IPv6 hint: interface MTU to 1280 - he.net Webcast 39

[rchandraonline]

Oh, yes...same thing happens if they're not careful to people running PPPoE (typically DSL). MSS clamping unfortunately only helps for systems that pay attention to it. Unfortunately all systems behind the access point (e.g. tunnelling router) have to adj their MTUs. Otherwise, fragmentation may result.

[pbrutsche]

The same problem can happen in any sort of tunnel environment. For the purposes of discussion, IP vs IPv6 is irrelevant - I've seen the problem this video describes when passing internet traffic over a GRE tunnel between two Cisco 2610 routers.

Adjusting the interface MTU is one way to do solve the problem; another is what some systems (mostly Linux and *BSD) call "TCP MSS clamping". Cisco has a feature for IP called "TCP MSS Adjustment" that is the same thing. They need to add it for IPv6.

[bassbacke]

Just checked on my Cisco 1812 connecting to my ISP.

router-isp#show interfaces dialer 0

My Dialer0 interface MTU size is set to 1492 bytes manually.

My Tunnel0 shows a MTU 17920 bytes, but Tunnel transport MTU 1472 bytes

I guess I'm fine since 1492-1472=20 octets tunnel header and I didn't even have to configure the tunnel MTU manually.