[hobbit] strange graph behavior - random machines & graphs

Gary Baluha gumby3203 at gmail.com
Wed Dec 5 20:05:51 CET 2007


cd hobbit_data_dir/host_machine
rrdtool dump clock.rrd > clock.xml

I know any number that shows up greater than "e+1nn" is bogus, so I search
for "e+1".

One of several bogus data lines:
<!-- 2007-11-26 19:00:00 EST / 1196121600 --> <row><v>
3.9551632477e+169</v></row>

Same line, changed to NaN (repeat for all affected lines):
<!-- 2007-11-26 19:00:00 EST / 1196121600 --> <row><v> NaN </v></row>

rrdtool restore clock.xml clock.rrd

On Dec 5, 2007 11:57 AM, Kern, Thomas <Thomas.Kern at hq.doe.gov> wrote:

>  Could you give a short example of a bogus and a changed (NaN) entry, just
> in case that is also happening to some of my data files?
>
>
> /Thomas Kern
> /301-903-2211 (O)
> /301-905-6427 (M)
>
>
>  ------------------------------
> *From:* Gary Baluha [mailto:gumby3203 at gmail.com]
> *Sent:* Wednesday, December 05, 2007 11:53 AM
> *To:* hobbit at hswn.dk
> *Subject:* Re: [hobbit] strange graph behavior - random machines & graphs
>
> I am now completely convinced that the strange behavior of the graphs is
> due to some bad data getting inserted into the .rrd database files.  The bad
> data is always the same value: 5.1776682516e+170.  That's what the value
> looks like when you do an rrddump on the .rrd database file.
>
> I still have no idea where this value is coming from, but I have at least
> determined how to fix these graphs.  I'm working on a script to do this, but
> for now, I manually do an rrddump of the file, change all bogus values to
> NaN (basically, searching for "e+1", since none of the values I trend
> generally get that large, so I know these entries are just averaged values
> of correct data and the 5.17... number), and then do an rrdrestore from
> the modified xml file.
>
> It would be nice to determine where this problem is coming from, though.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20071205/3007050a/attachment.html>


More information about the Xymon mailing list