[Orca-users] plot accuracy

[Jon A. Tankersley]

> ------------------------------
>
> Message: 7
> Date: Tue, 17 Jun 2003 12:56:18 -0700
> From: Blair Zajac <blair at orcaware.com>
> Subject: Re: [Orca-users] Data Accuracy problem or is it averaging?
> To: Santana Fernando <santana.f at mellon.com>
> Cc: orca-users at orcaware.com
> Message-ID: <3EEF7262.12AC8E6D at orcaware.com>
> Content-Type: text/plain; charset=us-ascii
>
> > Santana Fernando wrote:
> >
> > Hi, I edited RRDFile.pm and ImageFile.pm so that Orca
> > would not average the data for weekly, monthly and so forth.
> >
> > Also, I edited the Constants.pm file and made
> > @RRA_PDP_COUNTS =   (    1,     1,     1,   1); so that Orca
> > would not average at all for any of the plots.  After that, I
> > deleted all rrd files and started Orca.
> >
> > The problem now is that I see the minimum idle percentage to be a
> > number that doesn't match the graph.  For example, here in the weekly,
> > plot I would have had expected to see an idle minimum of 29.2 or a smaller
> > number.
> >
> > Does someone knows what's going on?  I would like to maintain data
> > accuracy on the plots before I present them to
> >
> > my supervisor.
>
> Santana,
>
> If you want to remove consolidated data, I would completely remove the
> consolidations.  I would modify lib/Orca/Constants.pm to have something
> like this;
>
> use vars         qw(@RRA_PLOT_TYPES @RRA_PDP_COUNTS @RRA_ROW_COUNTS);
> push(@EXPORT_OK, qw(@RRA_PLOT_TYPES @RRA_PDP_COUNTS @RRA_ROW_COUNTS));
> BEGIN {
>   @RRA_PLOT_TYPES = qw(yearly);
>   @RRA_PDP_COUNTS =   (                1);
>   @RRA_ROW_COUNTS =   ( 3*365*24*2400*12);
> }
>
> Let me know how this works.
>
> It'll probably be really slow this way though.
>
> Best,
> Blair
>
> --
> Blair Zajac <blair at orcaware.com>
> Plots of your system's performance - http://www.orcaware.com/orca/
>

I thought we had to update the .pm files to use MAX instead of the AVERAGE settings for
the rrdtool creation.
I've got the hooks in to allow the config file to allow that to be selectable, but I
can't get one of the perl modules to 'see' this setting, so I have a working copy that
will have both MAX and AVERAGE (and MIN if that were desired)  This will double the RRD
size, basically.

I've also been looking at the 'date' information.....  the Week XX is pretty useless to
my users....   I have to build a lookup table for them.  So, I've looked at doing
something like MM/DD instead.  This would be in rrdtool, that I'm making as a compile
option.  I haven't had a chance to test it, so I don't know what things will look like.
Also, putting MMM YY for the year, so more of the date information is in the graph
instead of the 'timestamp' at the bottom.