[rrd-users] [rrd] Can't I turn off data smoothing for GAUGE?
philip at vogon.net
Fri Aug 20 05:08:22 CEST 2010
On 8/19/2010 3:35 PM, Larry Adams wrote:
> Sent from my Daughters iPad
> On Aug 19, 2010, at 1:27 PM, Philip Peake <philip at vogon.net> wrote:
>> On 8/19/2010 8:15 AM, Tobias Oetiker wrote:
>>> Today Philip Peake wrote:
>>>> Yes ... I am actually measuring two parameters.
>>>> One of them has fairly high numbers (in the 100's), the interpolation
>>>> (or whatever you want to call it) is ok there. The actual number
>>>> displayed is not too important as long as its something like right.
>>>> The second set of samples are more like:
>>>> 0 0 0 0 0 0 1 0 0 0 1 1 0 0 2 0 5 0 0 0 0 3 0 0 1 0 0 0 0 0 1
>>>> They are integer values. Displaying them as (for example) 0.666732134 is
>>> please elaborate ... time based integer data ? never seen it ...
>> Ok ... lets try.
>> This is a queue.
>> In the trivial case, its empty.
>> In other cases there are an integer number of items in the queue
>> (partial items don't exist).
>> Obviously, the number of items in the queue varies with time.
>> Because of the short timescales involved in processing these items, any
>> reading of the number of items in the queue is only a snapshot. Fraction
>> of a second earlier or later and the number could be significantly
>> different. But it will always integer.
>> The data is not time based, its based upon request rates and processing
>> rates. It will change with time but is essentially not itself directly
>> related to time.
>> Does that make sense?
> If you dont want consolidation, then dont have more than one rra. The rrd is going to be big, but you wont have to worry about averages nearly as much. You can handle display (gprint) anomalies using a preset.
In this particular example, there is just one RRA ...
What I found most perplexing was that Tobi seemed to have had the (IMHO
:-) ) right idea to begin with.
To leave gauge readings alone, no further processing just enter them. He
went so far as to document this behavior.
Then somehow, that "divide by interval" bit crept into the processing
when storing the value.
At what point did that get added, and why??
I take Alex's point about a speedometer already being a rate, so don't
play with it, but that's not true of all gauges, for example, one
measuring water pressure, or voltage or <whatever>.
I suspect that the "right" answer (if there is one) is maybe not to
create a set of different types, but probably to add a "do not process
value" flag to *any* data type? The basic concept of the data types
remains the same.
I think the basic concept we are talking about here is that not
everything is a rate, some are just values.
More information about the rrd-users