[rrd-users] Last value always nan

[Alex van den Bogaerdt]

On Tue, Sep 04, 2007 at 05:37:27PM +0200, Michael Jürgens wrote:

> > Solution:
> >
> > Integer divide 1188916748 by your step size (60?), then multiply
> > by the same amount.  The result is 1188916740. Use that as end time.
> >
> 
> 
> 
>     > /usr/local/rrdtool/bin/rrdtool fetch
>     > /home/mrtg/data/dolly.log4j.warn.rrd AVERAGE   -s 1188919920  -e
>     > 1188919920
>     >                             ds0                 ds1
>     >
>     > 1188919980: nan nan
> 
>     but
> 
>     > /usr/local/rrdtool/bin/rrdtool fetch
>     > /home/mrtg/data/dolly.log4j.warn.rrd AVERAGE   -s 1188919919  -e
>     > 1188919920
>     >                             ds0                 ds1
>     >
>     > 1188919920: 0.0000000000e+00 0.0000000000e+00
>     > 1188919980: nan nan
> 
>     and
> 
>     > /usr/local/rrdtool/bin/rrdtool fetch
>     > /home/mrtg/data/dolly.log4j.warn.rrd AVERAGE   -s 1188919919  -e
>     > 1188919919
>     >                             ds0                 ds1
>     >
>     > 1188919920: 0.0000000000e+00 0.0000000000e+00
> 
>     So it should be "int(ts/60)*60-1"  - right?

Erm... yes and no.

This has been discussed in the past (several times) and the conclusion
was:

* A timestamp itself has no duration.
* Intervals (in rrdtool) are defined by a duration and an end time.
* If you want information *ending* at 1188919920, you are NOT asking for
  the interval *starting* at 1188919920.

In other words: time 1188919920 "belongs to" interval 1188919919..1188919920.

But apperently one of these is true:

-1- you are using a version which still contains the bug, and you should
    upgrade

or

-2- the bug was introduced again, and should be fixed again

or

-3- Tobi changed his mind, and I'm not aware of it


Several off-by-one errors will result from such mistakes. Try
showing a huge spike at the very end of your time interval, and
compare with output of GPRINT.

In the mean time: yes, int(ts/60)*60-1 will work. But it should not
be necessary.

HTH
-- 
Alex van den Bogaerdt
http://www.vandenbogaerdt.nl/rrdtool/