Default route in iBGP dual homing environment

[paolo.acquaviva]

  Hello,

The scenario that I have is two CPEs connected with internal link (iBGP routing) and then both are connected to different ISPs. I have a doubt about the implementation of a prefix-list in order to allow the announcement of default route 0.0.0.0 to each iBGP peer. I think it's important to have a failover when the internal LAN is not reachable for the active HSRP router.

Another doubt is regarding to the redistribution of static routes in iBGP. What do you think about it?

Thank you very much.

Best regards.

Paolo

[Jose Jara]

Paolo,

As far as I understand, you will have a Dual CE running BGP and you will receive the 0/0 from different ISPs, and in the LAN you will have HSRP + static routes.

In that case, you will need to redistribute the static routes towards eBGP or via network command. It is not necessary via iBGP because each peer will have the static routes configured locally. The default route will be received via the WAN and you will need to pass it via iBGP, in case there is a failure of HSRP tracking or the LAN interface. If you are just receiving the 0/0 via the WAN, you do not need any filter in the iBGP session. But, if you want to allow just the default, assuming you're receiving other routes, you may apply the prefix-list allowing just the 0/0:

ip prefix-list DEFAULT permit 0.0.0.0/0

Then, apply it to the neighbor:

neighbor iBGP prefix-list DEFAULT out

Hope this helps,

Jose.

[paolo.acquaviva]

Hello Jose,

thank you for you answer.

The scenario that I have is composed by two CE connected to different ISPs. There is a iBGP session between them and eBGP sessions between their respective ISPs.

Now, we have a prefix-list configured as you said but my doubt is if this default route exchange is really needed.

Another question is if we have to redistribute into BGP internal IP addresses.

Let me know if you need further clarifications.

KR.

Paolo

[blau grana]

Hello Paolo,

Exchange of default routes between iBGP peers is needed, because of failover. If exchange is no permited and internet link between ISP1 and CPE1 fail, CPE1 will not have information that internet is accesible via CPE2->ISP2.

From your description I understand that you have not any IGP in your LAN, am I right? It could change some things.

Also you mentioned some static routes, what are they for?

Best Regards

Please rate all helpful posts and close solved questions

[Jose Jara]

Paolo,

as blau grana states, the exchange of the default route is needed.

If you have an IGP running in the LAN, it will be needed to do bgp redistribute-internal as you are running the iBGP with a different connection than the LAN, in case the LAN interface of the Main Router (assuming a Primary-Backup scenario) fails. Be careful with that, controlling what prefixes will be needed to advertise to BGP.

Otherwise, using static routes, nothing more is needed.

Best Regards.

[paolo.acquaviva]

Hello,

Thank you very much for your support.

But I still have a doubt regarding the importance of default route announcement because if we already have the network summary echanged between CPEs, each of them knows the backup path. Am I right?

The redistribution of static routes are used to reach some IP addresses not declared in BGP and to reach the L0 of the respective iBGP peer.

Please let me know what fo you think about the importance of default route announcement in my scenario.

Kind regards.

Paolo

[blau grana · Accepted solution]

Hello Paolo,

But I still have a doubt regarding the importance of default route announcement because if we already have the network summary echanged between CPEs, each of them knows the backup path. Am I right?

Do you receive from your ISPs just default route, partial or full BGP? If you receive just  default route or partial BGP table, you need to advertise default route between iBGP peers. That way one iBGP peer will be backup for another iBGP peer and vice versa.

The redistribution of static routes are used to reach some IP addresses not declared in BGP and to reach the L0 of the respective iBGP peer.

You answered your question yourself -> to have full reachability in your LAN network, you need to redistribute static routes into BGP (redistribution, network command).

Static routes for reaching loopbacks of iBGP peers does not need to be redistributed into BGP.

Best Regards

Please rate all helpful posts and close solved questions

[paolo.acquaviva]

Hello Blau,

Thank you very much for your answer. I have for you the last question. To reduce the eBGP convergency time could I configure the BGP fast external fallover command on each CE router? Is there any technical problem with this scenario?

Best regards.

Paolo

[blau grana · Accepted solution]

Hello Paolo,

There is no need to configure it if you did not disable it. In cisco IOS it is enabled by default, but remeber that this feature works only for directly connected eBGP peers.

Best Regards

Please rate all helpful posts and close solved questions

[paolo.acquaviva]

Hello Blau,

Thank you very much for your support.

Have a nice day.

Best regards.

Paolo