We are going to convert to BGP, However there are few things I am not sure which road I should go down at the moment, it would be grateful if anyone here could give me some suggestions.
We have two ASR routers connecting to Transit provider via EBGP sessions and receiving full Internet routing table.
1. EIGRP will be runn on all routers within the AS to provide detailed routing, currently there about 500 internal routing entry within the AS
2. iBGP sessions between ASR1-ASR2 will be propagating full internet routing table that each ASRs take from transit provider
3. iBGP between ASR1 to 6500-1 will be
ASR1 sends default route to 6500-1 with Local preference 130
6500-1 doesn't announce anything to ASR1, as the internal routing will be carried by EIGRP
4. iBGP between ASR2 to 6500-2 will be
ASR2 sends default route to 6500-2 with Local preference 130
6500-2 doesn't announce anything to ASR2, as the internal routing will be carried by EIGRP
5. iBGP sessions between 6500-1 and 6500-2
6500-1 announce default route to 6500-2 with Local preference 90
6500-2 announce default route to 6500-1 with Local preference 90
6. hot standby protocol running between 6500-1 and 6500-2 with 6500-1 acting as active router for most of the subnet
1. would this design work?
as from I've read at the moment, that the iBGP needs to be fully meshed, however I can't see there is any reasons I need to establish an iBGP sessions between ASR1 and 6500-2 or ASR2 and 6500-1.
And also is it a good idea to use the EIGRP to progate the internal routing as there would be lots of changes, will this make the network unstable in the future?
2. for future expansion purpose, for instance if we add another physical site with a pair of 6500 and connecting to the ASR router core in the same manner,
should I use route reflector and make 6500-1 as the ASR1 client and 6500-2 as ASR2 client?
Any response will be really appreciated. Thank you all.
1. It is too difficult to say Yes or No. You IGP design is rather unclear - do you 65 pairs (1-2 with 3-4) have another way to talk to each others except through ASRs? How ASRs announce your AS'es networks toward ISPs?
2. Yes. iBGP never sends any route that received through iBGP itself. You might want to make 65-1 rr-client of 65-2 and vice versa. Same way for 65-3 and 65-4.
3. As you do not want any BGP routes to come to ASRs from 65s, you do not need to make anything towards ASRs, I think.
I probably made it too complicated, so I have updated what I have put down with a simplified network, can you take a look again please to see if that make more sense. Thank you.
Yes, that is good.
If you want to use iBGP between your devices you need full mesh. Default that is generated by ASR1 will come to 65-1, but it does not send it to 65-2 not being a route-reflector. Same thing is about ASR2 and 65-2. So, you might have 65s to be a route-reflector-clients of each other's.
In my opinion it is better to advertise your networks in BGP on 65s, not on ASRs, but this is not very importaint.
I dont think any route reflection is required here. because both the 6500-1 and 6500-2 receive a default route via ASR1 and ASR2. each. Now the 65's routers are in HSRP. So only one router will be forwarding at a time.
Lets say 65001 is Active HSRP router. All the traffic from the LAN will use that router and the traffic wil lflow from it to ASR1k and above.
Now lets say that the lnk between ASR1k and 6500-1 goes down then HSRP will failover to 6500-2. It will become the Active router and it wil start forwarding the traffic to ASR2 as it receives the default route.
Yes, it's an idea. But David wanted to get two defaults on each of 65s - one from upstream ASR, another from second 65, which in turn got it from upstream.
Thanks for the reply.
If I am not using iBGP to carry the default route 65-2 learned from ASR2 to propagate to 65-1 But only use
neighbor default-originate command with lower local preference, would that eliminate the requirement for the RR?
Thank you all again.
if you don't want to remove EIGRP in the near term you could think of avoiding the iBGP sessions between ASRs and C6500 devices and you could inject a default route in EIGRP instead with both ASR devices.
There is no added value in having those iBGP sessions just to provide a defaut route.
If you want to remove EIGRP, it is better to have a full mesh of iBGP sessions between the four devices and the C6500 have to advertise the internal networks in iBGP as noted by Sergey.
A design wiith RR is possible, but to achieve redundancy both C6500 should be clients of both ASR in order to be able to exchange iBGP routes between the two C6500 even if one ASR has failed.
See this as an additional safety measure.
Hope to help
Yes, probably this will solve your problem. But also look at Giuseppe's post. May be really it is better use EIGRP for default?
Hi Kishore and Sergey,
Thank you very much for all your replies guys. Really appreciated.
what Kishore said is what I intended the network to do.
However I was confused that do I actually need to make the network as full mesh network for iBGP, such as establish an iBGP sessions between ASR1 and 65-2. I can't see a must requirement, but everything I've read so far, it says it need a full mesh iBGP environment otherwise you will need RR or confederation.
But given the fact here I am only using default route within iBGP, I can't see why do I need a full mesh, I am really confused.
As far as I can see you dont need to have a iBGP full mesh for your case. You dont need to use IBGP full mesh for every scenario. It depends on case by case topology.
If you are happy with the answers you got please mark the question answered so that others can benefit from it :-)
It is an iBGP requirement that a full mesh should be done between neighbors so yes you would need it if you want BGP on all four devices otherwise you will run into routing issues. The is no much to understand here, it is the way that BGP works.
RR is there to minimize neighbor sessions by bypassing the rule stated above. RR is really useful when dealing with a loot iBGP in networks. You only have 4 devices on yours so I don’t think you should consider RR in this scenario.
The real question here is do you need to run BGP on all devices? I don’t think so, if you follow Giuseppe's advices you will have a more simple routing topology if you let EIGRP deal with default routes.