[collectd] enhancment for threshold (make a notice when value is in range)

Florian Forster octo at verplant.org
Fri Mar 14 15:01:15 CET 2008


Hi,

sorry for this flood of mails, but I finally got around to read mails
again. Hooray for fridays! ;)

On Fri, Mar 14, 2008 at 10:59:14AM +0100, Sebastian Harl wrote:
> Afaik, this has not been implemented so far.

Right.

> This could be done in a very flexible way by introducing some kind of
> consolidation function which would be applied to the values before
> deciding whether to send a notification. I have some very vague ideas
> about that but nothing yet which would be worth mentioning - so any
> further ideas would be very welcome :-)

I've a couple of ideas that might be interesting to talk about:
- Some sort of grace time: I. e. only send a notification if `n'
  consecutive values are ``bad''. Or, more general: Alert if at least
  `m' out of `n' values are ``bad''.
- Some sort of moving average. I. e. check if
    (f * last_value) + ((1 - f) * current_value)
  is okay or not.
- Check average of `n' samples.

Now I'd like to have some flexible way to formulate as many of the above
things as possible. However, I didn't come up with anything nice yet.
Ideas, anyone?

What about something like that:
  Consolidate CountSamples
  Samples 7 OutOf 10
- or -
  Consolidate MovingAverage
  Factor 0.1
- or -
  Consolidate FixedAverage
  Samples 6

Regards,
-octo
-- 
Florian octo Forster
Hacker in training
GnuPG: 0x91523C3D
http://verplant.org/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://mailman.verplant.org/pipermail/collectd/attachments/20080314/748a2d09/attachment.pgp 


More information about the collectd mailing list