This morning a client has asked me to assist them in attempting to redesign the way a redundant path to a Business Partner is working.
The client has a HQ site in location A, and a DR site in Location B. There is a server in Location A (Primary Server) and a server in Location B (Secondary). These sites are connected together via a MPLS link.
There is a WAN connection to the Business Partner in Location A, as well as a WAN connection to the Business Partner in Location B.
The current configuration works so that both the Primary server in A as well as the Secondary server in B use the WAN connection in A to connect to the Partner. Unfortunantly in the current config, it seems that the Secondary server in B cannot connect to the Business partner over the WAN connection to the Partner in Location B.
I am seeing on the WAN routers at the client that they are getting EIGRP updates from the client in both locaitons. I think that both WAN connections are T1's even though they are from AT&T (Location A) and Verizon (Location B).
The client has asked me to make recommendations for how to configure so that the Primary server in Location A will have a preference for the WAN connection to the BP in Location A, but would also be able to use the connection in Locaton B if the WAN in Loc. A went down.
Likewise, they want the Secondary in Location B to have a preference and be able to communicate to the BP via the WAN in Locaiton B, but in the event that it fails, that the Secondary server in Loc. B could use the WAN in Loc.A to communicate back to the BP.
I would think that if the BP is advertising his networks to us, that EIGRP would know which one to use. Here are the EIGRP statements from Loc.A and Loc. B:
router eigrp 13
network 172.27.6.128 0.0.0.15
neighbor 172.27.6.130 GigabitEthernet3/0/15
distribute-list EIGRP_OUT out GigabitEthernet3/0/15
distribute-list route-map eigrp-rm out
distribute-list eigrp in
router eigrp 13
network 172.28.6.128 0.0.0.15
neighbor 172.28.6.130 GigabitEthernet1/0/4
eigrp stub connected summary
The 172.27 and 172.28 networks respectively are the networks that connect my client to the BP. The destination networks at the client site are the
I am curious about the "eigrp stub connected summary" statement in Location B, and if this may be why EIGRP does not think that it should send data to the 206.223. networks via its own G1/0/4 interface to the BP.
Thanks for any help on this issue.
Here are the change recommendations that I am currently considering publishing to the client. Please confirm Jon:
1. We do not need to have the 206.223 networks in our Router eigrp 13 statements. This is because "the BP" advertises these to us.
2. We may not want the EIGRP "stub" statement on the EIGRP config on the DR side.
3. We may want to add the "no auto-summ" statement on the DR side so that we ensure that 'classless" BP networks are correctly advertised.
let me know if you see anything else I should explicitly recommend to the client.
1) Yes, i can't see why these are needed and if there are no interfaces on the HQ/DR routers using these IPs then they are not being used by EIGRP either.
2) If you want to have HQ A use DR B if HQ lose their connection to the BP site A then you will need to remove the eigrp stub config because otherwise DR B will not advertise the 206.223.x.x networks to HQ A.
A word of warning here - it's always best to assume that the config is there for a reason. Removing eigrp stub may mean that other subnets are then advertised to HQ and you may find that routing for other subnets then takes different paths. I suspect this may be there because they don't want DR advertising any HQ learned routes out of any WAN links it has. Now you obviously need to advertise the HQ server subnet out of DR for failover but be careful of any other networks that may also be advertised. It's tricky to say without knowing the full setup but you may find that all of sudden traffic starts using DR and not HQ even if HQ is up. So this does need a bit of investigation.
3) Yes you need to enable no auto-summary but again as above you need to be careful.
What you may have to do is use a route-map/distribute-list at DR under the eigrp config to only advertise out those networks you want to such as the server subnet in HQ. Sorry to be vague but you do need to be aware of the rest of the routing.
As for the BP. It looks like they are not advertising their 206.223.x.x networks as /24s but are advertising smaller subnets. So you need to see exactly what they are advertising and just as importantly how they are interconnected because in effect you are trying to emulate at their site what you want at yours ie.
HQ A = BP A and DR B --> BP B under normal circumstances but then each can use the other as failover.