Troubleshooting (not so) advanced routing topics

So...

I was stuck into this MPLS scenario for about 4 days. Really. That's an eternity...

It turns out a router on the middle of the path was blocking TCP 646. TCP 646 is LDP port for label exchange. It's the port used to exchange labels between peers. If peers can't exchange labels, it doesn't matter of course, if they are neighbors.

I had seen the output of the show ip cef for that vrf, but that didn't help me much, because I had no idea there was an ACL blocking LDP in the middle and I had no idea that this was causing the issue. That was nasty (for me at least).

Lesson of the day is - your routing protocol runs on top of IP to exchange data? Then pay attention to the chit-chat, gods above and below. Had I seen this 4 days ago my weekend would have been happier.

Worst of all, I rewatched all MPLS videos and still couldn't get it. It was only when I found a similar discussion and analysed it that I got it working. Because I found an ACL in the path, took it off, put it back on and logged the drops. Horrible reverse engineering, I should be better than that.

Anyway, moving on.

Mixing OSPF interface network types.

So, curiously, and contrary to CCNP (R&S) knowledge, you can mix and match OSPF interface network types.

You need to pay attention to:

1) Timers
2) DR/BDR electiveness

For the first topic, quite self-explanatory. The timers must match. Hello and Dead interval, that was obvious.

For the second, not so easy. If the participant interfaces have the same DR/BDR electiveness type (ie, participate or not in DR/BDR elections) they will be compatible. That's broadcast and non-broadcast for DR/BDR elective; point-to-point, point-to-multipoint and point-to-multipoint non broadcast for not DR/BDR elective.

The not-so-easy part is that if the timers match and the DR/BDR electiveness is mismatch. The frakking things will establish the frakking adjacency! This drove me nuts for two days. TWO DAYS. But they won't exchange a frakking LSA.

Frak me.

(and I found a copy of John T. Moy's OSPF book. I should read it, I guess)