0

I have some custom munin graphs that measure an aspect of our sales in our webapp. So I measure "Number of sales so far today". The graph goes up and then reset to 0 at midnight. This is what I want. It works well for the "by day", 5 minute one. Munin by default makes 3 extra graphs showing a longer term view, it has a "by week", "by month" and "by year" view which is based (AFAIK) on the combined data.

However the default seems to be an average of the values, so the total number of sales for a day is an average. I'd much much rather if the combined weekly/monthly/yearly data is based on the maximum/largest for that period. So rather than averaging, it'll 'maximise'?

I'd prefer if this can be done without losing our existing data, however I'm willing to lose the existing data/graphs if that's what's needed.

Amandasaurus
  • 31,471
  • 65
  • 192
  • 253
  • I too would like to have a better solution to this. Currently I use an additional 'daily max' plugin for those type of metrics, which just spits out the maximum value from the day before. This makes the week/month/year graphs meaningful, but it's a bit of a hack, requires your previous max value to be cached somewhere etc. – Moby Duck Dec 18 '13 at 13:24

2 Answers2

1

I would love to know this answer too.

I suspect it would be useful to review what RRDTool can do (this looks possibly useful: http://www.vandenbogaerdt.nl/rrdtool/total.php ) and then hack munin? Or perhaps the CDEF function will let you do this: http://munin-monitoring.org/wiki/fieldname.cdef has some useful info and a link to a tutorial.

You could also try asking your question on the munin-users mailing list. I looked through the plugin exchange and didn't see anything that seemed appropriate.

mooreds
  • 111
  • 3
1

I might be wrong, but isn't then Munin/RRDtool the wrong tool for your needs? As far as I know RRDtool is exactly meant to keep fewer data samples the further away the date is. (sorry no native english speaker, hope you understand what I mean).

To clarify, even if you could select the maximal value, you might not be sure that this was the absolute value, as not all data samples were kept in the rrd files.

weeheavy
  • 4,089
  • 1
  • 28
  • 41