[c-nsp] EoMPLS VC up on one side, not on the other.

[Stephen Fulton]

Gert,

When I decided to get into the networking field, I knew that MTU was the bane of 
the industry, I just did not realize I'd still be dealing with it years later :) 
    EoMPLS hell, indeed.

If Ytti's reading, I'd like to know what those details are too.  Right now I 
have 7600/RSP720's in the core, not the edge, so it is of limited use to me 
right now, but it's worth knowing, especially when I roll out VPLS.

I *do* wish that I could change the MTU of SVI's on the 3750 Metro, like I can 
do on the 7600/RSP7200 SR train.  *hint hint*..

Anyway, thanks all :)

-- Stephen

Gert Doering wrote:
> Hi,
> 
> On Sat, Nov 15, 2008 at 11:58:07AM -0500, Stephen Fulton wrote:
>>     MTU: local 9000, remote 1500
> 
> Welcome to EoMPLS hell.
> 
> If MTUs do not match, the VC won't come up.  And you can't change the
> "raw" MTU on a dot1q subinterface.
> 
> I seem to remember that there is a knob in recent SR* IOS trains that
> permits you to ignore MTU mismatches, but I have forgot the details - Ytti
> will know all the details (how to configure it plus availability).
> 
> Workaround: make sure the "base interface" MTU is the same on both sides
> ("gig 0/1", with no ".123").
> 
> gert