0 200 400 ?

*                   10.1.1.2                               0 200 300 400 ?

R2#

on R3 output of that command is


Network          Next Hop            Metric LocPrf Weight Path

*  30.1.1.1/32      11.1.1.2                               0 200 400 300 i

*>i                 3.3.3.3                  0    100      0 200 300 i

*  30.1.2.1/32      11.1.1.2                               0 200 400 300 i

*>i                 3.3.3.3                  0    100      0 200 300 i

*>i50.1.1.1/32      2.2.2.2                  0    100      0 i

*> 100.1.1.1/32     11.1.1.2                               0 200 400 ?

*> 100.1.2.1/32     11.1.1.2                               0 200 400 ?

R3#

My question is that

1. In R2 why 30.1 n/ws is not learnt from neighbor 3.3.3.4 on R3 as well?

2. Same in R3, why 100.1 n/ws not learnt from neighbor 3.3.3.3 on R2 as well?

3. When neighborship between R1 and R2 is broken, why R1 doesnot know abt 30.1 n/w routes from R3?

I' very confused now and please clear my doubts.

Thanks

Roman

2 Accepted Solutions

Accepted Solutions

Giuseppe Larosa
Hall of Fame
Hall of Fame

Hello Roman,

let's take IP prefix 30.1.1.1/32 as example.

If we examine R2 BGP table and R3 BGP table we see that:

R2 best path to prefix 30.1.1.1/32 is

*> 30.1.1.1/32      10.1.1.2                               0 200 300 i

via eBGP session with 10.1.1.2

on R3 we see the following different scenario:

*  30.1.1.1/32      11.1.1.2                               0 200 400 300 i

*>i                 3.3.3.3                  0    100      0 200 300 i

the iBGP path via 3.3.3.3 is considered the best path (> symbol) because of its shorter AS path attribute

200 300 compared to 200 400 300.

As a result of this, R3 installs the BGP path coming from R2 and does not advertise back its best choice to R2.

For this reason on R2 we don't see two paths for the prefix but only one.

The same behavior happens with inverted roles for IP prefix 100.1.1.1/32.

This time is R2 that installs the iBGP learned path coming from R3 for the shorter AS path attibute and as result of this it does not propagate the BGP path received on the eBGP session.

This is normal behaviour for BGP, because a BGP router does propagate only best path choice and there is also a sort of split horizon so a router does not send back the best path to the BGP speaker that has sent the best path to it,

Hope to help

Giuseppe

View solution in original post

Hi Giuseppe,

I agree with you but IMHO it would be worth to rephrase your sentence

" ... and there is also a sort of split horizon so a router does not send back the best path to the BGP speaker that has sent the best path to it."

which might be a little misleading.

The BGP split horizon rule is related to iBGP only, it says: "Prefixes received from an iBGP neighbor are not advertised to iBGP neighbors."

Which is also the behaviour noticed in this case.

View solution in original post

2 Replies 2

Giuseppe Larosa
Hall of Fame
Hall of Fame

Hello Roman,

let's take IP prefix 30.1.1.1/32 as example.

If we examine R2 BGP table and R3 BGP table we see that:

R2 best path to prefix 30.1.1.1/32 is

*> 30.1.1.1/32      10.1.1.2                               0 200 300 i

via eBGP session with 10.1.1.2

on R3 we see the following different scenario:

*  30.1.1.1/32      11.1.1.2                               0 200 400 300 i

*>i                 3.3.3.3                  0    100      0 200 300 i

the iBGP path via 3.3.3.3 is considered the best path (> symbol) because of its shorter AS path attribute

200 300 compared to 200 400 300.

As a result of this, R3 installs the BGP path coming from R2 and does not advertise back its best choice to R2.

For this reason on R2 we don't see two paths for the prefix but only one.

The same behavior happens with inverted roles for IP prefix 100.1.1.1/32.

This time is R2 that installs the iBGP learned path coming from R3 for the shorter AS path attibute and as result of this it does not propagate the BGP path received on the eBGP session.

This is normal behaviour for BGP, because a BGP router does propagate only best path choice and there is also a sort of split horizon so a router does not send back the best path to the BGP speaker that has sent the best path to it,

Hope to help

Giuseppe

Hi Giuseppe,

I agree with you but IMHO it would be worth to rephrase your sentence

" ... and there is also a sort of split horizon so a router does not send back the best path to the BGP speaker that has sent the best path to it."

which might be a little misleading.

The BGP split horizon rule is related to iBGP only, it says: "Prefixes received from an iBGP neighbor are not advertised to iBGP neighbors."

Which is also the behaviour noticed in this case.