PLEASE NOTE THE CROSS-POSTING when responding. Please direct responses
to the authors and the PIM WG only, so the broader community doesn't get
bombarded with responses. This message is also being sent to the
Routing-Discussion and ISIS WGs.
draft-lts-pim-hello-mtu-01.txt was inspired by customer reports of
multicast routing hangups caused by MTU mismatches at the two ends of a
link. The end with the larger MTU sends a message within its MTU size
that is discarded at the other end because it is larger than that end's
MTU size. The proposal in draft-lts-pim-hello-mtu-01.txt was for
adjacent routers to exchange information on their MTUs on each link,
then restrict message sizes to the smaller of the MTUs.
Working Group discussion in Vancouver raised the obvious question: do
MTU mismatches cause problems for other routing protocols? And do any of
these protocols have a solution to the problem?
I raised the issue of the effect of MTU mismatches on the NANOG list.
There were quite a few responses. Amongst other comments:
- OSPF hangs at a particular point in the state machine and fails to
come up. Debugging can take a while if you don't think of MTU mismatch
as a possible cause.
- Cisco has published a note on the subject, to be found at
http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a0080093f0d.shtml
(sorry about the wrapping).
- BGP can hang after a link flap when two specific products [identities
omitted] are at the two ends of the link and the configured MTUs are
identical, because one end includes the ethernet overhead and the other
doesn't.
- a good MTU debugging tool can be found at
http://www.elifulkerson.com/projects/mturoute.php
This message is written to ask the broader routing community what, if
anything, routing protocols should do about identifying MTU mismatches,
and beyond that, accommodating them. As requested above, please direct
responses to the authors and the PIM WG only.
Tom Taylor
_______________________________________________
OSPF mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ospf
