This plugin does not use the standard Munin SNMP plugin interface, and
is, nowadays, duplicating the functionality present in the main
snmp__if_ plugin, which uses the 64-bit counters by default if
present, and fall back to the 32-bit ones if not.
Instead of just having a single "errors" value on the per-interface
error graphs, split it in errors and discards, then provide a total
for the combined graph.
This way it's possible to see more details of what's going on at the
interface level.
This basically almost replaces snmp__if_multi in full, as sub-graphs
with interface traffic are generated as a set of detailed graphs. It
doesn't yet compose graphs with the interface errors though.
This plugin uses a single requests to parse all the statistics instead
of making three of them, reducing the overhead, and uses the
multigraph capabilities to generate the same graphs. The naming is
compatible so that it should be a clean update from the previous
plugins.
This also merges in apache_average_requests (and apache_request_rate)
which, albeit not as relevant, also parsed the same file.
Note that the configuration file syntax has changed. You need one "url"
variable per URL now, numbered starting with 1, e.g. url1, url2, url3.
Each one can be configured by suffixing its variable, e.g. url1_proxy,
url2_proxy, etc.
This new plugin is written in Perl rather than sh+awk, executes
ipmi-sensors only once per call (config/fetch still require two calls
total), and with one call it produces all the available graphs
depending on the available sensors.
This still allows for monitoring a remote system, and that becomes
even more interesting since it only fetches data once.
The plugin will also now fail autoconf if there is no sensor output
that can be monitored.
This is designed as a replacement for the sensors_ plugin in Munin; it
accesses the same data as that plugin, but does so through direct
access to /sys, following the kernel-mandated interface, instead of
using the `sensors` command.
This allows much faster and reliable monitoring, especially on
dme1737-based systems where the driver is prone to failures.
Compared to the sensors_ plugin, this does not let you change the
warning/critical levels from the node configuration, nor to ignore any
particular sensor, as it provides whatever the kernel is providing.
Also, this plugin will show humidity and current readings, if present.
