MPLS BGP Scenario based tricky interview question

[Naive]

Hello Everyone,

I need your help to understand one scenario.  Below is the simple MP-BGP scenario, 

Where MP-BGP configured between PE1 to PE2 through their loopback addresses. and everything working perfectly. 

But then you delete this neighborship and make a new MPBGP neighborship with PE1 and PE2 's inside interface.

Then what are the reasons to MPBGP not came up with new interfaces or CE1 can't able to reach CE2. 

When Physical connectivity, IP routing is good.. 

[Harold Ritter]

Hi @Naive ,

It is always strongly recommended to establish the MP-BGP session using the loopback interface. In your scenario, you established the BGP session using the core facing interface address. The issue is that this subnet is shared with the P router and the P router therefore sends an implicit null label for that subnet to ingress PE. Having received an implicit null label, the ingress PE does not push a service label. This breaks the end to end LSP between the ingress and egress PE and therefore the IP connectivity between CE and CE2.

,

Regards

[Naive]

@Harold Ritter  Thank you so much for your swift response.

Now I Got some new questions from this. 

• I believe P was doing implicit null operation before too[when loopback neighborship], why on that time there was no ping drop.
• If I insert 1-2 more P routers in this scenario[between PE1 and PE2] and do all required configuration according to P router, after this can CE1 Ping to CE2.

This is just a hypothetical situation, I am not going to use it in real network.

[Harold Ritter · Accepted solution]

Hi @Naive ,

> I believe P was doing implicit null operation before too[when loopback neighborship], why on that time there was no ping

> drop.

P sends implicit null for its connected prefixes, but not for the PE's loopback interface, for which it will send a label.

> If I insert 1-2 more P routers in this scenario[between PE1 and PE2] and do all required configuration according to P router,

> after this can CE1 Ping to CE2.

This does not change anything to the described problem. In this scenario, the P router directly connected to the egress PE advertises the implicit null for the subnet shared by the egress PE and the P router and causes the upstream P to perform a penultimate hop popping (PHP), which causes the end to end LSP to be broken.

Regards,

[Naive]

Thank you so much for your explanation