In the lesson, when configuring MPLS you state “Once again I will configure iBGP between PE1 and PE2 but this time I will use their loopback interfaces. You will see why in a minute”, however I’m not seeing why you decided to use the loopback interfaces, please explain.
Also, I created a 5-router lab just as the one you’re using in that lesson and tried to do MPLS but configuring iBGP using the physical interfaces along with the next-hop-self option and couldn’t get it to work. I then changed the configuration to use loopback interfaces and still couldn’t get it to work. Then I noticed that I did not advertise loopback 0 on router P and once I advertised that network all of a sudden MPLS started working. It’s as if it requires the loopback interfaces of all routers to be advertised, is this in fact the case?
The main reason to use loopback interfaces is that loopbacks can’t go down unless you add a shutdown command to them. Physical interfaces can go down. When you use a loopback interface for iBGP and one of your physical interfaces goes down, your IGP can always find a new route to the loopback interface (assuming there is a backup path).
A loopback is not a hard requirement. The only thing that is required is that you have connectivity between the IP addresses that you use to establish your iBGP peering. If you use loopback interfaces, you need to advertise them in your IGP otherwise they are unreachable.