[Systems] Fwd: Munin alert for bender

[Bernie Innocenti]

El Wed, 05-08-2009 a las 06:38 -0700, Luke Faraone escribió:
> On Aug 5, 2009, at 6:14, Bernie Innocenti <bernie at codewiz.org> wrote:
> 
> > I think the figure is in %...  Still, 60% of _system_ load is
> > suspicious.  Could it be that all the VM CPU usage gets accounted on  
> > the
> > host system?
> 
> CPU load is, iirc, how many times faster the CPU would have to be to  
> handle all the requests made over PERIOD in "realtime".
>
So 1 CPU load would mean that the processor was dully utalized,  
> whereas 60 would mean you would need 60 times more power to handle the  
> requests without slowdown.

Yes, this is what human readable load usually looks like in UNIX.
However, munin determines it like so:

awk '/^cpu / { print "user.value " $2 "\nnice.value " $3 "\nsystem.value " $4 "\nidle.value " $5 "\niowait.value " $6 "\nirq.value " $7 "\nsoftirq.value " $8 }' < /proc/stat

So I'm actually surprised that we're seeing numbers as small as 60 in
the warning email.  On my laptop, I see:

user.value 278987
nice.value 15231
system.value 404446
idle.value 10797874
iowait.value 233824
irq.value 39
softirq.value 11845



-- 
   // Bernie Innocenti - http://codewiz.org/
 \X/  Sugar Labs       - http://sugarlabs.org/