[rrd-users] spikes or junk in rrd files

[Steve Bishop]

I have been using rrdtool since August, or maybe even before.  I
have followed it since the 1.0 series releases.

I like what the idea, and what it does, but I have had
problems with random spikes getting into the data and messing
up the rrd's.  These spikes consist of large negative numbers, large
positive numbers, and also very small negative numbers.

At first, I thought it was a problem with my system, or the
way I was using the rrd tools, but now I don't think so.
I have been trying to figure this out for several months now.

For the last month and a half, I have been running using the 1.0.7 release.  My system
enters the data into the rrd using the shared library perl module (RRDs.pm).
I am also running on FreeBSD 3.3-RELEASE.

So far, I have been able to determine that it seems to occur when a data
point is missing.  I collect my data every 15 minutes, and update the rrd's
The collected data has 5 minute resolution.

Occasionally, I found that one of the 5 minute data points was missing.  At first,
I had my heartbeat interval set to 750.  Then, I decided that maybe that was
too long, and the heartbeat interval was the culprit.  So, I shortened it to 600
seconds.  When that didn't work, I changed my program so that it would report
the 5 minute data EXACTLY when it was expected (e.g. 10:00:00, 10:05:00, ...)

This still didn't fix it.  I changed my heartbeat interval to 300 seconds, and I'm still
having problems.

As a last resort, I put limits (min,max values) into my rrds.  This hasn't
worked either.

Here's my RRD definitions.

                                "DS:load:GAUGE:300:0:5000",
                                "DS:out_oct:COUNTER:300:0:12500000",
                                "DS:drop_oct:COUNTER:300:0:U",
                                "DS:out_pkts:COUNTER:300:0:50000",
                                "DS:drop_pkts:COUNTER:300:0:U",
                                "RRA:AVERAGE:0.5:1:864",      # collect 5 min data for three days
                                "RRA:AVERAGE:0.5:3:1344",     # collect 15 min data for two weeks
                                "RRA:AVERAGE:0.5:12:744",     # collect hourly data for one month
                                "RRA:AVERAGE:0.5:288:91",     # collect daily data for three months


Here's a list of the random spikes that I have found in 15000 rrd files.
Each line is preceded by a count indicating the number of times it occurred.
There were a total of 3209 files w/ errors.




-- Attached file removed by Listar and put at URL below --
-- Type: text/plain
-- Size: 92k (94790 bytes)
-- URL : http://www.ee.ethz.ch/~slist/pantomime/fix.unique


--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-users