[c-nsp] Cisco 7500, output errors on FastEth interfaces

[Pete Templin]

Ray Van Dolson wrote:

> Investigating this it seemed we weren't using the proper switching method.
> Turned on "ip cef distributed" which apparently offloads a lot of the
> switching decisions to the VIP cards.  Immediately the load on the RSP4
> dropped to 2%.  Hoorah right?

Good choice.

> Things seem to be working OK, but now I am beginning to see output errors on
> the FastEthernet interfaces.  I'm wondering if this is due to the cef entry I
> enabled above.

Aside from the half-duplex problem others mentioned, this may be the 
"problem".

> Should I turn on ip route-cache distributed for the Hssi port only and use
> another method for the FastEthernet ports?  If I understand things correctly,
> turning ip cef distributed on globally will enable this on all interfaces.

With VIP4-50s, I doubt you want to do that.  You should have a decent 
amount of PPS capacity on them, so I wouldn't want to punt the switching 
decision to the RSP if you have the distributed capacity.

Another possibility, though not likely, is that you are in fact maxing 
out the very-short-term capacity of the VIPs.  Normally, a burst of 
packets will appear as input-side ignores (because the VIP didn't have 
enough CPU, and ignored the packet), but if you have various QOS 
features enabled then the drop gets pushed to the output side for some 
reason.

pt