[c-nsp] 12.2SX vlan-mapping

[Anders Marius Jørgensen]

Den 05/04/2011 kl. 12.32 skrev Daniel Holme:

> On 5 April 2011 10:49, Reuben Farrelly <reuben-cisco-nsp at reub.net> wrote:
>>> Approximately 200kbit/s throughput.
>> 
>> Sounds somewhat like the translation is being punted to software/CPU rather
>> than done in hardware.  I was seeing similar figures when I was working the
>> CPU hard to do a similar setup, until I figured out the right way to
>> configure it on the switch...
>> I'm using 12.2(33)SXI5 (soon to be SXI6) on an ME6524.
> 
> Having looked a little further I see that the PBR pushing packets into
> the trunk doesn't appear to be working in hardware.
> 
> I've got PBR, in the global context, doing this on other SVIs and
> that's okay. this PBR however is inherent-VRF and I'm matching on a
> /32. So yes, it does look like it's being punted for some reason.

I have seen an issue where PBR traffic was being punted to the CPU.
The traffic was sourced from within a VRF and was suppose to stay in the VRF.
In order to trigger hardware forwarding, I had to change the route-map to:

set ip vrf <vrf> next-hop x.x.x.x

(instead of set ip next-hop x.x.x.x, as expected)

Regards,

Anders