[collectd] Network plugin - ensuring delivery and local caching during network outages

[Spike Spiegel]

Hi,

we're currently reviewing collectd as an alternative to ganglia for
our infrastructure and ran into a problem with the way data is handled
during network outages and collectd is configured to use the network
plugin. Basically what we're after is a local cache for the duration
of the outage so metrics can be stored on the local node and then
replayed once the connection is re-established. Given that I'm new to
collectd I'd like to hear from people recommendations on how to
implement this feature, either as an option to the network plugin or
maybe as something entirely different (output to rrd or something and
then resend via the network plugin if any data is found?

I'd also be interested to hear opinions on a tcp based implementation
in case the user wanted delivery to be guaranteed. I appreciate the
overhead of TCP, especially for high sampling rates, but if designed
to be an option the user can turn on or off I believe it could be
useful for important nodes where data loss is not acceptable.

thanks for your input,

Spike

-- 
"Behind every great man there's a great backpack" - B.