Thanks for reporting this issue !
According to your config file, you don't have the WriteQueueLimit{Low,High}
global options enabled. In the case where this issue would be caused by one
of the write plugins not being able to keep up with the data collection
rate, collectd would indeed buffer up values forever. So one lead would be
to set both of these options to the same value (for ex 5000) and observe
evolution of memory consumption over the same time range as without.
As you suggested, selectively disabling plugins could help narrowing down
which one is causing trouble *assuming the problem is in a plugin, not in
the core engine*. While doing this, try starting with the write plugins
(network, graphite and rrdtool) and consider setting a much lower global
Interval (such as 0.1 instead of the default 10). This should make the
memory allocation functions run more often, and the leak to happen faster.
The fact that this issue appears after upgrading from 5.6 is clearly
something that warrants investigation, thanks for your help !
Marc
