[rrd-users] rrdfetch resolution problem

Simon Hobson linux at thehobsons.co.uk
Wed May 2 22:54:47 CEST 2007

Matt Pounsett wrote:

>my @TrackRRTypes = qw( A AAAA ANY CNAME MX NS PTR SOA );
>my @args = ( "rrds/$host.rrd",
>          qw(--step 300 ),
>          "-b", $start_time-300,
>          (map { "DS:$_:ABSOLUTE:600:0:10000000" } @TrackRRTypes ),
>          'DS:OTHER:ABSOLUTE:600:0:10000000',
>          'DS:TOTAL:ABSOLUTE:600:0:10000000',
>          'RRA:AVERAGE:0.5:1:2016',
>          'RRA:MIN:0.5:12:17250',
>          'RRA:MAX:0.5:12:17250',
>          'RRA:AVERAGE:0.5:12:17520',
>          'RRA:MIN:0.5:288:3650',
>          'RRA:MAX:0.5:288:3650',
>          'RRA:AVERAGE:0.5:288:3650',
>RRDs::create( @args );

You only store 7 days worth of data at 300s resolution. Your basic 
step is 300s, your first RRA is 2016 time 1 step 
(RRA:AVERAGE:0.5:1:2016) which works out to 168 hours or 7 days.

Beyond that you CANNOT get 300s resolution data because it no longer exists.

The next step is to store min, max, and average for a little under 
two years at 1 hour resolution. RRA:AVERAGE:0.5:12:17520 means to 
store one consolidated value for every 12 steps (12 x 5m = 1h), and 
to store 17250 values which works out 718.75 days.

After that you have 24hr samples (5min * 288 = 24h) for 10 years (3650 days).

So if you ask for data covering anything from 7 days to nearly 2 
years ago you will get 1 consolidated value per hour. If you go back 
2 years or more then you will only get one value per day. RRD tools 
will NOT give you mixed data - if you really wanted to plot a graph 
over say the last 2 weeks and use the highest resolution data then 
you would have to extract two sets of data (up to a week, beyond a 
week) and plot them using some other tool.

More information about the rrd-users mailing list