[rrd-developers] Re: (bug?) Incorrectly selecting low-res RRA
P.Radden at rl.ac.uk
Fri Mar 28 17:49:14 MET 2003
On Thu, 27 Mar 2003, Alex van den Bogaerdt wrote:
> [very helpful explanation]
Thanks very much for your mail yesterday; I believe I now understand what
you're suggesting (sorry for being slow!), so I've added some additional
test to my monitoring script to try some extra things.
Unfortunately this debugging is going very slowly because I have to wait
until the strange behaviour hits each time...
Based on other suggestions, I'd previously tried shifting back a time
period equivalent to one row, to avoid what you described as the
'yet-to-be-archived-partially-filled-timeslot' (but not rounding to a
multiple of that time period, as I now understand you to be suggesting)
but this made the situation worse, in that the lower resolution RRA was
chosen for about twice as long last night.
This caused me to realise that the 'coverage before resolution' rule is
also applying at the beginning of time, not just the end of time, and that
my 'month' rra is only specified to include just the right amount of data
- if I understand things better now! That's probably a ganglia bug if
anything, and I'll let them know.
In the meantime, tonight I'll be testing a variety of counts including
LINES=`rrdtool fetch "$CHECK" AVERAGE -s -2419200 -e N | wc -l`
LINES2=`rrdtool fetch "$CHECK" AVERAGE -s -2429280 -e -10080 | wc -l`
LINES4=`rrdtool fetch "$CHECK" AVERAGE -s end-2419200 -e $RNDNOW1 | wc -l`
LINES5=`rrdtool fetch "$CHECK" AVERAGE -s -2419200 -e $RNDNOW1 | wc -l`
LINES6=`rrdtool fetch "$CHECK" AVERAGE -s end-2419200 -e $RNDNOW2 | wc -l`
LINES7=`rrdtool fetch "$CHECK" AVERAGE -s -2419200 -e $RNDNOW2 | wc -l`
to see if I really understand.
Thanks once again,
Unsubscribe mailto:rrd-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-developers-request at list.ee.ethz.ch?subject=help
More information about the rrd-developers