[j-nsp] MTU on RSVP tunnel

[Eric Van Tol]

> -----Original Message-----
> From: Harry Reynolds [mailto:harry at juniper.net]
> Sent: Monday, October 27, 2008 1:47 PM
> To: Eric Van Tol; juniper-nsp at puck.nether.net
> Subject: RE: [j-nsp] MTU on RSVP tunnel
>
> Are you pinging with DNF set?
>
> Could be ingress node is seeing Ip and trying to frag based on the
> smaller inet mtu. Does seem frag would be based on packet's egress
> next-hop, which is mpls in this case, but not sure we do that at
> ingress.
>

Hi Harry,
I am pinging with DF-bit set, but also without.  Either way, I get no message that that the packet was dropped because it was too large.  I know that 1500-byte IP packets can get through because any destinations not using the RSVP tunnel work fine.  It's only when the packet traverses the tunnel that they get dropped.  For instance, if I ping x.x.224.2 while sourcing from lo0.0 on R3, everything works because both lo0.0 and x.x.224.2 are seen through the IGP.  If I source from an interface whose return path takes the RSVP tunnel, it dies.  I can see the pings getting to x.x.224.2 by the 'monitor traffic interface' tcpdump, but they don't ever make their way back.  Same happens in reverse.

-evt