[rrd-users] Direct RRA updates

Simon Hobson linux at thehobsons.co.uk
Sun Nov 13 22:29:18 CET 2011

krcmar at datinel.cz wrote:

>  > or if you want to go even further, set the MRHB
>>  to 60s and then supply three fake updates, 1s min, 1s max and a
>>  large 58s with whatever is necessary to create the right average
>This was my very first idea but I haven't dared to even send it to
>this list. Sould this "ugly tripple update" be approved by You, Tobi,
>I am going to use it :-)

We don't have to approve it, it's a free (and in speech as well as 
beer) tool and you can use it however you wish.

As it happens I have something not too dissimilar to this at work. 
Our UPS reports the supply voltage, plus the min and max values over 
the previous minute. So I collect these three values (plus others) 
and I have three DSs for voltage - voltage, min_voltage, and 

>  >> A side note: why, under usual conditions (not the example above),
>>>  should I define RRAs with steps=1 for all the aggregates I want to
>>>  then use with steps>1? (examine any RRD file with "rrdtool info" to
>>>  see several RRAs with steps=1). RRA with steps=1 does not consolidate
>>>  data, it only stores them: it is enough to have only one such a RRA...
>>  yes it is enough ...
>So is it a common mistake of scripts built upon rrdtool to define
>several RRAs with steps=1?

It's not a mistake if that is what you require.

>Will the "rrdtool graph" take (for example)
>data from RRA:AVERAGE::1 when requested to draw recent MAX of a DS
>(and there is no RRA:MAX::1 present)?

Sorry, I don't understand that.

krcmar at datinel.cz wrote:
>  > That is what you would need to do.
>>  Set your step to 1 second, update as often as your program allows,
>>  and then have a consolidation for 60 second periods. You would then
>>  get 1 minute reporting periods, each with a min/max/average for the
>>  60 x 1s step times it contains.
>This is something I would like to avoid. I am not interested in all
>the 60 values, only MIN, MAX and AVERAGE of them.

You don't have to actually use them. In fact I'm not sure you even 
have to store them !
If (say) you only stored 60 samples with a step of 1s, then unless 
you try to retrieve or graph less than one minutes worth of data, the 
rrd tools won't use that and will instead use the coarser 
consolidated data.

Simon Hobson

Visit http://www.magpiesnestpublishing.co.uk/ for books by acclaimed
author Gladys Hobson. Novels - poetry - short stories - ideal as
Christmas stocking fillers. Some available as e-books.

More information about the rrd-users mailing list