Constantly incrementing data population delays

[goyette]

Hello. This is kind of a general question regarding how RRDTool is going to handle a certain type of data population, so I know whether Cacti/RRDTool can solve my problem.

The basic idea is that we use Nagios to monitor performance data, and then feed it into RRDTool/Cacti. The trouble is, because it takes Nagios a second or two to execute the command that generates the data, the results of my minutely checks do not occur exactly on the minute. Instead, they occur in a constantly incremementing pattern. For example, here's a sample list of when my results might come back, and be timestamped:

12:00:00
12:01:01
12:02:01
12:03:03
12:04:04
12:05:06

The problem occurs when the seconds "wrap", creating a missing bit of data, for example:

12:45:58
12:46:59
12:48:00
12:49:01

So, if I wanted to analyze my data, in minutely intervals (Say, plotting a point representing the data from 12:00:00 to 12:00:59, and the next point would represents 12:00:01 to 12:01:59), I will end up with a gap in my data. That is, the the second list of values, I have results for 12:45->, 12:46->, and 12:48->, but I'm missing the data from 12:47->. I'm not really missing ANY data, but just the fact that delays exist in populating the data I will always end up with cases where there will be this skip.

I know RRDTool uses a heartbeat to prevent this kind of thing, but it seems that, eventually, incoming data will all have the heartbeat applied to it. So, RRDTool will wait up to an extra minute beyond when it expects the data. But, given this delay, eventually it will consider that it has waited too long, and enter an *UNKNOWN* value for that period.

Anyone know how this is reflected in the graph? Or a way around this? What do people do? Or maybe this is a more global graphing theory question.

Will RRDTool/Cacti end up creating a graph with missing points every once in a while when the delay wraps around enough and resets? Otherwise, how does RRDTool keep from eventually entering an UNKNOWN value for results that are always coming back slightly longer than the theorectical check interval?

Thanks,

-Dan

[gandalf]

If the heartbeat time is not yet exceeded, rrdtool will provide for correct adjustment of data that was gathered with a little time jitter, see http://www.rrdtool.org tutorials section. Under special circumstances rrdtool will interpolate missing readings, but this can be suppressed using the "correct" rrd definitions. Again, this is described at the rrdtool site. Both aspects are pure rrdtool issues
Reinhard