During the test o mpls vpn,field engineer notices that it's impossible to exchange private route and ping between two PEs.
The part of topology is as follows：
Each PE has one loopback interface,and bind a vpn instance test.
The loopback addresses are 22.214.171.124 and 126.96.36.199
We check the possible reasons one by one：
1,We already config the OSPF between the two PEs,there is loopback address of the other PE in the routing-table.
2,Check the mpls and lsp,no problem.
3,Check the BGP and BGP peer,no problem.
4,There is peer 1.1.1.X enable command in the view vpnv4,using dis bgp vpnv4 all peer, the peer and the sending of private route 1.1.1.X are all OK.All the possible reasons are clear.
Check the private routing-table again,find that there is only one route:
This is its own loopback address.At this time,we find that the mask length isn't 32 but 24.Check the config,it's 24 indeed.
In this case,the two loopback address of the two PEs are in the same subnet.Actually,the router has received the private route,but it's in the same subnet of its own loopback address.So the router regards them as the same route got in different ways.One way is direct,the other is BGP.Then the router choose the direct because it has higher preference.So there is no BGP private route in the routing table,then we can't ping.
Changing the mask length to 32 solves the problem.
From the penomenon we can figure out that's due to the config,the possible reasons are as follows:
1.No loopback address of the other PE in the routing-table.
2.Wrong config of mpls.
3.Haven't configured command connect-interface-loopback,so the peer isn't established.
4.In the view of vpnv4,there is no command peer 1.1.1.X enable,so PEs can't exchange private route.
During our configuration,we have to pay attention to details.There are many critical problem just caused by the neglect of details.