[rrd-users] Daily consumption

[Alex van den Bogaerdt]

> You have "--step 60" and update every 10 seconds. Your finest RRA will
>> still contain a composed number built from 6 updates.
>>
>
> Update every 10 seconds is a little bit "aggressive" but it is not really
> needed, you're right. The reason is that when I run my tests, I only need
> 10s to have an update and check that everything works (graphs update is
> also every 10s)
> But do you confirm that I can update with a higher freq than de DB ? Let's
> say update every 30 sec still have sense ?

At some point the timing error will be noticeable but yeah, you can.
Personally I would also have an RRA containing 10 seconds per interval, or
not update this often. But this is not required.

Unless memory / disk space is limited, I would probably want to store at
least one hour worth of 10-second data, so that I have time to switch on
some equipment and then can still view the graph. But that's just me. You
can do what you want, just as long as you remember that RRDtool will do
what you tell it to do, not what you want it to do :)

>> If you mean: one vertical line / bar per day, then you need to be aware
>> that RRDtool works in UTC time. Your "days" are off by one hour during
>> winter, and by two hours during summer.
>>
>
> Thanks for the warning, but it is not a very big issue in my case (unless
> for the 23h and 25h days which will show abnormal low or high
> consumption).

For RRDtool, these two Sundays are just 24 hours long (UTC time).
The one in March will start at 1am local time and end 2am local time,
the one in October will start at 2am local time and end at 1am local time.
Still, both days are 86400 seconds and contain power consumption during
those 86400 seconds. So, unless you always have a much larger amount of
power usage between 1am and 2am, I don't expect to see a big difference
when compared to the other 50 Sundays in a year.

My point was: don't expect to get midnight local time to midnight local
time. You won't.

cheers,
Alex