[rrd-users] Absolute values..
Alex van den Bogaerdt
alex at vandenbogaerdt.nl
Thu Nov 13 02:25:35 CET 2008
If you set step to 1, and multiply every amount of steps per entry in your
RRA by 60, each RRA contains the same amount of time, the same amount of
time per entry, but no normalization needs to take place. Have you tried
that?
Perhaps now, perhaps some day or perhaps never, subsecond precision may be
in your way. That's why you want to provide explicit timestamps, e.g.
'1226538361' instead of 'N'. Have you tried that? You want to do so anyway
if you want to be able to store the update commands somewhere else and
recreate your database (certainly during debugging this is very helpful).
Now look at consolidation vs. your remark 'I want no response alterations'.
For instance: the RRA "24 hours = .Steps = 60 'Minutes per Hour". I take
it you mean every RRA entry should be 60 minutes. This means one rate
should be built from 60 separate 1-minute rates or, after the modification I
suggested, be built from 3600 separate 1-second rates. If you set
consolidation function MAX, you will get the highest of those 60/3600 rates
and, thus, loose the other 59/3599 rates. This is what you ask RRDtool to
do, it may also be what you want it to do but it sure is an alteration of
your input. Please be specific when you describe this, it seems to be very
important.
You also write there's a difference between the maximum computed by rrdtool
and the maximum computed by sql. Are you aware that timestamps in rrdtool
are describing a time interval? E.g. the rate stored at 1226541600 is valid
from 1226541600-3600 upto 1226541600. That same rate could be found in sql
at timestamp 1226538000 (I dunno, I'm guessing here, sort of).
Why don't you generate a simple database (it doesn't need to be large),
perform a couple of updates on it, run rrdtool fetch. Put all the commands
inside a script so that anyone can reproduce it. Also add a description
what the numbers are intended to represent, what the consolidation should
(in your opinion) look like, and why. Post the script on this list, and I'm
sure you'll get plenty of support.
Don't waste time. Keep the database very simple and the amount of updates
low. It needs to be enough to show what you mean, it does not need to
reflect your current project.
Most likely I won't take a look at it very soon but I'm sure others on this
list are equally capable of looking at your script and your train of
thought.
It may indeed be that rrdtool won't do what you need for your project. You
may want to use a regular database and a graphing program. But in most cases
I've seen problems resulted from a lack of understanding (user side), not
capabilities (rrdtool side). We can work on that.
HTH
Alex
----- Original Message -----
From: "Gavin Landon" <Gavin.Landon at ignitetech.com>
To: <rrd-users at lists.oetiker.ch>
Sent: Wednesday, November 12, 2008 11:54 PM
Subject: Re: [rrd-users] Absolute values..
> Thanks for the response. However, after looking over the email sent
> out by Alex and looking through the link he posted, it makes me think
> that RRD isn't going to handle what I'm trying to do. I want no
> response alterations, only exactly what I'm putting in. Based on the
> setting of GAUGE and the use of MAX, shouldn't it do that? Maybe I'm
> misunderstanding the documentation.
>
> Thanks again,
> Gavin
>
>
> -----Original Message-----
> From: andy.riebs at hp.com [mailto:andy.riebs at hp.com]
> Sent: Wednesday, November 12, 2008 3:18 PM
> To: rrd-users at lists.oetiker.ch
> Cc: Gavin Landon
> Subject: Re: [rrd-users] Absolute values..
>
> Gavin,
>
> Please see the mailing list archives from the past 5 days. Alex B (?)
> posted an outstanding, and brief, explanation.
>
> Andy
>
> On Wednesday 12 November 2008 04:04:30 pm Gavin Landon wrote:
>> Was this a newb question and no one wants to answer, no one has ever
>> experienced, or no one has ever validated the data so if it's happen,
>> no one noticed?
>>
>> -----Original Message-----
>> From: rrd-users-bounces at lists.oetiker.ch
>> [mailto:rrd-users-bounces at lists.oetiker.ch] On Behalf Of Gavin Landon
>> Sent: Wednesday, November 12, 2008 11:14 AM
>> To: rrd-users at lists.oetiker.ch
>> Subject: [rrd-users] Absolute values..
>>
>> I'm having a few problem which has interrupted my train of thought to
>> the point I'm starting to doubt myself.
>>
>> (Problem 1)
>> I've been writing a monitoring system for software we have written
>> that report back to us on a minute by minute bases. I'm only in the
>> development stages so the data is small and in one case it always
>> returns 1. I find the chart that's looking at this stat is in flux as
>
>> if the values are more than 1. I've deleted the RRD file then
>> recreated it with data I have in SQL, logging everything that goes
>> into RRD and I never put in anything greater than 1. Yet my chart
>> still has spikes at the same place up to 20.
>>
>> (Problem 2)
>> The answer to this one may fix problem 1.. I can't seem to find the
>> balance in the configuration so that the numbers on the side of the
>> left side of the chart match up with the data that is actually in the
>> RRD file. Since I'm also storing the data in SQL at the same time
>> it's being placed into RRD, I know I have an exact match up. However,
>
>> when I tell SQL to give me the MIN and MAX of a specific time frame,
>> then tell RRD to do the same, I get two different values.
>>
>> I do have 5 archives.
>> Min = .Steps = 1
>> 24 hours = .Steps = 60 'Minutes per Hour
>> Week = .Steps = 1440 'Minutes per Day
>> Month = .Steps = 10080 'Minutes per Week
>> Year = .Steps = 43829 '525948.766/12 'Minutes Per Month
>>
>> I'm using Gauge when creating it since Absolute, Counter, and Derive
>> all seem to skew the charts as well. On my archives I'm using MAX
>> function because I found that Average was throwing me way off.. Since
>
>> this is a monitoring interface, I need it to be exact.
>>
>> Maybe it's me and I just don't have it configured correctly.. If so,
>> can someone tell me the configuration needed to show what is exactly
>> in RRD, because showing my clients 189 for SQL, but looking at a chart
>
>> that doesn't go over 40 is only going to prompt questions I can't
>> answer.
>>
>> Thanks for your time,
>>
>> Gavin
>>
>> _______________________________________________
>> rrd-users mailing list
>> rrd-users at lists.oetiker.ch
>> https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
>>
>> _______________________________________________
>> rrd-users mailing list
>> rrd-users at lists.oetiker.ch
>> https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
>
>
>
> --
> Andy Riebs, HP Scalable Computing and Infrastructure / Linux R&D
> (w) andy.riebs at hp.com Technology for better business outcomes
> +1 (978) 748-0043 (Mon,Thu,Fri)
> +1 (508) 467-3002 (Tue,Wed)
> (h) andy at candooz.com My opinions are not necessarily those of HP
>
> _______________________________________________
> rrd-users mailing list
> rrd-users at lists.oetiker.ch
> https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
>
More information about the rrd-users
mailing list