Cause for interruption of some services:
After g5/0/0 on NE2 was down, NE2 failed to receive OSPF packets, and OSPF neighbors between NE1 and NE2 were down.
RT1 and RT2 controlled network-to-user traffic based on the MED value in BGP routes. Certain traffic from NE1/NE2 to RT1/RT2 need to pass between NE1 and NE2 before arriving at RT1/RT2. When OSPF neighbors were down between NE1 and NE2, IGP routes were re-iterated by IBGP routes for the traffic. Therefore, the traffic was routed to SRs or BRASs connected to the NE5000Es. The SRs and BRASs, on which the route to NE1 and that to NE2 had the same cost value, sent the traffic to NE1/NE2. As a result, a rooting loop occurred between the NE5000Es and SRs/BRASs and the services corresponding to the traffic were interrupted.
Cause for the down state of OSPF and LDP neighbors:The TTL value of a packet decreased by 1 each time it passes a router. When the TTL value was reduced to 1, NE5000E sent the packet to the CPU of the LPU board. Because protocol packets were sent a dedicated channel, no OSPF packets were lost. Upon receiving OSPF and LDP packets with the TTL value being 1, the LPU board put the packets into an IP_INPUT queue before sending them to processing modules. Query results indicated that the CPUs of LPU 9 and LPU 12 of NE5000E were high. Once the queue was full, protocol packets were discarded and protocol timeout occurred.
1. Delete Eth-trunk1 and create two OSPF neighbors.2. Set valid-ttl-hops to 1 for IBGP neighbors, so IBGP neighbors are down upon OSPF down between NE5000Es. In this manner, the routing loop can be prevented.