[rrd-users] A reason there isn't a CF SUM function?

[Alex van den Bogaerdt]

----- Original Message ----- 
From: "Rafal Gwizdala" <rafal.gwizdala at gmail.com>
To: <rrd-users at lists.oetiker.ch>
Sent: Tuesday, August 21, 2012 9:17 PM
Subject: Re: [rrd-users] A reason there isn't a CF SUM function?


> On Tue, Aug 21, 2012 at 8:00 PM, Simon Hobson 
> <linux at thehobsons.co.uk>wrote:
> (...)
>
>> Which you can get with :
>>
>> cdef:data_sums=data_avg,${res_step},*
>>
>> Where ${res_step} is the length of the consolidation interval being
>> used. In your example, 1800 seconds.
>>
>> What might be more useful might be an option to handle unknowns
>> differently during consolidation.
>>
>> the only problem is that when rrdtool selects the consolidation interval
> automatically you have no easy way to obtain that value in your cdef
> calculation. Even if you specify the interval in the command line, rrdtool
> can silently use a larger value if RRA data resolution is lower than
> specified and you'll get invalid sums on the graph.

Someone, which will not be me, could expand the set of VDEF computations and 
make this value available. There are more values which would be welcome for 
some, e.g. width/height of the graph area, number of intervals being graphed 
(which does not necessarily equal the graph width), perhaps number of 
unknown intervals.

> And I'm not yet sure what to do with unknowns - probably you could prevent
> unknowns by using large heartbeat parameter during rrd creation and large
> xfiles factor for RRAs

That is the inherent problem with unknowns. You just don't know what you 
should have gotten so it's hard (if not impossible) to substitute an 
alternative. The X-Files factor has its name for a good reason...

cheers,
Alex