[collectd] [FR] notification severities += MISSING
Andrés J. Díaz
ajdiaz at connectical.com
Tue Mar 1 18:00:05 CET 2011
On 1 March 2011 14:21, Michael Shigorin <mike at osdn.org.ua> wrote:
Hi Michael et al,
> Just got a thousand and a half of paired email notifications like:
> hostname/ipmi/temperature-Baseboard Temp system_board (7.1) has not been updated for 20 seconds.
> Received a value for hostname/ipmi/temperature-Baseboard Temp system_board (7.1). It was missing for 20 seconds.
Yep, it's a veeery annoying behaviour :)
My advise here is 1) to disable notifications for non interesting values,
setting the "Interesting false" keyword into the threshold block, and 2)
increase the missing timeout with the Timeout variable.
> ...and thought that it might make sense to differentiate between
> known bad and unknown value -- maybe declaring missing ones as
> MISSING and not FAILURE.
I'm not very sure about how usefull could be this change, but anyway I
attach a patch which implements this feature. The patch is for version
in master branch, due to it uses the new (and great) threshold missing
dispatcher, added by Florian in commit e595975, and, of course, it's
not much tested yet ;)
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 2748 bytes
Desc: not available
More information about the collectd