Xymon Mailing List Archive search

graphing questions

2 messages in this thread

list Phil Wild · Fri, 28 Mar 2008 17:00:36 +0900 ·
Hello hobbit users,

Three items....

We monitor the size of oracle tables in a similar way as the base disk
monitoring is performed. We use a custom script at the back end to create
the rrd's as new tablespaces appear (i.e. one tablespace per rrd)

The tablespace sizes are graphed in a similar way to the disk graphs as well
with the standard hobbit method of dividing the tables into smaller groups
(I think it is 7) and charting them on one graph.

The issue we have is that some tables are very small (say 100MB and some are
large 30GBytes). When these are plotted on the same graph, the plots for the
small tables are of little value as they look like zero...

I assume the only way to address this would be to plot each table
individually. Anyone have any ideas how I can force this?

Next....

I want to stack all the tables onto an area chart to show total database
growth. However, the first value needs to be area and the rest of the values
stack. How can I achieve this when a regex is used to pick up the rrd's?

And last, I am toying with the idea of trending growth rate. It would seem
that the rrd database would make a good source of historical data for this.
Has anyone done something here? One idea I am playing with is if a
filesystem breaches a capacity threshold out of hours, but the greatest of
the last five minute and 30 minute growth rates would indicate that it is
not going to run out of space before 10AM on a business day, then don't send
me an sms to wake me up. Thoughts?

Phil
list S Aiello · Fri, 28 Mar 2008 08:15:05 -0400 ·
quoted from Phil Wild
On Friday 28 March 2008, Phil Wild wrote:
Hello hobbit users,

Three items....

We monitor the size of oracle tables in a similar way as the base disk
monitoring is performed. We use a custom script at the back end to create
the rrd's as new tablespaces appear (i.e. one tablespace per rrd)

The tablespace sizes are graphed in a similar way to the disk graphs as
well with the standard hobbit method of dividing the tables into smaller
groups (I think it is 7) and charting them on one graph.

The issue we have is that some tables are very small (say 100MB and some
are large 30GBytes). When these are plotted on the same graph, the plots
for the small tables are of little value as they look like zero...

I assume the only way to address this would be to plot each table
individually. Anyone have any ideas how I can force this?
Well you could do as the disk trends graphs, graphing by percent used. Percents won't help you with your other 2 questions, since with them I would assume you would want the amount used. You could have 2 data stats in your RRDs %Used & Used. Then define different graphs to used each stat for the different purposes.

As to your other 2 questions, I have no past experience to pull from for this. I would think it is possible, would just need to figure out the dynamics to do this within rrdtool, then apply that to hobbit's graph definition.

 ~Steve