From: Kirby Krueger <kirbyk@corp.webtv.net>
Jeff Allen wrote:
>
> From: Jeff Allen <jra@corp.webtv.net>
> > data? So is it better to have a complete different cricket for different
> > polling intervals?
>
> It's definately possible to use a separate Cricket setup to do fast
> polling.
>
> Cricket is designed to be able to poll different subtrees on different
> rates, but this is not a function we have completely tested and gotten
> working. A critical change to make this work was to reformat the
> targetType dictionary, which is done. You need to be certain to reset
> "rrd-poll-interval" in the root Defaults file to reflect the shorter
> polling rate.
>
I'm actually doing this. It's not really too hard. I'm graphing some
data about blocked mail from syslog output, which is due to the nature
of the logfiles easiest to get at in 12 hour bundles. My Defaults file
in the spam-stats directory contains:
# Polling every 12 hours, for 600 days of data
rra 12hour AVERAGE:0.5:1:1200
targetType spam-stats
rra = "12hour"
graph --default--
default-ranges = "m:y"
We decided that, due to the relatively small number of data points
needed to keep 600 days, we didn't need to roll the data for the
longer-term graphs. Note the default-ranges set to m:y - that makes the
graphs that show up default to the long-term values, which are more
interesting for data coming in this sparsely.
The only other difference to make the polling rate different is to
create a different set in subtree-sets (or subtrees, for pre-0.65
installations) and have cron run that at the correct increments.
This is a fairly experimental fringe of cricket, admittedly, but it
seems to be working for me just fine so far.
-- Kirby
------------------------------------------------------------------------
Campaign 2000 is here!
http://www.onelist.com
Discuss your thoughts; get informed at ONElist. See our homepage.
This archive was generated by hypermail 2b29 : Mon Mar 06 2000 - 19:00:50 PST