Hi Community Members,
I´ve a design-related and comprehensibility question I´d like to ask everyone here in the forum who is interested in routing!
I`m currently dealing with a requirement in which 2 locations are to be connected directly and on the otherside via MPLS from an ISP (= network diagram in simplified form attached)!
The IP Traffic between location A <-> location B (e.g. 10.1.1.0/24 <-> 10.2.2.0) should primarily pass directly and all other traffic should use the default route which is accessible through the ISP MPLS! If an ISP link or the direct connection fails, a corresponding backup routing (e.g. 10.1.1.0/24 <-> 10.2.2.0/24 statt direkt- über MPLS) should take place...
I have the following questions that I hope you can help me with with your feedback or input... thank you very much in advance for any kind of answer!
My questions so far:
-> is OSPF in general the "right/best" dynamic routing protocol to use it in that case here?
-> is the OSPF design with multiple areas the "right/best" design to use?
-> how can a potential routing loop be avoided when redistributing the IP prefixes announced from OSPF into BGP?
- it`s not possible to influence the routing of the ISP/MPLS
- static routing is not an option due to scalability
- the BGP backdoor feature should´nt be used due to scalability
This way by default without even introducing any traffic engineering, location A/B traffic will go direct between each other (shortest As-PATH, and all other traffic will traverse via MPLS?
@MHM Cisco World yes - correct, each of the routers connecting to the MPLS as well as the ones connecting to the direct link should announce the local connected 10.x.x.0/24 networks; the 0.0.0.0/0 will be learned only via the MPLS routers and should be used from the locations primarly but in case of a problem, each location should be able to access ressources with the default IP-Prefix as backup through the respective other location!
I would have set the distance for OSPF lower, the question for me now is whether I should use OSPF at all or go with eBGP as mentioned by @paul driver
M02@rt37 in my case each area will contain 2-3 routers so no large network, but the point with the control of LSA advertisments is something I´d in my mind...
I thought about using route-maps with communites - eg.
route-map BGP-LOOP-PREVENT-OUT permit 10
match source-protocol ospf 1
set community 1:100
route-map BGP-LOOP-PREVENT-IN deny 10
match community 1:100
... maybe that will work as expected?
You can not make bgp not advertise the prefix' in this situation you end that prefix learn only from ospf' no backup.
Keep all config only change AD of ebgp'
Ebgp is 20 and ospf is 110 so router use lower AD ebgp but if we make ebgp higher than ospf we can force router use bgp not ospf and insame time if it failed then router will use bgp.
NOTE:- cisco recommend not modify the iBGP keep is AD 200
NOTE:- this will effect all prefix learn from BGP
this lab for you, no need any other filter, what you need only change the AD of eBGP to be higher than OSPF (110),
attach show you how router change protocol for learn prefix 184.108.40.206 from bgp to ospf after I config distance under the BGP
Using OSPF with multiple areas can be a good design choice, especially if you have a large network or want to control the scope of route advertisements. Areas help in organizing the network, reducing OSPF database size, and improving routing efficiency.
When redistributing OSPF routes into BGP, it is crucial to prevent routing loops:
--Use route filters or route maps to control the routes being redistributed.
--Implement route tagging or attribute manipulation to differentiate OSPF routes from BGP routes and avoid loops.
--Implement proper route redistribution policies and filtering to prevent unwanted routes from being redistributed.
--Utilize route summarization and aggregation techniques to reduce the number of routes being redistributed.
if you don’t want to start messing around changing AD values and mutual redistribution filtering then I would say keep is as simple as possible.
Then at each site create a IBGP peering between each site rtr, EBGP peering for rtr 1 for MPLS and EBGP peer rtr2 towards location x
Each site rtr will be aware of the others ebgp connection, they will each advertise the local lan subnets to their respective ebgp peers, thus provides resiliency for the site incase either the MPLS or LES link fails and at the same time both will only advertise a default route into the L3 core ospf switch for egress traffic to exit the LAN, and you would not need to traffic engineer location traffic to traverse the les link it will do that by default due to the bgp best path selection process.