Author Topic: PMTU(D)  (Read 304 times)


  • Are we there yet?
« on: October 28, 2014, 10:15:24 am »
You're finding that communications between A & Z are having issues due to black hole(s), do the usual probes and client MTU fiddling, and find that yes,  intermediate device(s) MTU is the issue for end-to-end comms.

You end up trying to isolate which device(s) in the chain of A, B, C, D...Z, & do some traceroutes to establish what the 'consistent' IP chain of A, B, C, D...Z is, and then start targetting each IP successively for expected response, eventually expecting to hit on the bad boy(s) in the chain, according to their lack of approriate response.

So, you find the bad boy(s), and for the LOLZ keep probing the rest of the chain to Z...whereon successive devices at BB(n)+1 respond in expected manner.

a) Howzat then? Either BB(n) should FAIL when itself AND devices beyond it are addressed, as it's a black hole, or BB(n) is managing to pass on traffic. Scratches head...

b) Can't see that you can tell the difference between duff setups at BB(n), or (hostile) policy in return chain actively blocking ICMP responses - they both look equivalent at this level; surely the only 100% would be to camp on each device in turn from A-Z, and probe solely to next device in chain?

I'd offer you some moral support - but I have questionable morals.