cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
400
Views
5
Helpful
5
Replies
Highlighted
Beginner

BGP redistribution

hello

we have below scenario

ospf-bgp.png

in task asked us to aggregate and given already 1 prefix(and mask) for each router suppose :

aggregation for :
R1 :  1.0.0.0/8
R2 :  2.0.0.0/8

then asked each router must propagated those prefixes which received via aggregation through a eBGP into their respective IGP domain:

so i think:

it means each router received that prefixes we aggregated on each side should redistributed into their IGP protocol

solution is :

R2:
router eigrp 1
network <something>
no auto-summary
redistribute bgp 65400 metric 10000 100 255 1 1500
-------------------------------------------------
router bgp 65400
redistribute eigrp 1

and on R1 :

router ospf 1
network <something > a 0
redistribute bgp 65200 match int ext 1 ext2
-----------------------------------------
router bgp 65200
redistribute ospf 1 subnets

My question is Why mutually redistribution ?

as each router received aggregated prefixes via eBGP , then redistributed to their IGP (R1 redistribute BGP into OSPF ) and (R2 as received do it for EIGRP)

 

just i think in this case which we did not aggregated all prefixes from each sides , so each side (each ASs wants other ASs prefixes)

and NOT all of them sent via aggregation we did on R1 and R2, just this reason comes into my mind

what you think ? is(are) there any things else for mutually redistribution ?

because by just redistributing BGP into IGP we can achieve to our goal...

 

thanks and regards

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted

Hello cisc0.ameer,

I try to answer to your questions

1)

my question is Why we redistribute BGP into IGP protocol ?

Do we do it because those route(s) carried by BGP(with network command) , we want to injected into IGP?

 

Yes I agree the redistribution of BGP into each IGP is for providing the routes into the routing domain of the receiving IGP.

 

2)

with using Network and aggregate what's need for summary-address ?

and what happened if We don't use summary-only keyword?

 

If you don't use the summary-only keyword BGP advertises the summary route and all alive component routes.

So the summary-only is needed if you don't want BGP to propagate component routes in addition to the summary route.

 

summary-only will be used  
whenever we have bunch of Prefixes 
10.1.0.0/16
10.2.0.0/16
10.3.0.0/16
.
.
.
and just sending summary only but condition is we have this bunch routes in our Table [in this case BGP table] correct?

Your understanding is correct the aggregate-address command requires that at least one component route to be in the BGP table in the local node by any means (network commmand, learned by another BGP peer or the result of a redistribution on local node)

 

>> 3- sorry for this silly Question : which prefixes should be advertised through a Network command ? those we want to aggregate or NO some internal we want other Side have that too.

 

First of all, with new BGP default of no auto-summary the network command requires an exact match to work so we need to advertise at least one component route with the exact mask it has in the IP routing table.

Advertising multiple component routes is recommended to avoid the summary route state to depend on a single component route.

How many component routes to advertise depends on the usage or not of the summary-only keyword.

We don't need to advertise all the existing component routes if we are going to use summary-only.

A subset of network commands with exact match on subnet mask is enough, because the summary route will cover all the missing component routes: it is used to represent all the address block.

The sending of some component routes is required when we have multiple paths / neighbors and we would like to perform load sharing of inbound traffic based on component routes: on Router RA we send out the summary route plus some component routes (we can use the suppress-map to invoke a route-map to decide what component routes to send)

on RB we can send the same aggregate and a different subset of component routes (again using a route-map invoked in a suppress-map option in aggregate-address command).

Because the most specific route is used RA will attract traffic directed to the component routes it advertises and the same will happen on RB.

 

Hope to help

Giuseppe

 

View solution in original post

5 REPLIES 5
Highlighted
Beginner

i mean this not mentioned in TASK
but they want each side (each ASs) have completely full all prefixes from each other
means
OSPF(in AS 65200) has EIGRP routes
and
EIGRP(in AS 65400) has OSPF routes
for this reason we should do mutually redistribution....
Highlighted

Hello cisc0.ameer,

there are some problems of interpretation of this lab exercise.

As far as aggregate routes are involved mutual redistribution is not needed.

