[collectd] Having problems with notifications

[Brian Long]

Hi,

I'm having some problems getting notifications to work consistently or
sometimes work at all. Every machine in our cluster runs collectd and sends
its data via the network plugin to a central master server. The master
server writes the data using rrd and also contains the <threshold/>
definitions for alarming. An interval of 30 is used everywhere at the moment
(10 was thrashing the master disk too much). The cluster runs a mixture of
4.5.x to 4.6.x and the master is running 4.6.2.

In this particular instance, one of the machines runs a custom <exec> plugin
that reports on a queue backlog size. The data is definitely making it to
the server becaues the RRD graph shows a nice fluctuation of queue size
(typically 0 and 10). On the central server I have the following definition:

<Threshold>
  <Type "current">
     WarningMin    0.00
     WarningMax    1.00
     FailureMin    0.00
     FailureMax    1.00
     DataSource    "value"
     Invert        false
     Persist       true
     Instance      "FetchQueueCount"
  </Type>
</Threshold>

The queue fluctuates between full and empty quite often, so this alarm ought
to be getting triggered and then okay and so on repeatedly. Instead, I have
never seen this alarm on a value range problem. Once in a long while it will
alarm on not receiving data, but never on data out of threshold.

The notification exec looks like:

<Plugin exec>
        NotificationExec "dotspots:dotspots"
"/dist/collectd/bin/notify-exec.rb"
</Plugin>

I have a logfile running on the master server and it is not indicating any
errors, and I don't see any info about notifications in the log (except when
I get a failure due to to the occasional incident of data not being
received.)

Am I doing something wrong?

Thanks,
Brian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.verplant.org/pipermail/collectd/attachments/20090415/06c8c70c/attachment.htm