[collectd] Notifications Flood in plugins

[Pierre Baillet]

Hello,

Following-up more or less what has been said in
http://mailman.verplant.org/pipermail/collectd/2011-March/004380.html one
year ago :). I'm having the exact same problem here.

I originally wanted to integrate Russell Poyner plugin (cf
http://mailman.verplant.org/pipermail/collectd/2012-August/005310.html )
but ended up facing some issues (a simple run and control-C of my collectd
resulted in regular segfault in the perl engine).

At the same time I saw that my nagios was flooded by notifications and
started suspecting nagios of being unable to play correctly the passive
checks.

So as to investigate, I took some already working python plugin I use and
added the notification callback. In the callback, I simply log the
notification in the syslog (
http://uu.zoy.org/v/butayo#clef=ogpmzdefiwczkfkf )

With collectd running in debug (foreground in console), I see 1
"Dispatching", one notification arriving in my plugin. And 10 seconds later
(my regular collectd interval, syslog proudly announces there has been
43297 others notifications (cf
http://uu.zoy.org/v/gagima#clef=papztgesfesjueyl ).

The corresponding threshold sections look like this:

<Threshold>
  <Type pending>
        WarningMax 5
    FailureMax 10
  </Type>
</Threshold>

I'm not using persist because I think the default is correct (ie not
supposed to send additional notifications if the state has just changed).

Am I doing something wrong here ?

Thank you very much :)
-- 
Pierre Baillet <pierre at baillet.name>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.verplant.org/pipermail/collectd/attachments/20120912/62727401/attachment.html>