[c-nsp] PPP Mulitilink on 3620

[Rodney Dunn]

Just for the record.

You should never have to enable CEF on a per
interface basis when you turn it on globally.

If you do it's a bug.

Leave the switching flags on a per interface
basis as default (which will be CEF if you
enable it globally as is recommended).

In some old code when you turn on CEF globally
you would see "ip route-cache cef" go under
the interface too.  That was a dumb move on our
part.  Because when it was removed and made the
default switching vector on the interface people
started thinking they had to manually enable it
on a per interface basis wich is wrong.

Turn it on globally and that's all you have to do.

As for the member links of a MLPPP bundle.  When
you put them in a bundle you deactivate any L3
capability of those member links.  Don't apply
any configuratin to the member links (the only
one I have used on the member links is to turn
the keepalive timers down to get faster link
down detection).

Let us know how the CPU does now since you have it
on and it's working.

Rodney


On Thu, Feb 24, 2005 at 10:22:04AM +0000, David Simmons - EliteUKServe.Net wrote:
> 
> Thanks for that. When you say interface level ... just the multilink int or 
> each searial. What about the ethernet port?
> 
> Cheers,
> 
> Dave
> 
> 
> On Thursday 24 February 2005 10:18, Gert Doering wrote:
> > Hi,
> >
> > On Thu, Feb 24, 2005 at 09:34:28AM +0000, David Simmons - EliteUKServe.Net 
> wrote:
> > > Is this right?
> > >
> > > #sh adj detail
> > > %CEF not running
> >
> > No :-) - you want to switch on CEF.  (Global: "ip cef", and on the
> > interface level "ip route-cache cef").
> >
> > gert