Re: [hobbit] weird rrd output (using the NCV module)

["Gary Baluha" <gumby3203 (at) gmail.com>]

On 9/13/07, Henrik Stoerner <henrik (at) hswn.dk> wrote:
>
> On Thu, Sep 13, 2007 at 11:20:03AM -0400, Gary Baluha wrote:
> > I'm seeing a weird output for one of my custom RRD graphs, using the NCV
> > module.  The collector script on the client
> > is creating an ncv-style output based on the bind 9 "rndc stats"
> command.
> > These numbers are counter-style numbers,
> > always increasing (until named is restarted).  As such, the default
> DERIVE
> > format of the NCV module is correct.
> >
> > The problem I'm seeing is that the numbers showing up in the RRD
> graph--as
> > far as I can tell--have nothing at all to do
> > with the actual "rndc stats" values.  For instance, the "last" and
> "current"
> > values of the data was:
> >
> > success : 283  ->  294  (11)
> > referral : 0  ->  0  (0)
> > nxrrset : 149  ->  156  (7)
> >
> > The number in () above is the actual difference between the numbers, and
> the
> > values I'm expecting to show up in the RRD graph.
> > But the actual values that are showing up for the above example (for the
> > "(cur)" value) are, respectively:
> > 34.0m
> > 0.0
> > 20.8m
>
>
> Everything is behaving the way it should, but it doesn't match your
> expectation.
>
> The problem is that rrdtool treats counters (and derive's) as RATES,
> i.e. "this many X per second". So when you have 11 requests over a 5
> minute period, rrdtool records (11 / 300) = 0,037 requests/second.
> That's why you have that "34.0m" recorded - apparently there was
> slightly more than 300 seconds between the two readings.
>
> If you want the graph to show how many requests occurred over the 5
> minute period, then the simplest way is to define the graph to use
> a value which is 300 times the actual value recorded by rrdtool :-)



Oh dear...  I completely misinterpreted what the RRD doco was trying to tell
me.  As always, thanks for your help.