[c-nsp] 6500 stable code

[Phil Mayers]

A.L.M.Buxey at lboro.ac.uk wrote:
> Hi,
> 
>> FWIW we had problems finding a stable release, but have taken the plunge  
>> and rolled out 12.2(33)SXI. Remarkably, it seems to be a solid release.
> 
> hmmm, interesting. we have several test systems running SXI (and
> they've been good to us so far)...will probably go for the full rollout
> now.  dont forget, 12.2(50)SYA soon...  ;-)

For what platform?

SXI is nice. There are a few very minor bugs we've run into:

  * DHCPv6 relay doesn't work inside a 6vPE VRF, and I've had the devils 
time explaining to TAC what the problem is - they claim it's a feature 
enhancement to use the VRF rather than global routing table :o(

  * Some scary %MLSM-6-LC_SCP_FAILURE messages have started appearing in 
the logs, indicating the Sup is unable to punt some multicast groups 
into hardware CEF, and will software-forward them. I hope that doesn't 
happen with a 20mbit/sec video stream...

  * Some cosmetic / non-service-affecting %DATACORRUPTION messages when 
fiddling with the TCP/xml syslog stuff

We also ran into an early issue where the BGP path MTU discovery wasn't 
working between our 6500s and our J-series junipers (on baby giant 100M 
ethernet circuits) but that is probably more juniper/cisco MTU madness - 
not a bug, and easily worked around by disabling PMTU on the 
peer-session template.

One of our routers shows considerably higher SP CPU usage in the "slcp 
process" but I'm not very worried.

I'm sure I'll find more issues given how new it is, but the big stuff 
works absolutely fine.

sh run | inc section ^track
sh run part router bgp 65000

...and the like are very welcome additions, as is LLDP (no MIB though - 
boo) and the config archive / logging / diff utilities.

I'm a bit disappointed some of the features e.g. HSRP multiple group 
optimisation, BFD etc. are useless on SVI/switchport based deployments - 
does anyone really run 6500s with sub-ints? - but there we go.