Is it good practice to run multiple routing protocals over same p2p links acrross 2 locations. Ex:if I have 3 different conenctions across 2 locations, can is it recomended to run BGP/EIGRP over those links? (forming neighbours & advertising same networks via both protocols). My guess is yes, but would like to confirm.
Thank you in advance
Not sure i would call it good practice. Is there a reason you need to run 2 routing protocols to advertise the same networks. Not only are you placing additional overhead on the routers but you are also using extra bandwidth which doesn't come cheap.
if you do this during a migration to avoid redistribution it can be a good thing.
If you have decided to use BGP normally this implies that you want to keep the IGP local to each site in this case you can
advertise IGP site subnets using network commands.
redistribute BGP into IGP to learn other sites' routes.
Or you can choice to use only one IGP, EIGRP in your case and this is a good choice too.
All is dependent on your scenario: size of each site, multivendor context (where EIGRP cannot be used), a migration phase can lead to run two routing protocols over the same links.
For example if you are going to move from dedicated links where you have only EIGRP to a L3 MPLS VPN service you may consider moving to BGP on the inter-site links where in the near future the BGP session will be with a Service Provider PE node.
This can be seen as a way to be ready to switch to L3 MPLS VPN.
Someone could see running two routing protocols as a form of redundancy on the signalling plane: should one routing have a problem the other one is there to provide a backup.
Hope to help
Thank you both for your quick replies and sorry for the late reply. I was out. Here is the scenario in nutshell:
Sites A, B and C. Site C is Main data center and OSPF is main protocol.
A& B connects site C via gig links and OSPF being used. between A and B there are 3 MAN links and both BGP and EIGRP has been configured via those links (for both protocols #3 neghbors -one via each link). those networks. Final requirement from client is...
--> A & B should talk each other directly
(BGP will do that)
--> Incase link between A&C goes down, A should be able to reach C via B and incase the B&C goes down, B should be able to reach C via A.
--> Incase 3 links between A&B goes down (which is may not happen) the reachability will be via A <-->C<-->B.
I know its hard to give a solution with out giving the existing configs here, but in general my undesranding is, with OSPF & BGP (EBGP between A&B) with OSPF redistribution in to BGP and little tweaking of BGP weights (as A&B got one L2/L3 device)the above can be achieved without EIGRP inplace. Any suggestions..?
Thank you in advance
It's is difficult to comment without full topology details but is there any reason why you can't just run OSPF between A & B ?
If you did this then each site would have 2 paths to every other site. Because of the direct connections between A & C and B & C both A & B should go direct to C for data centre access.
You may need to tweak the cost of the links between A & B so that this path is favoured for direct communication between A & B or you may not - you don't say how fast the links between A & B are.
I am perhaps missing something or there is additional information needed but the requirements you have outlined above could be met with just one routing protocol as far as i can see which would certainly make things easier.
Thats a great point. I by tweaking the ospf path cost on interfaces , we can control the paths. Is that correct..? also, the paths..
Between A&B the Bandwidth learned on the i/f:
Path1:Primary:1000000 Kbit <--> 1000000 Kbit
Path2: Sec: 900000 Kbit <--> 1000000 Kbit
Path3: L3 Switch to RTR <-DS3-> RTR to L3 Switch
between A&C / B &C its Gig path.
Not sure why both BGP & EIGRP added here. I was thinking BGP is needed but after your posting looks like it is not needed eighter.
Thank you for the information.
as Jon has correctly noted you can easily fulfill your client's requirements using only OSPF with little tweaking on OSPF costs
on the specific you need to make
DS3 path between A and B preferred over the A-C + C-B links to avoid that link is never user when first two paths had failed.
under each router and multilayer switch
router ospf 1
auto-reference bandwidth 10000
to make reference bandwidth 10 Gbps otherwise all links with BW >= 100 Mbps cost 1 regardless they are FE or GE or more
do it in all routers and multilayer switches.
then using show ip ospf interface take note of cost of path 3.
to make path 3 preferred over A-C + C-B you need to increase ospf costs on A-C , C-B links
do this on both sides using
ip ospf cost value
2*value > cost L3switch to RTR + cost DS3 link
All this under the hyphotesis of having all links in the same OSPF area so costs can be compared.
Hope to help
Thanks for the reply.
"All this under the hyphotesis of having all links in the same OSPF area so costs can be compared".
But all the locations A, B and C are in different areas. So the interconnected links (5 in this case,#3 between A&B, 1 each to C)
will not be under same area. Any additional suggestions.?
the locations can be in different non backbone areas while the backbone links can be in area 0 or are you using any form of MPLS L3 VPN ?
have you got L2 services over the GE links or L3 VPN?
if they are L3 VPN and only the DS3 allows to build an intra-area connection you are in a trouble with OSPF because intra area routes are preferred over O IA routes.
if all routes of interest are O IA you can still play with costs to achieve the desired behaviour (the problem should become that of managing costs to ABR router-ids ip addresses)
Hope to help