If your PE-PE is not a trunk port, which is normaly the case, and you want to successfully transport a payload of 1500 bytes, you should consider setting IP MTU as 1500 + 20 = 1520 bytes. mostly two labeled are stacked for vpn traffic but there are cases when 3 label may also be used so you should consider 12 bytes for mpls header ( 4bytes each), it will make the mpls mtu as 1532. Your physical interface mtu should be equal or larger than 1532 + 18(Ethernet header) bytes. Try this out and share the results.
/AHQ -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Chris Hale Sent: Monday, June 01, 2009 8:42 AM To: cisco-nsp Subject: Re: [c-nsp] strange behavior over MPLS network - remote desktopwon't work On Sun, May 31, 2009 at 9:04 PM, Ray Burkholder <[email protected]> wrote: > > > > > > What does that indicate to you? 1472 + VLAN tag plus MPLS < 1500? > > > > When provisioning MPLS circuits, one has to be careful. Basic MPLS will > attach one or more 4 byte labels on to each packet. Psuedowires attach > additional bytes onto each packet. WAN circuits running MPLS need to be > provisioned such that the interface MTU is set to 1500 PLUS any pseudowire > overhead plus any MPLS label overhead. If you try to run MPLS stuff across > a standard 1500 MTU WAN interface, you get the problems you are now > encountering: fragmentation, drops, corruption, ... Some protocols can > handle it, but I've read that RDP sets the no-fragment bit, thus dropping > the packets. > > STM-1 and DS3 circuits run by default at 4470 bytes so easily accommodate > MPLS overhead. Ethernet circuits are at 1500, and you have to work with > upstream vendors to ensure their networks can handle MTU's greater than > 1500. Cisco switches need a reboot after setting a system mtu setting. > Routers can change interface mtu settings on the fly. > > You could try setting your MTU setting on your pc to 1300 and see if things > work. If they do, then you know you have an upstream mtu problem. > I have an available DS3 interface on each of the POP H routers. Maybe I will set that up tomorrow and push the MPLS traffic across this interconnect to see if that helps. Maybe the mpls mtu setting on the PA-FE-TX interfaces just isn't working. I have also forced the GigE MPLS MTU settings on the backbone link between the POPs to 1538 as they were at the default of 1500 before. Thanks again, Chris _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
