Re: bgp pmtud

Hi Imran,

1) DF bit is not set for the BGP update message.

From my experience, You can see BGP flapping in case you have device in
between like transmission devices or a switch working as a pure layer 2
device which are dropping the packets if they dont support packet of 1500
bytes). The MTU supported by the routers will be 1500 bytes so MSS would be
1460(Excluding 20 bytes IP and 20 bytes TCP)

I had faced this issue for one of my customer. BGP neighbor was flapping
every 3 minutes(Keepalives getting lost). After troubleshooting we found
out that the keepalive messages were getting queued behind the BGP update
messages(Considering BGP update message payload to be 1460 bytes.). We had
to ping and check the MTU supported by the underlying devices. Then we
manually adjusted the value of tcp mss using the global command on cisco
router(The interface level didnt help as it wont affect locally originated
traffic).

2) I am not sure about how path MTU discovery will work in this scenario
as it will only come into picture when you have devices that understand IP
in between the BGP neighbors. I am not sure what will happen if the
transmission devices or layer 2 devices in between are dropping the packets
of 1500 bytes silently(It means the IP router at the other is not receiving
icmp packets hence there will not be any ICMP unreachable message generated
from the IP router).

Regards.

On Thu, Nov 1, 2012 at 8:24 PM, Tom Kacprzynski <tom.kac_at_gmail.com> wrote:

> I think this a very good question to be tested in GNS3 with wireshark.
> Doing things you'll learn so much better than just reading about it. Please
> let us know what you find out.
>
> Regards,
>
> Tom
>
> On Thu, Nov 1, 2012 at 8:51 AM, Imran Ali <immrccie_at_gmail.com> wrote:
>
> > Hi all,
> >
> > By default when bgp router sends update ...does it sets DF bit ?
> > question 1
> >
> > in this case if the size of update packet let say is larger than
> > MTU of any of transit links ..then any transit router upon reception
> of
> > DF marked bgp update packet send an ICMP error , if icmp is filtered
> due
> > to any firewall then originating router will not reduce the MSS and
> this
> > results in oscilaitng bgp sessions
> >
> >
> >
> http://nagendrakumar-nagendra.blogspot.com/2010/03/bgp-path-mtu-discovery.html
> >
> > does any one have experinced the same or can comment on this ?
> >
> >
> > 2) if bgp path mtu is enabled by default ..then do the router undergo
> > path mtu discovery ..and figures out lowest mtu of transit links and then
> > adjusts the [mss = lowes mtu - (ip headers) - (tcp headers) ] ..?
> >
> >
> > Blogs and organic groups at http://www.ccie.net
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Thu Nov 01 2012 - 22:51:49 ART