Wed Feb 29, 2012 11:41 pm
I have had same problem. in my setup, it was due to wrong labeling. I also had more than one point with same MPLS label, causing this stuff.
ospf is indeed buggy, but its quite ok, if u tweek it a bit. Its still problems with different maste flag, if link between some of the routers drop. I think. it maybe due to some bug in LSA not timing out, and that the LSA gets different value after a state change, but other side, dont see the new lable untill old somhow times out (not scientificly proved, but a taught)
Why all this use of pppoe sentralised, (eoip, vpls etc) when its a nice feture to run pppoe on each AP, and use BGP to let the central router know about customer IP. central radius is still also possible. Then all hassle with eoip, vpls etc is minimal. in my setup, i have subnets to a router (lets say just /21 to make it easy) addresses in this subnet is by a managment software, given to spesific customer. Its updating a database, and tells the same to radius. My ap is connected to radius, and get information about ip/user/pass/ques etc. Customer logon to AP, get an ip from pppoe server, AP makes an interface, and BGP redistrubute this trough the MPLS net to the central router.