Problem with Time series data entry

[Chris G.]

I have some historical time series data.  The data was collected approx. every 15 mins, I say approx. because the times are not exactly 15 mins to the second apart.  In some cases there could be 10-15 seconds of delay  between each of the 15 min collections.  In some cases if a collection cycle was skipped a time entry less then 15 mins will occur.   For example, data collected at 10:00:01, 10:15:07, 10:30:03, 10:45:11, 10:48:13, 10:51:12, 11:00:32.

I have created the following:
rrdtool create test.rrd --start 1179721292 --step 900 DS:value:GAUGE:1830:-5:5000 RRA:LAST:0.5:1:672 RRA:AVERAGE:0.5:4:2160

I expected this to allow me to keep the actual measured value and the time the value was collected for 7 days and then the hourly average for the previous 90 days.

I update test.rrd with the following:

rrdtool update test.rrd 1179722192:0.359
rrdtool update test.rrd 1179723089:0.343
rrdtool update test.rrd 1179724001:0.329
rrdtool update test.rrd 1179724892:0.359
rrdtool update test.rrd 1179725797:0.36
rrdtool update test.rrd 1179726695:0.375
rrdtool update test.rrd 1179727604:0.36
rrdtool update test.rrd 1179728493:0.359
rrdtool update test.rrd 1179729394:0.375
rrdtool update test.rrd 1179730293:0.344

When I perform a rrdtool fetch the first data entry I get has a timestamp of 1179721800 this is not the date/time I passed in with the update.

Is it not possible to collect/store and then graph time series data using the exact time the measurement was taken.

Any help/guidance would be appreciated.

Chris G.

[Simon Hobson-2]

Chris G. wrote:

>I have some historical time series data.  The data was collected approx.
>every 15 mins, I say approx. because the times are not exactly 15 mins to
>the second apart.  In some cases there could be 10-15 seconds of delay
>between each of the 15 min collections.  In some cases if a collection cycle
>was skipped a time entry less then 15 mins will occur.   For example, data
>collected at 10:00:01, 10:15:07, 10:30:03, 10:45:11, 10:48:13, 10:51:12,
>11:00:32.
>
>I have created the following:
>rrdtool create test.rrd --start 1179721292 --step 900
>DS:value:GAUGE:1830:-5:5000 RRA:LAST:0.5:1:672 RRA:AVERAGE:0.5:4:2160
>
>I expected this to allow me to keep the actual measured value and the time
>the value was collected for 7 days and then the hourly average for the
>previous 90 days.

>When I perform a rrdtool fetch the first data entry I get has a timestamp of
>1179721800 this is not the date/time I passed in with the update.

No it won't be the timestamp you passed in the update UNLESS that
timestamp is a multiple of the step size you used when creating the
rrd. Unless every update is timestamped exactly on a step boundary,
then it will be normalised to fit those steps.

>Is it not possible to collect/store and then graph time series data using
>the exact time the measurement was taken.

No, this is not what rrd was designed for, if you want to do that
then use a more appropriate tool. Rrd was designed to do one task,
and it does it very well, to store historical data which is
normalised and consolidated as it is entered so as to a) minimise the
processing required for data extraction/graphing and b) to minimise
the data storage requirements.

>Any help/guidance would be appreciated.

As well as the project web pages, there is some excellent information
at Alex van den Bogaerdt's site http://www.vandenbogaerdt.nl/rrdtool/ 
In particular, see under the tutorial section where it says :

>>Rates, normalizing and consolidating explains how RRDtool processes
>>its input data, normalization and consolidation are described there
>>as well as some generic stuff about the properties of an RRA


_______________________________________________
rrd-users mailing list
rrd-users@...
https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users