[hobbit] mangled and/or assumed data in rrds with 4.3.0-0.20080327 snapshot

[Tom Georgoulias]

Tom Georgoulias wrote:
> I think I'm seeing some rrd bugs in our Hobbit snapshot, 
> v4.3.0-0.20080327, or something in our config is really out of whack.
> 
> 1.  When a new host is added to bb-hosts, the first data that goes into 
> the rrd for the conn test is mangled and wildly inaccurate.  The trend 
> charts have an Y axis with a scale like 10 ?, 20 ?, etc., and the rrd 
> file is filled to the 576th day with data, most of which looks like this:
> 
> <!-- 2007-09-30 20:00:00 EDT / 1191196800 --> <row><v> 8.5485381823e-72 
> </v></row>
> 
> 2.  When a host goes offline and stops reporting, sometimes the rrds 
> associated with the disk trend charts simply fill in the missing data 
> with 0s instead of NaN.
> 
> Any ideas?  I didn't find anything useful in the rrd-status.log.\

I'm also seeing this on some memory and procs graphs as well.  For 
example, attached is the memory graph for newly added rhel5 based 
server.  Notice how the "?" found its way into data and now the chart 
and rrd file are essentially worthless.  I also attached an image of the 
tcp ping test to better illustrate #1 from my previous message.

This seems to happen to systems that are new, fresh additions to hobbit, 
not to systems that have long histories.

I am using rrdtool 1.2.23 and hobbit snapshot 4.3.0-0.20080327 on a 
redhat enterprise linux 5.1.  Debugging tips or help is appreciated.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: memory.png
Type: image/png
Size: 15814 bytes
Desc: not available
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20080528/13165290/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ping.png
Type: image/png
Size: 12624 bytes
Desc: not available
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20080528/13165290/attachment-0001.png>