[rrd-users] rrdupdate

rihad rihad at mail.ru
Tue Dec 25 13:49:33 CET 2007

Alex van den Bogaerdt wrote:
> On Tue, Dec 25, 2007 at 02:31:07PM +0400, rihad wrote:
>>> If you REALLY need to fiddle with your historical data, then you will 
>>> need to export the data, alter the values in the XML file, then 
>>> import that into a new database.
>> Wow, absolutely amazing! So these are the hidden costs of MRTG -> 
>> rrdtool migration? :-) Thanks for the solution, though, I shall try that.
> You get extra protection for your historic data and are complaining?
> Anyway, don't forget you need to edit all RRAs covering the timespan,
> not just one.  MRTG-style RRAs have:
> Each row 5 minutes, AVERAGE, and a MAX
> Each row 30 minutes, AVERAGE, and a MAX
> Each row 120 minutes, AVERAGE, and a MAX
> Each row 1440 minutes, AVERAGE, and a MAX
> You will need to modify each AVERAGE, and you will at least have to look
> if each MAX should be changed.
> An example:
> You are going to remove a rate "55" from the database.  You find it
> in the 5-minute RRA and remove it. You notice the surrounding rates are
> all set to "1".
> If you would only alter that "55" into "NaN", then you didn't do your
> job properly. The 30-minute AVERAGE RRA still contains a rate "10",
> which should be corrected to "1". The 30-minute MAX RRA still contains
> rate "55", which should be corrected into "1". And so on.
Thanks for a thorough explanation, but as you probably guessed I simply 
grepped for the timestamp and NaN'ned them all :) Good enough for now, 
as the graphs changed...

> Make sure you understand consolidation before you edit your RRD. You need
> to do these computations yourself, and store the results.

I've read the RRD manpages covering the theory etc. over the past few 
years but forgot most of it :) It's probably easier for Tobi to consider 
implementing this Much Wanted (c) feature in rrdupdate for some MRTG 
folks like me :) It's so much easier to open the classic MRTG logfile in 
vi and edit it to your liking. Most people just want to get their job 
done quickly without having to wade through the docs for a deeper 

More information about the rrd-users mailing list