[rrd-users] Problem when restarting "device" - Getting huge negative values using DERIVE

[Jose Oscar Olivares Ocampo/Mexico/IBM]

     Hello everyone.
I have a situation that others might have already faced.
Our application collects data every half hour, we use Derive as our DST. 
Normally, we expect the values to increase during the day; however, by the 
end of the day, these values tend to be very high, so the "device" is 
reset every day at midnight.
The problem with this is that we get a gap at that time the device is 
reset.
We use a minimum value of 0 for the Data Source, but I know that the 
calculations done by rrd are generating a negative value and since less 
than 0 is not acceptable, it is filled with NaN.

What can I do to take this reset of the device and allow a smoother 
transition? (do not show a gap but instead maybe a peak within positive 
numbers)
Can I indicate to rrd that the count was started all over again?
What other suggestions do you have?


The detail information to reproduce the problem as follows.

- The file is created this way:
RRDs::create ("FileSample.rrd",
         "--start", 1278219600,
         "--step", 1800,
         "DS:Accesses:DERIVE:3600:0:999000111222",
         "RRA:AVERAGE:0.5:1:672"
);

The updates are done with this data (Time / Value):
22:16:11        831433273
22:39:25        881374388
23:18:04        911450591
23:47:32        944041059
00:17:20          16590140
00:45:30          78026088
01:13:57        110711741
01:46:25        135500180
02:15:35        157698405
02:43:35        177473994
03:12:45        196935506
03:41:31        215507899

The problem occurs on the time the counter is decreased from 944041059 to 
16590140

Best Regards!

José Oscar Olivares Ocampo
Infrastructure Tools
Tie Line: 877-3115
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20100708/b9fa23d0/attachment-0001.htm