 

You can even build a working configuration that does not use any form of redistribution into BGP but just network commands +

aggregate-address 1.0.0.0 255.0.0.0 summary-only on one side

 

and network commands+

aggregate-address 2.0.0.0 255.0.0.0 summary-only on the other side

 

both in router bgp environment

 

then you need to redistribute BGP into OSPF on router R1 and BGP into EIGRP with a seed metric on R2.

 

However, if they ask for the routers to see all the original routes in each IGP the need for mutual redistribution arises and also the aggregate-address command needs to remove the summary-only option.

 

The lab may have two phases the first one with usage of aggregate address and the second phase asking for mutual redistribution.

You need to read carefully ALL the lab including all the requirements before to make implementation choices.

 

Edit:

be aware that the proposed solution for R1 looks like not correct

 

router ospf 1
network <something > a 0
redistribute bgp 65200 match int ext 1 ext2
-----------------------------------------
router bgp 65200
redistribute ospf 1 subnets

 

the match internal ext1 ext2 applies when redistributing OSPF into another protocol.

Only OSPF requires the subnets keyword in redistribution.

 

The correct configuration for R1 is:

 

R1:

router ospf 1

redistribute bgp 65200 subnets

 

router bgp 65200

redistribute ospf 1 match internal ext1 ext2

 

 

Hope to help

Giuseppe

 

 

Highlighted

too much thanks for perfect answer

---------------------------------------------------------

some doubt i have , i mean i did not understand well

using Network Command advertise some prefix(s) through a our BGP connection(eBGP) .

my question is Why we redistribute BGP into IGP protocol ?

Do we do it because those route(s) carried by BGP(with network command) , we want to injected into IGP?

 

2-with using Network and aggregate what's need for summary-address ?

and what happened if We don't use summary-only keyword?

summary-only will be used  
whenever we have bunch of Prefixes 
10.1.0.0/16
10.2.0.0/16
10.3.0.0/16
.
.
.
and just sending summary only but condition is we have this bunch routes in our Table [in this case BGP table] correct?

 

3- sorry for this silly Question : which prefixes should be advertised through a Network command ? those we want to aggregate or NO some internal we want other Side have that too

 

again thanks for correcting my mistake sir

Highlighted

Hello cisc0.ameer,

I try to answer to your questions

1)

my question is Why we redistribute BGP into IGP protocol ?

Do we do it because those route(s) carried by BGP(with network command) , we want to injected into IGP?

 

Yes I agree the redistribution of BGP into each IGP is for providing the routes into the routing domain of the receiving IGP.

 

2)

with using Network and aggregate what's need for summary-address ?

and what happened if We don't use summary-only keyword?

 

If you don't use the summary-only keyword BGP advertises the summary route and all alive component routes.

So the summary-only is needed if you don't want BGP to propagate component routes in addition to the summary route.

 

summary-only will be used  
whenever we have bunch of Prefixes 
10.1.0.0/16
10.2.0.0/16
10.3.0.0/16
.
.
.
and just sending summary only but condition is we have this bunch routes in our Table [in this case BGP table] correct?

Your understanding is correct the aggregate-address command requires that at least one component route to be in the BGP table in the local node by any means (network commmand, learned by another BGP peer or the result of a redistribution on local node)

 

>> 3- sorry for this silly Question : which prefixes should be advertised through a Network command ? those we want to aggregate or NO some internal we want other Side have that too.

 

First of all, with new BGP default of no auto-summary the network command requires an exact match to work so we need to advertise at least one component route with the exact mask it has in the IP routing table.

Advertising multiple component routes is recommended to avoid the summary route state to depend on a single component route.

How many component routes to advertise depends on the usage or not of the summary-only keyword.

We don't need to advertise all the existing component routes if we are going to use summary-only.

A subset of network commands with exact match on subnet mask is enough, because the summary route will cover all the missing component routes: it is used to represent all the address block.

The sending of some component routes is required when we have multiple paths / neighbors and we would like to perform load sharing of inbound traffic based on component routes: on Router RA we send out the summary route plus some component routes (we can use the suppress-map to invoke a route-map to decide what component routes to send)

on RB we can send the same aggregate and a different subset of component routes (again using a route-map invoked in a suppress-map option in aggregate-address command).

Because the most specific route is used RA will attract traffic directed to the component routes it advertises and the same will happen on RB.

 

Hope to help

Giuseppe

 

View solution in original post

Highlighted

thanks too much