[Orca-users] Re: orca not graphing weekly properly

Blair Zajac blair at orcaware.com
Mon Jul 29 14:19:44 PDT 2002


Yves Dorfsman wrote:
> 
> On Mon, 29 Jul 2002, Blair Zajac wrote:
> 
> > To fix this you may want to try this:
> >
> > 1) Stop Orca.
> > 2) Delete the orca.state and *.rrd files.
> > 3) Start Orca.
> 
> No that doesn't help.
> 
> Looking at it more closely, this problem might have to do with the
> "interval" feature. Am I right thinking that "interval" is the number of
> seconds in between two datapoint ??

Yes.

> 
> If I am right, why do we need this, since there is a timestamp ?

When Orca sets up the RRD files, the RRD files need to be told the
normal interval for measurements so it can be initialized properly.
RRD files are constant size and saves a bucket for each measurement
interval.

> Does that mean that ORCA won't be able to graph if the data comes in at
> random interval (say every five minutes, then switch to every 10 minutes,
> then change again...) ??

Yes.

> 
> What is the effect of getting the wrong interval number ?

If you're measurements occur in a time period longer than the Orca/RRD
`interval', then you'll end up with blank period in the RRD files and
you're plots will have holes in them.  If your time period between
measurements is shorter than `interval', then I believe RRD will
perform some averaging, min, or max on the values.  You'd have to
look at RRD's manual pages to specifically what happens here.

By default, the Solaris orcallator.se takes care to measure on
integral multiples of `interval' to avoid these problems.

> What is the effect of having data points at random time ?

Just as above depending upon the measurement rate with respect to
the `interval'.

Best,
Blair

-- 
Blair Zajac <blair at orcaware.com>
Web and OS performance plots - http://www.orcaware.com/orca/



More information about the Orca-users mailing list