[rrd-users] Re: how to deal with COUNTER-reset at unusual values?
greg at mwater.ca
Mon Jan 10 16:31:09 MET 2005
From the rrdtool man page:**
DERIVE will store the derivative of the line going from the last to
the current value of the data source. This can be useful for gauges,
for example, to measure the rate of people entering or leaving a
room. Internally, derive works exactly like COUNTER but without
overflow checks. So if your counter does not reset at 32 or 64 bit
you might want to use DERIVE and combine it with a MIN value of 0.
peter pilsl wrote:
>I use the COUNTER-DST for monitoring the number of spam in my mails. The
>data comes from analyzing a log-file that is daily rotated by the system.
>So the values delivered are like
>100 200 350 800 1399 1850 50 129 ....
> logrotate here
>So what happens between the 1850 and 50? According to the manpage the
>COUNTER-type can deal with a counter-overflow, but it expects it at
>16bit and not at an even unknown value like "a bit more then 1850". If
>rrdtools asume a 16bit-boarder then it would assume a value of
>2**16-1850+50=63736 for the step between 1850 and 50, which would not
>look good in the graph.
>How to deal with this? A possible approach would be to use a statistical
>correct estimation for the overflow-point (that number of spam when
>logrotation appeared). IMHO there are two possibilities: to use 1850 as
>boarder (and therefore 50 as step between) or to use 1850 + the average
>previous step(s) as overflow-point.
>But of course this is only theory, cause I've got no idea how to
>implement this ..
>thnx for any idea,
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
More information about the rrd-users