Hi All,
I am facing issue to understand the EIGRP SSO attribute in MPLS VPN senario.
Lets consider the senario
CE1-----PE1------PE2------CE2 and CE1----CE2 (backdoor link)
CE1 loopback (1.1.1.1/32) and CE2 loopback(2.2.2.2/32).
I am trying to explain the race condition below.Please correct if I am wrong.
When the CE1 loopback goes down it will send the query message to both PE1 and CE2. PE1's current successor is CE1 and having no EIGRP neighbor so it will send the unreachable message back to CE1 and remove the route from topology table and routing table.This will cause the BGP to send a withdrawn message to PE2.
When CE2 gets the query message from CE1,Since It is already having matching route from PE2, it will send back a reply immediately back to CE1 about the route and CE1 install the new route via CE2 and send the update to the neighbor (to both PE1 and CE2).So PE1 router will again install that route and send update message to PE2.
Lets say after this, the withdrawn message from PE1 reached PE2,so PE2 will remove the route from BGP table and redistribution to EIGRP will not happen.since the route is lost, PE2 will send the query to CE2 and CE2 will propagate the query to CE1,since CE1's current sucessor is CE2,he will propagate the query to PE1 and PE1 will send the unreachable back to CE1 and send the withdrawn to PE2.
This loop will continous untill the maximum hop count reach.
How can EIGRP SSO avoid these kind of loops .(Expecting detailed explanation)
Thanks,
Arun Mohan