BGP Keepalives are very short, but Updates can be very long. It looks like 
they can be 4096 bytes from RFC 1771 (not counting headers).

BGP relies on TCP and IP, as you know, of course. Those layers would have 
to make sure that the IP Don't Fragment bit was set to 0 (which means May 
Fragment). I checked a few BGP packets from a Cisco router and they do seem 
to have that bit set to 0.

I still think it's worth discussion, though. There may be some 
implementations that don't set the bit to 0. MTU problems also crop up in 
weird places due to tagging, although you might not expect to see that with 
BGP. The poster seems to have run into actual problems, though, maybe.

Sorry if my message was a bit punchy. I just thought you sounded so 
imperious that I had to sound that way too. I'm glad that you are 
impervious to TLAs though. ;-)

Priscilla

At 10:57 PM 4/18/02, nrf wrote:
>""Priscilla Oppenheimer""  wrote in message
>[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> > He didn't say that BGP negotiates the MTU in any of its PDUs. He just
says
> > that mismatched MTUs can be a problem, which is all I mentioned in my
> > message about OSPF also (although OSPF does in fact also include the MTU
>in
> > database description packets and refuse to become adjacent with a router
> > that doesn't agree on the MTU).
>
>That's what I'm talking about - OSPFv2 does in fact include an MTU field in
>its PDU's (Ok, Ok, it's not really a negotiation per se, but still...), so
>if somebody starts a discussion about MTU and BGP, then it would stand to
>reason that BGP includes an MTU field somewhere, which I am not aware of.
>
>And besides, the idea of MTU problems in BGP is an interesting one, because
>of the fact that BGP peering often occurs between non-adjacent routers.
>What is the relevant MTU size of such a peering arrangement?   The routers
>do not share a common network, so is it really relevant to talk about MTU?
>
>
>
> > Did that have enough TLAs for you? ;-)
>
>I've read enough RFC's in my day to be impervious to TLA's.
>
> >
> > Priscilla
> >
> > At 09:53 PM 4/18/02, nrf wrote:
> > >Really?  I had never heard of this problem.  I'm not aware that BGP
> > >negotiates MTU in any of its PDU's.  Can you provide the RFC that
>discusses
> > >this problem?
> > >
> > >
> > >""suaveguru""  wrote in message
> > >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> > > > If I am not wrong this problem also occurs for BGP
> > > > peers with unmatched MTU sizes which causes BGP to
> > > > flap when they exchange routing tables , especially if
> > > > one neighbour is configured with full-routes
> > > >
> > > >
> > > > regards,
> > > >
> > > > suaveguru
> > > > --- Priscilla Oppenheimer  wrote:
> > > > > The problem happens when the routers try to exchange
> > > > > database description
> > > > > packets. One side can send packets that are too
> > > > > large for the other side to
> > > > > receive. Then the routers never achieve adjacency.
> > > > > It's an infamous
> > > > > problem. I was glad that Kevin brought it up. I was
> > > > > thinking we should have
> > > > > mentioned it in that other thread about OSPF Hellos
> > > > > (although this problem
> > > > > happens after the initial hellos).
> > > > >
> > > > > More here:
> > > > >
> > > > > http://www.cisco.com/warp/public/104/12.html
> > > > >
> > > > > Priscilla
> > > > >
> > > > > At 11:33 AM 4/17/02, Kane, Christopher A. wrote:
> > > > > > > The most frequently mismatched parameters
> > > > > relevant for OSPF
> > > > > > > configuration
> > > > > > > seem to be dead intervals & mtu sizes.
> > > > > >
> > > > > >OSPF doesn't care about MTU size.
> > > > > ________________________
> > > > >
> > > > > Priscilla Oppenheimer
> > > > > http://www.priscilla.com
> > > > [EMAIL PROTECTED]
> > > >
> > > >
> > > > __________________________________________________
> > > > Do You Yahoo!?
> > > > Yahoo! Tax Center - online filing with TurboTax
> > > > http://taxes.yahoo.com/
> > ________________________
> >
> > Priscilla Oppenheimer
> > http://www.priscilla.com
________________________

Priscilla Oppenheimer
http://www.priscilla.com




Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=41984&t=41613
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]

Reply via email to