From: Bert Driehuis <driehuis@playbeing.org>
Date: Fri, 14 Jan 2000 03:41:07 +0100 (CET)

This script will read an XML file produced by

       rrdtool dump foo.rrd >in.xml

and look at the $maxspike highest samples per datasource. It then finds the
records with the most hits and ditches the data. The resulting file can be
read back into the RRD database with the command

       rrdtool restore out.xml foo.rrd

The whole idea is to find and eradicate "spikes" caused by erroneous
readings affecting entire records.

This tool is not for the faint of heart, will require tweaking per case
(even though that should just be picking values for cutoff and to a lesser
extent, maxspike). It will cause data loss, for obvious reasons.