our company ( we are not a MPLS VPN provider) has self deployed MPLS i.e. EoMPlS, MPLS TE & fast reroute for FCIP and MPLS VPN for Guest network.
We also have a need to distribute our labels to remote sites so that we could deploy EoMPLS. remote sites are connected via a MPLS VPN provider which provides a BGP handoff.
i tested a design where labels would be advertised to ISP-MPLS-CLOUD and then to remote sites via BGP. ( "neighbor x.x.x.x send-label" on CE and PE BGP process).
(all links in campus - from distribution block to core swtich and from core switch to CE router are enabled for tag-switching.)
I tried to implement that design with the ISP's help yesterday: i was able to extend labels to remote sites successfully, however one remote site created a big problem. From the headend CE when i traced route to remote site core switch i was seeing very odd behavior. There was a loop: packet was going out the PE router and then back to CE router ( ce router has a remote site(Z) connected via point-to-point T1 and it is NOT MPLS enabled) and then to remote site Z and then back to the CE router and from here the same trip again.
once i reverted back the changes every thing went back to normal.
do i need any of following commands:
?mpls ip default route:
- we are advertisint a default route to all remote sites.
? do i need to configure "neighbor x.x.x.x explicit-null"
Any other ideas - i have not been able to re create this problem in my dynamips routers.
appreciate any input