[c-nsp] me3600 copp

Jason Lixfeld jason at lixfeld.ca
Thu Feb 27 11:30:01 EST 2014


Amazeballs.  Clear as mud.  Might be an action item for your SE to clarify.

On Feb 27, 2014, at 10:40 AM, Nick Hilliard <nick at foobar.org> wrote:

> On 27/02/2014 14:13, Jason Lixfeld wrote:
>> Does CoPP actuaally work now?
> 
> I don't know yet.  Separate to this, the copp documentation doesn't make it
> clear how copp interacts with the cpu queues.
> 
>> We've been bitten by the RPQ issue many times.  ARP and routing
>> protocols should now be split across two different CPU queues as of
>> 15.5(3)S2.
> 
> yeah, i've seen this happen in production.  Not pretty.
> 
>> In the mean time, you can use this to allocate more resources to that
>> queue to prevent the routing protocols from going away during an ARP
>> storm:
>> 
>> test platform qos policer modify cpu 1 5000000 100000
> 
> which is not exactly reboot persistent, nor does it solve the problem.
> 
> The documentation is very confusing.  The 15.3 documentation says:
> 
>> •The ACLs used for classification are QoS ACLs. The supported QoS ACLs are IP standard, extended, and named.
>> 
>> •Only IP ACLs are supported in hardware. 
>> 
>> •MAC-based matching is done in software only. 
> 
> But the 15.4 documentation says:
> 
>> •STP is the only supported protocol for control-place policy-maps.
>> 
>> Note IP ACL is not supported.
>> 
>> •MAC-based matching is done in software only. 
> 
> wat?
> 
> Nick
> 




More information about the cisco-nsp mailing list