[rrd-users] Newbie question: Always missing the last datapoint in rrdtool graphs

Ole Holm Nielsen Ole.H.Nielsen at fysik.dtu.dk
Wed Feb 25 23:05:41 CET 2009

Hi Simon,

Thanks a lot, your suggestions seem to make a lot of sense, although
I could never have guessed this based on the rrdtool documentation !
I'll redefine my RRD databases tomorrow and see if things improve.

One point makes me really worried, though, namely your explanation
of the timezone problem.  I do live in the UTC+1 timezone (and UTC+2
during the summer), so you're telling me that I'll *never* be able to
map my daily 24-hour data into a single timestep in an RRD database ??!!

Surely, there must exist an elegant solution where daily (24h) data can
be mapped one-to-one into an RRD database, no matter which timezone
you live in ?  Can one define an RRA with a 1 hour step and use that
in some way ?  There must be lots of people who have needed to solve
this problem previously !  Any suggestions would be most welcome.

Thanks again,

>> Question 2: What am I doing wrong in the above database definitions,
>> and how should I create the database so that it'll support my needs ?
> Try changing your updates to midnight instead of 1 second before midnight.
> Two reasons :
> 1) 1 second before midnight isn't on a step boundary so you are 
> forcing rrd tool to normalise the data - 86399 seconds worth into one 
> step, and one second into the previous one.
> 2) Your step period has not ended at 23:59:59 so you won't get any 
> data out for the preceding 23:59:59 - until you do another later 
> update and rrdtool knows what value is being fed in for that last 
> second.
> Also, bear in mind that steps are ALWAYS relative to UTC - so if your 
> timezone (not mentioned but your email says +1) isn't UTC, then the 
> data will be normalised accordingly - eg 'midnight' is really 
> 23:00:00 UTC and so you will be putting 86339 seconds into one step, 
> and 61 seconds into the previous step.

More information about the rrd-users mailing list