[rrd-developers] Re: rrdgraph & step sizes

[Burton Strauss]

(  Tee Hee Hee... Told you Tobi that people would want this... :-)  )

No, you can't.  The code looks for the smallest step size that has both the
start and end points within it.  The docs and code blather about 'best' but
that's not really what happens.  It doesn't take into account coverage (NaNs
etc.) just whether it has the range of time from start to end in it.  If you
want to look, it's the block of code in rrd_fetch.c around line 165
(1.0.49), beginning  /* find the rra which best matches the requirements */

I played with some code for 1.0.49 that adds a feature to force selection of
a specific rra, but held off
 (1) because Tobi didn't see a need for it 
and
 (2) because at the time, Tobi wanted me to rework it for both 1.0.x and
1.2.x.  Which I didn't have the time, nor experience w/ 1.2.
and
 (3) Because the 'right' answer is (Quoting Tobi):

"well but in this case the fetch code should be improved so that it can
assemble data from RRAs with different resolutions ... filling in 'holes'
from higher resolution RRAs ... eg take the 5minute RRA for the first 72h
and fill in the rest with data from the 30 day RRA ?"


And of course, what you are asking for is #3.  Never-the-less, I do have
some code which might help you, if you can live with forcing a specific RRA
and suffering it's associated gaps/NaNs.  The patch I have - unsupported, of
course - would add a --rra=# parameter to force selection of a specific rra.
If you are interested, let me know off-list...

-----Burton (burton at ntopSupport.com)


-----Original Message-----
From: rrd-developers-bounce at list.ee.ethz.ch
[mailto:rrd-developers-bounce at list.ee.ethz.ch] On Behalf Of Greg Prosser
Sent: Friday, June 10, 2005 2:36 PM
To: rrd-developers at list.ee.ethz.ch
Subject: [rrd-developers] rrdgraph & step sizes

Hi All,

I increased my rrd's to contain 60000 5 minute average (step=300) rows in
order to do some statistical analysis on this data.

As a result, rrdtool is now picking this sample set to generate all graphs
(other than one year ones) as its start/end times match the requested
start/end times.  The problem inherent in this is that I'd much rather have
the 30 minute average data set (for example) used to generate the weekly
graph, as I don't have enough 5 minute average data to populate the graph.
The same goes for the monthly graph.

Is there a way I can force rrdtool to pick a certain RRA?  When consulting
the documentation (and the code) it seems that a DEF: line causes rrdtool to
try to find something with the proper start/end times with a step size less
than the requested step size (with --step).
However, rrd doesn't pick my 30 minute average RRA's (or others) because my
requested graph runs beyond the "last update" of the 30 minute RRA.

I could, I guess, just live with it for the 3 weeks it would take to fill in
the data, but I'd much rather have a way to work around that.
Any comments?

BTW: After reading the list, and seeing some of the strangeness in 1.2, I'm
still using 1.0.  (1.0.49 to be exact)  I'd rather not upgrade unless this
specific feature is found in a later version, as I have a setup that works
right now and don't want to break it.  :-)

-gnp


--
Unsubscribe
mailto:rrd-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-developers-request at list.ee.ethz.ch?subject=help
Archive     http://lists.ee.ethz.ch/rrd-developers
WebAdmin    http://lists.ee.ethz.ch/lsg2.cgi

--
Unsubscribe mailto:rrd-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-developers-request at list.ee.ethz.ch?subject=help
Archive     http://lists.ee.ethz.ch/rrd-developers
WebAdmin    http://lists.ee.ethz.ch/lsg2.cgi