[rrd-users] Filtering of RRD Data prior to graphing ?

[Robert]

Hello,

I found a problem with RRD files with Counter DS not having a Upper or 
Lower limit for the DS.

Historical Data is collected by RRD. We have quite some data here how 
and the data is huge (some hundred GB) and a valuable asset. We found 
that some data collectors were set up wrong, not properly following the 
rule "Always use MIN and MAX values". However - this happened - so we 
have data that has huge peaks in the graphs. This is due to counter 
wraps on the system beeing monitored. RRD interprets this as very high 
change rate / frequency if no Upper or Lower values are specified.

Those peaks make the long time graphs really useless, because the real 
interesting stuff is masked by the peaks.

There are different ways to work around the issue:

1) Change the DS definition and life with the peaks for historical data 
(rrdtune does not touch the archived data right ? )
2) Try to figure out the max value and use max value for graphing 
effectivly cutting the peaks (difficult, also disables autoscaling of 
the graph, which is nice)
3) Dump the rrd's, filter out the peaks and import them again (quite a 
lot of work with some hundred GB's and also modifying historical data - 
not good)
4) filter the values greater than X before graphing  (this would be 
nice, while keeping the auto scaling of the graphs!!)

Is there a way to to (4) with current implementation of RRD ?

Thanks,
Robert