[c-nsp] MTU on XR

[Vitkovský Adam]

Yeah it's very confusing.  
Personally I don't trust the auto adjustment as it doesn't always work out as expected. 
So I hardcode the MTU on sub-interfaces(L2 and L3) based on the number of tags. 

Also my experience is that on the (l2transport) interfaces the Pseudowire MTU is adjusted automatically based on the tag rewrite operation performed so on the remote(IOS) end of the PW you need to adjust the interface MTU based on the XR interface MTU - L2 overhead and + or - the number of push-ed or pop-ed tags respectfully. 


adam
> -----Original Message-----
> From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
> Mikael Abrahamsson
> Sent: Monday, August 25, 2014 1:36 PM
> To: cisco-nsp at puck.nether.net
> Subject: [c-nsp] MTU on XR
> 
> 
> Hello.
> 
> http://www.cisco.com/c/en/us/support/docs/ios-nx-os-software/ios-xr-
> software/116350-trouble-ios-xr-mtu-00.html
> 
> What I can check personally with machines I have available, the above article
> is correct in saying that vlan tags does not count when it comes to configuring
> MTU. XR will automatically add number of vlan tags when programming MTU
> to the hardware, no care has to be taken to configuring it, at least not for L3
> interfaces.
> 
> On an interface with MTU 9200 on the main interface, the subinterface has
> this:
> 
> #show im database interface Te0/3/0/0.101 Interface TenGigE0/3/0/0.101,
> ifh 0x06001340 (up, 9204)
>    Interface flags:          0x0000000000800597 (ROOT_IS_HW|IFINDEX
> 
> |SUP_NAMED_SUB|BROADCAST|CONFIG|VIS|DATA|CONTROL)
>    Encapsulation:            dot1q
>    Interface type:           IFT_VLAN_SUBIF
>    Control parent:           TenGigE0/3/0/0
>    Data parent:              TenGigE0/3/0/0
>    Views:                    GDP|LDP|L3P|OWN
> 
>    Protocol        Caps (state, mtu)
>    --------        -----------------
>    None            vlan_jump (up, 9204)
>    None            dot1q (up, 9204)
>    ipv6            ipv6_preswitch (up, 9186)
>    ipv6            ipv6 (up, 9186)
> 
> so this single tagged subinterface has automatically had its MTU raised to
> 9204 to accomodate the vlan tag, even though config says 9200. The IPv6
> MTU ends up being 9186.
> 
> This for instance contradicts what AMSIX has written on this page:
> 
> https://ams-ix.net/technical/specifications-descriptions/config-guide
> 
> "5.7. MTU Config
> 
> On Cisco IOS, the interface IP MTU should be set to 1500. MTU configurations
> for IOS-XR include the Layer-2 headers, and need to be adjusted when using
> VLAN tags."
> 
> So which one is correct? I have talked to numerous people and they have
> different experience and results are contradicting.
> 
> --
> Mikael Abrahamsson    email: swmike at swm.pp.se
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/