[c-nsp] Recommended steps to avoid 100% CPU while executing "debugip nat"

[Oliver Boehmer (oboehmer)]

> 
> We have executed this morning "debug ip nat" in a 7206VXR with ver bad
> results.
> The router was overloaded for a while and at the end we needed to
reboot
> it.
> 

well, any "busy" debug can kill a box, but the following recommendations
can help to mitigate the risk (not get rid of it completely)
- disable console logging (this should be a best practice)
- disable sending debugs to syslog (logging trap informational)
- don't do "term moni" while turning on debugs, to reduce the
telnet/ssh/tcp load, instead log the debugs into the logging buffer
instead (need to increase it), and use "show log" to view the results.

> I was doing some research but I would like to hear from you too.
> As a plan we have deployed CoPP configuration for management traffic,
> and I am thinking on play with the "scheduler" command, the problem is
> that I don't know what could be the correct "scheduler"
configurations.

on the 7200, I've seen scheduler allocate 3000 1000 and 4000 1000 being
used successfully (default is 4000 200), to give more response to user
processes. There could be a minor performance dip as you devote more CPU
time to user processes, not sure if you'll be able to observe any in
real life..

	oli