Ok riddle me this (I think its a V7 bug tbh) and I can't reproduce it as i've rebooted the router and everything working, but do me a sanity check and tell me how this is even possible
Router A has a 0.0.0.0/0 link with a cost of 4 and it is using it
Router B has a connection to Router A and that link has a cost of 10
Router B also has other paths through the network it could use, backup transit paths, the closest of which has a cost of 84
Everything is in Area0. Everything has a valid adjacency and peer states are all Full
RouterB decides to use the backup path with a cost of 84....... for everything, including the loopback of RouterA. It just outright refuses to use the path with a cost of 10/14. No amount of bouncing the link works, only after rebooting RouterB does it finally decide to use the lower cost path
As said I can't reproduce this problem as the reboot fixed it (no config changes obviously), i'm chalking this one up to a bug. But is there any scenario i'm not thinking of with the above setup that could cause this? ALL routes were using the backup, even the loopback of RouterA was going via an extremely long backup route. I did not see any LSA's from RouterA in the table of RouterB or vice versa, so its like they were in the Full state but no actual LSA exchange happened, and no errors were shown in the log
RouterB is running 7.23.2 and RouterA is using 6.48.6
Where is the /export of the devices?
text description = no sanity check
also because you might be typing it incorrectly, or maybe you think it's that way, but the configuration is different...
Fair point, however as this is OSPF and in the backbone area there's really not a lot of ways to manipulate routes even if I wanted to. By design of the protocol itself I can't do route summarization or manipulation other than what gets introduced into the LSD at those 2 specific routers. And since this is a default route that originates from an ASBR upstream of RouterA I can't even do anything to that
My question is more one of theory 'Is there anything about the OSPF protocol i'm not aware of that could cause this situation, even if I tried'
In any case here are the trimmed configs with relevant sections. Note that though RouterA has MPLS-TE area configured in the OSPF instance. It does not have any MPLS/LDP config at all and is not participating in MPLS. Neither is RouterB - it has some legacy config but LDP is turned off - nor any routers in the longer backup path between them
The other routers aren't relevant, as otherwise the network itself would have been down. The only situation here is that RouterA and RouterB despite having a Full adjacency, did not route via each other at all. They chose the much longer backup route for any and all routes until I rebooted RouterB