If you look through the source, the RRD support modules are easy to spot. If you are using custom graphs, then you need to review the ncv method, or the "roll your own" method.
But, since "garbage in -> garbage out" you might be on the right path.
GLH
From: Gary Baluha [mailto:gumby3203 at gmail.com]
Sent: Monday, April 07, 2008 1:15 PM
To: hobbit at hswn.dk
Subject: Re: [hobbit] strange graph behavior - random machines &graphs
I thought I'd revisit this issue again. A new thought hasoccurred to me... Where does Hobbit generate the RRD files? I wonder what parameters Hobbit is using to pass to rrdtool, and if something there might be acting funny with some of the data I'm providing to that Hobbit module.
On Wed, Dec 19, 2007 at 10:59 AM, Gary Baluha<gumby3203 at gmail.com> wrote:
It's interesting that it seems the CPU Load and Usersand Processes graphs are the graphs that are most likely to have this strange corruption. I have also seen it on a few Disk graphs, but not nearly as many as the other two graphs. Interestingly, the CPU Utilization, Network I/O, and TCP Connection Times graphs have _never_ had this corruption. I'd also like to say the Memory Utilization graph hasn't had this issue either, though I can't recall with complete certainty that that is the case. I wonder what the main difference between the 3 graphs that do have the issue is, and the 3 (possibly 4) graphs that have never exhibited this issue. There must be some physical difference, as I can't imagine it is all due purely to luck...