[collectd] Change network plugin reporting interval
jan-jonas.saemann at saenet.de
Sun Jul 5 21:32:50 CEST 2015
I have the same problem.
Also trying to monitor a _small_ set of data.
Can't get the network Plugin to work properly with only one value to be
I followed the Guide https://collectd.org/wiki/index.php/Troubleshooting
In my case rrd's getting created on the server but updated every 3-4
minutes without a graph displayed.
My client collectd.conf:
> Hostname "***"
> FQDNLookup true
> TypesDB "/usr/share/collectd/types.db"
> Interval 5
> LoadPlugin syslog
> LoadPlugin network
> LoadPlugin exec
> <Plugin exec>
> Interval 5
> Exec "pi:pi" "/bin/bash" "/etc/collectd/test.sh"
> <Plugin network>
> Server "10.0.0.1"
> while true; do
> echo "PUTVAL ***/env/gauge-test interval=5 N:2"
> sleep 5
test.sh is running, no errors at syslog on client side.
If I use tcpdump on client side like it's described at the guide it
flushes every 3-4 minutes a huge
block containing all informations to the server at once.
So all updates to one rrd in between the 3-4 minutes get transferred to
the server simultanously. None of them get saved to the rrd on the
server because: (servers log)
> [2015-07-05 18:14:26] rrdtool plugin: rrd_update_r
/var/cache/rrd/***/env/gauge-test.rrd: illegal attempt to update using
time 1436112865 when last update time is 1436112865 (minimum one second
I think there is a bug with output buffers inside the network plugin.
Changing the Interval didn't affect this behavior at all.
I'am trying to fix it by transmitting useless stuff too by enabling
serveral unneeded plugins.
I think the buffer gets filled faster with different values wich speeds
up the flush rate and pads the values to prevent two updates to one rrd
in one request...
It really looks like the way to deal with and proves my theory with
buffering issue inside the network plugin.
More information about the collectd