This topic is extremely important for our blog followers who are interested in learning more about Service Provider’s technologies and new features which are introduced to optimize the design of any ISP network.
We all experienced studying MPLS the most power full technology which you may spend time in reading 3 or 4 books for understanding what will happen after writing “mpls ip” command under an interface, WELCOME TO NETWORKING !!
This introduction is very related to what I am going to explain here. mpls ldp sync is a feature that firstly introduced in RFC 5443 aiming to reduce or eliminate the packet loss that may happen if we have a broken LSP.
We know that the IGP “ISIS/OSPF” chose the best path depending on the SPF calculations of the metric, so lets assume that our P router has an interface that is not mpls enabled and the IGP chooses this interface as the best path for the VPN traffic according to the metric and SPF calculations, so when a VPN traffic received at that P router it will find that the forwarding interface is not mpls enabled and will strip out all labels then start sending native IP packets and sure the next hop will drop all these traffic as it does not have the destination in its global RIB, here comes the power of mpls ldp sync the IGP will sync with LDP session, so that it will send the maximum metric for all links until LDP comes up.
This feature guarantees that when LSP is broken the IGP will exclude it from the RIB and only uses the complete LSP to prevent any packet losses that would happen for the VPN customers and may the ISP engineers will not recognize it and start troubleshooting on the CPE side.
What could be the reason?
– Forgetting to configure mpls ip for an interface.
– IGP is converged before the LDP.
– Or any reason could make the LDP not up such as: neighbor router ID is not propagated into the IGP or not valid…etc.
Recommended solution:
– Sure using mpls ldp sync feature will reduce the impact by avoiding the broken LSP, but using mpls ldp autoconfig beside mpls ldp sync will not only reduce the impact but will avoid it at all, also will reduce how many times you have to write mpls ip, so annoying 🙂
Configuration example on Cisco IOS:
| R1(config)#router isis 1 R1(config-router)#mpls ldp sync R1(config-router)#mpls ldp autoconfig |
| R1(config)#router ospf 1 R1(config-router)#mpls ldp sync R1(config-router)#mpls ldp autoconfig |
